Assessment of PET imaging devices: the case of a LSO/NaI PET-SPECT

Assessment of PET imaging devices:

the case of a LSO/NaI PET-SPECT prototype

070168 Urvi Joshi.indb 1070168 Urvi Joshi.indb 1 21-05-2007 15:14:5121-05-2007 15:14:51


VRIJE UNIVERSITEIT

Assessment of PET imaging devices:

the case of a LSO/NaI PET-SPECT prototype

ACADEMISCH PROEFSCHRIFT

ter verkrijging van de graad van Doctor aande Vrije Universiteit Amsterdam,

op gezag van de rector magnifi cus prof.dr. L.M. Bouter

in het openbaar te verdedigenten overstaan van de promotiecommissie

van de faculteit der Geneeskundeop maandag 25 juni 2007 om 13:45 uur

in de aula van de universiteit, De Boelelaan 1105

door Urvi Joshi

geboren te Lisburn, Noord-Ierland


Promotor: prof.dr. O.S. HoekstraCopromotor: dr.ir. A. van Lingen


Contents

Chapter 1 Introduction and thesis outline 7

Chapter 2 Clinical lesion detectability predicted by a simple phantom study: 19 an observer study of thorax phantom images of fi ve diff erent PET scanners

Chapter 3 In search of an unknown primary tumour presenting with extracervical 35 metastases: the diagnostic performance of FDG-PET

Chapter 4 Attenuation corrected versus non-attenuation corrected FDG-PET 49 in oncology, a systematic review

Chapter 5 Initial experience with a prototype dual crystal (LSO/NaI) dual head 63 coincidence camera in oncology

Chapter 6 Evaluation of pulmonary nodules: concordance of a prototype dual crystal 71 (LSO/NaI) dual head coincidence camera and full ring positron emission tomography (PET)

Chapter 7 Evaluation of new imaging tests suited for triage in oncology: the case 81 of a prototype dual crystal (LSO/NaI) dual head coincidence camera and mediastinal staging of non-small cell lung cancer (NSCLC)

Chapter 8 Summary and future directions 95

Chapter 9 Nederlandse samenvatting 103

Dankwoord/Acknowledgements/Remerciements 117

Abbreviations 123



C h a p t e r

Introduction and outline of the thesis

1




9

Positron emission tomography (PET) is based on imaging photons emitted during positron annihilation. Most positron emitters applied in medicine are produced in cyclotrons, which consist of large magnets where protons (positively charged nuclear particles) are accelerated to high velocities. These high speed protons are subsequently used to bombard clusters of target atoms, which are rendered unstable, i.e. radioactive. These excited, radioactive atoms regain their natural stability by emitting positrons, which are positively charged electrons that travel only a short distance (order of a few millimeters) in tissue before they give up kinetic energy and recombine with free negative electrons. As a result of this recombination, termed annihilation, the masses of the two particles are converted into energy corresponding to that of two coincident 511 keV photons, each emitted in a direction 180 degrees relative to the other.

Currently, the most commonly used PET tracer for oncological imaging is the glucose analog 18F-fl uorodeoxyglucose (FDG). The use of FDG is based on the early observation of Warburg in the 1930s that tumour cells exhibit increased glucose metabolism (glycolysis) as compared to normal cells (1). Unlike glucose, FDG remains trapped in the cell after it is taken up and does not undergo further metabolism. FDG uptake by tumours in animals was subsequently demonstrated by Som et al (2). Since many types of cancer cells have a high affi nity for FDG, which has a favourable biodistribution, it emerged as a highly eff ective tool for whole body staging of cancer. Furthermore, it appears that therapeutic cytotoxic and cytostatic agents can directly or indirectly aff ect the pathways, glucose transporters, and metabolic enzymes controlling glycolysis. Keeping pace with the advancement of our understanding of tumour biology, other PET tracers have been and are being developed to evaluate other fundamental aspects of carcinogenesis (cancer cell formation), such as proliferation, angiogenesis (blood vessel formation), hypoxia (lack of oxygen) and apoptosis (programmed cell death). Therefore, these molecular PET tracers hold great future promise in assessing disease status and directing cancer therapy (3).

Scanner evolutionThe fi rst positron detection device for localizing brain tumors was described by Bronwell and Sweet in 1953 (4). This was followed by the construction of a device to measure regional cerebral blood fl ow with PET tracers at the Brookhaven National Laboratory in 1966, which was later moved to the Montreal Neurological Institute. The concept of tomographic imaging was then introduced in the late 1960s and early 1970s. It was based on the placement or rotation of de-tection systems around the subject. Mathematical algorithms were applied to the collected data to reconstruct images of selected two-dimensional slice planes within the subject. The fi rst rudi-mentary tomographic PET imaging was performed by Kuhl and Edwards in 1968 (5). Following this, Ter-Pogossian et al developed a device with signifi cantly better image quality using fi ltered backprojection reconstruction principles in 1975 (6;7). An example of an early brain tomographic imaging device is demonstrated in Figure 1. Early devices were capable of only brain imaging due to the limited diameter of the detector ring, related to the high cost of the detector crystals and associated electronics. Advances in the manufacturing of detector crystals and faster electronics (to convert light signals to electronic signals and for data processing), in addition to the need for oncologic imaging resulted in the development of devices capable of performing whole body imaging at an acceptable cost in the late 1980s. These devices were only capable of


Chapter 1

10

imaging in 2D (dimensions) mode. Devices capable of imaging in 3D mode became commerci-ally available in the late 1990s. In 2D mode, thin septa of lead or tungsten separate each crystal detector ring and coincidence photons are only recorded between detectors within the same ring or lying in closely neighbouring rings. In 3D mode, the septa are removed, and coincidence photons are recorded between detectors lying in any ring combination. As a result, devices capable of imaging in 3D mode demonstrate higher sensitivity as compared to 2D mode.

Figure 1: Early PET imaging device (reprinted from Wagner HN, Semin Nucl Med 1998; 28: 213-220(7)).



11

FDG imaging with gamma camerasDuring the late 1980s and early 1990s, FDG-PET gradually gained acceptance in the scientifi c literature as an important diagnostic tool in the assessment of various malignancies. However, the high costs associated with PET imaging impaired its widespread use. As a result, low-cost alternatives suited for imaging both the high energy photons of positron decay and the lower energy photons of routine nuclear medicine were sought and hybrid cameras were developed. Initially, these cameras consisted of routine nuclear medicine gamma cameras fi tted with ultra high energy collimators capable of imaging the high energy photons of positron decay. Planar imaging was initially performed (8), followed by single photon emission computed tomographic imaging (SPECT) in which only one of the 511 keV photons is used (9-12). This technology appeared promising, especially in cardiology applications examining myocardial viability. (In fact, during that same period entire plenary sessions at international nuclear medicine congresses were dedicated to FDG imaging with ultra high energy collimators). The technology also appeared to have potential in oncological applications for response monitoring and the characterization of suspected lesions larger than 2 cm. However, the sensitivity for smaller lesions was poor, rendering the technique less suitable for staging purposes.

Since the cost and limited availability of full ring PET remained a problem, the search for cheaper alternatives continued and the next evolution appeared in the mid 1990s using dual head SPECT systems working in coincidence mode and without collimators to detect the two simultaneous 511 keV photons released in positron decay (13-16). The obvious advantage of these coincidence cameras (henceforth termed dual head PET) was their higher sensitivity as compared to cameras using ultra high energy collimators while still being able to perform routine nuclear medicine imaging, the latter of which resulted in a signifi cant cost saving.

Crystal technologyAt the same time, the use of various detector crystals was also investigated. Initially, bismuth germanate (BGO) crystals were used in the majority of dedicated full ring PET scanners while sodium iodide (NaI) crystals were used in routine nuclear medicine gamma cameras. In the late 1980s and early 1990s, new crystals such as lutetium oxyorthosilicate (LSO) (17) and gadolinium oxyorthosilicate (GSO) (18) were introduced. Unfortunately, NaI is suboptimal for positron imaging, largely due in part to its signifi cantly lower density, resulting in markedly lower “stopping power” for the higher energy photons released in positron decay (19). Suboptimal initial attempts to at least partially compensate for this problem in dual head PET cameras involved using thicker NaI crystals (5/8 inch crystals in place of 3/8 inch crystals (20). To further address this problem, a new type of dual head PET camera consisting of a dual crystal layer (termed a phoswich crystal) was developed. The fi rst model consisted of LSO and yttrium orthosilicate (YSO) crystals (21). This was followed by development of a LSO/NaI phoswich detector (LSO PET-SPECT) (22), the evaluation of which forms the basis of much of this thesis. A potential advantage of the addition of the LSO crystal was its higher density as compared to NaI enabling signifi cantly faster acquisition times than other dual head PET cameras on the market that utilized only NaI crystals. Subsequently, the newer crystals such as LSO and GSO were also incorporated into full ring PET cameras (23;24). Table 1 compares the physical properties of diff erent scintillator crystals (25).


Chapter 1

12

Table 1. Diff erent properties of scintillator crystals

Crystal Density

(g/cm3)

Atomic

Number

Hygro-

scopic

Rugged Primary

Decay

Constant

(ns)

Secondary

Decay

Constant

(ns)

Relative

Emission

Intensity

LSO 7.40 65 No Yes 40 75

GSO 6.71 59 No

No

(cleaves

easily)

60 600 30

BGO 7.13 75 No Yes 300 ~10 000 15

NaI(Tl) 3.67 51 Yes No 230 100

Source: Melcher, JNM 2000

Image reconstruction developmentsFinally, image reconstruction software was also continuously being improved. A common problem encountered in PET imaging is the scatter or absorption of photons originating from the patient prior to detection by the camera, resulting in non-detection or detection at a diff erent location, illustrated in Figure 2.

Figure 2. Schematic illustration of scatter and random counts

The end result is distortion of the true radioactivity distribution. Attenuation correction methodologies have been developed to correct for absorption eff ects in both full ring and dual head PET cameras (26;27). However, applying attenuation correction is associated with both advantages (such as the ability to perform quantifi cation) and disadvantages (such as increased statistical noise) and the added value in terms of increased lesion detection has been the subject of debate (28). In addition, performing attenuation correction with dual head PET cameras is further complicated by the relatively low density of the sodium iodide detector crystals and the suboptimal positioning of attenuation correction sources on the two detector heads compared



13

to a rotating ring of attenuation sources used on full ring PET cameras. The transmission scans obtained for attenuation correction are also used to quantify and correct for scatter. However, these techniques require a signifi cant amount of processing time. Despite the continual development of faster and more effi cient computer data processing systems, these techniques for scatter correction have not yet been incorporated into routine clinical imaging.

The challengeSince 1990, the demand for positron emission tomography has increased enormously for research as well as clinical applications. At the same time, however, there is concern about the impact on the overall cost of health care. Resources are limited and should be distributed among diagnostics and therapeutics. Notably, in the past few years, new classes of costly ‘targeted therapeutics’ have been developed, sometimes with huge success, thus necessitating their implementation in patient care. Therefore, as with therapeutic interventions, the introduction of new imaging tests also requires careful validation and implementation. To this end, we and others have developed hierarchical frameworks to evaluate the cost-eff ectiveness of imaging in a stepwise approach typically including a randomized clinical trial (29).

As shown above, refi ning the result of imaging tests (i.e. hardware and image reconstruction software) is a continuous process. Often, small progression prevails over actual paradigm shifts. However, with the introduction of PET-CT devices (30;31), it appears that going through such a cycle of evaluation may take too much time to keep pace with technological developments. The challenge for methodologists and clinicians is to develop intelligent and effi cient strategies of test evaluation such that randomized trials are necessary only in exceptional cases. In this thesis, such a strategy using the case of a prototype LSO/NaI PET-SPECT camera is explored.

Thesis Outline

The focus of this thesis is the performance evaluation of the prototype dual crystal (LSO/NaI) dual head PET camera (LSO-PS) versus the present standard of a full ring BGO-PET scanner. The evaluation is based on a methodology which attempts to rapidly yet eff ectively assess the performance of a new imaging device as compared to others that are already in clinical use.

Observational studies that provide the highest level of evidence measure test accuracy versus a pathological gold standard, and are followed, if necessary, by randomized controlled trials using patient outcomes as clinically relevant endpoints. The experience with FDG-PET for solitary pulmonary masses/nodules and for mediastinal lymph node staging has clearly demonstrated that performing appropriate observational accuracy studies is extremely diffi cult and that the literature is fraught with methodological shortcomings. Gould et al. performed two systematic reviews on these topics. In the fi rst review, the methodological quality of the evidence from 40 studies on pulmonary masses/nodules was classifi ed as only ‘fair’. This fi rst review included a large number of patients (n=1474) and covered the time span of a whole decade [1990-2000; (32)]. Later, the same group evaluated 32 studies for mediastinal lymph node staging of non-small cell


Chapter 1

14

lung cancer. Again, this review evaluated a large number of patients (n=1959) and a long time frame [8 years: 1994-2002; (33)]. They found that systematic sampling of mediastinal lymph nodes (the reference standard) had been done in only 47% of the studies, that only 51% of the studies been prospective, and that only 56% of the studies had adequate blinding of image interpreters. This highlights the fact that even in relatively simple clinical problems such as evaluating and staging a pulmonary lesion, serious methodological fl aws prevail (34). Therefore, alternative strategies should be considered.

Central to this evaluation methodology for the prototype camera is the consensus in the published literature that dual head PET (DH-PET) cameras have an inherently lower sensitivity than full ring PET (FR-PET), largely related to geometry limitations. Hence, it was anticipated that dual head PET cameras might be most useful in a setting of ‘triage’ for FR-PET. For example, patients with a radiologically indeterminate pulmonary coin lesion that is negative with FR-PET can generally be subjected to a watchful waiting policy. If DH-PET would have a comparable diagnostic accuracy to FR-PET, this expectative policy might also be followed after DH-PET. Patients with a positive lesion at DH-PET might require additional FR-PET for further staging unless, in the best case scenario, it would be determined that the positive predictive value of DH-PET be similar to that of FR-PET. In that case, only FDG avid pulmonary lesions with a negative mediastinum at DH-PET might require FR-PET to compensate for the expected limited sensitivity of DH-PET. It was reasoned that since technology is evolving rapidly, a set of in vivo “tracker trials” would be most helpful to provide a fi rst approximation of the relative merits of the DH-PET scanners.

As the diff erence in scanner sensitivity is the key variable, a head-to-head comparison with FR-PET is an attractive study design. In a systematic review by our group, (35), 9 studies were identifi ed and assessed according to their methodological quality (Cochrane criteria), and relevant data were extracted. The studies comprised a variety of tumours and indications. The reported full ring and dual head PET camera agreement for detection of malignant lesions ranged from 55 to 100%. We identifi ed several methodological limitations (blinding, standardization, limited patient spectrum). Mean lesion diameter was 2.9 cm (SD 1.8), with only about 20% of lesions measuring < 1.5 cm. The 3 studies with the highest quality reported concordances of 74-79% for the studied lesion spectrum. Contrast at dual head PET was lower than that of full ring PET, and contrast and detection agreement were positively related. At that time, it was concluded that in spite of methodological limitations, “fi rst generation” DH-PET devices detected suffi cient FR-PET positive lesions to allow prospective evaluation in clinical situations where the impact of FR-PET is not confi ned to detection of small lesions (< 1.5 cm). Furthermore, it was concluded that the effi ciency of head-to-head comparative studies would benefi t from application in a clinically relevant patient spectrum, with proper blinding and standardization of acquisition procedures.

As such, a choice was made to perform a comparative head-to head assessment of DH-PET versus FR-PET, comparing the LSO-PS camera with full ring BGO-PET as the gold standard. In addition, an attempt was made to target smaller lesions where dual head PET cameras have demonstrated the lowest sensitivity.



15

In vitro evaluation This approach begins with an in vitro analysis. Obviously, a comprehensive in vitro analysis might be quite useful in evaluation of test results obtained in diff erent cameras. The original NEMA (National Electrical Manufacturers Association, 1994) standards for acceptance testing of PET scanners were too limited for assessment of whole body 3D PET scanners. As a result, in 2002 new NEMA standards were introduced. However, these standards still fail to provide an adequate simulation of the routine clinical setting to enable conclusions to be drawn over the true image quality in actual patient settings. Furthermore, neither of these NEMA standards (i.e. 1994 and 2002) could be applied to dual head PET systems. In vitro testing must strive to replicate the actual clinical setting as much as possible. To this end, a phantom was used to simulate patients as closely as possible (e.g. similar tissue thickness, similar activity distribution, etc.). In addition, the phantom was designed so that it could be reproducibly used (i.e. fi lled) to enable comparative studies between diff erent cameras.

Chapter 2 outlines an attempt to relate lesion detectability to image contrast and noise for fi ve diff erent PET scanners that incorporate four diff erent detector crystals utilizing a thorax phantom with subsequent comparison of results to clinical data. An image acquisition protocol was designed for application to diff erent cameras in diff erent nuclear medicine departments. In addition, an observer study that minimizes observer bias was also designed. Specifi cally, the ability to predict the in vivo performance of the prototype LSO-PS camera as compared to full ring BGO-PET was examined.

An attractive feature of the prototype LSO-PS scanner is its suitability as a whole body scanner which was not feasible with previous dual head coincidence systems. A typical clinical setting in which this feature would be important is the patient presenting with a metastasis from an unknown primary tumour. In Chapter 3, the yield of full ring BGO-PET in patients presenting with extracervical metastases of unknown primary tumours is reported. In conjunction with the results of the thorax phantom study, the results of this study were used to estimate the potential yield of LSO-PS, primarily as a function of tumour size. So far, in the clinical oncology setting, size has been the limiting factor with dual head PET and FDG.

Since we had noted in the literature and from our own experience that the impact of attenuation correction on lesion detection was controversial, a systematic review was performed and outlined in Chapter 4 to compare the accuracy of FDG-PET with and without attenuation correction, for both full ring PET and dual head PET.

In vivo evaluationIn the second part of this thesis, an in vivo assessment of the LSO-PS camera was performed. In Chapter 5, an initial feasibility evaluation of the LSO-PS is described where its lesion detection sensitivity in an unselected group of patients with diff erent malignancies was examined. In Chapters 6 and 7, the performance of LSO-PS in specifi c clinical indications such as the evaluation of indeterminate pulmonary nodules and the mediastinal staging of non-small cell lung cancer (NSCLC) was investigated.


Chapter 1

16

References

(1) Warburg O. The Metabolism of Tumors. New York, NY: Smith Press, 1931.

(2) Som P, Atkins HL, Bandoypadhyay D, Fowler JS, MacGregor RR, Matsui K et al. A fl uorinated glucose analog,

2-fl uoro-2-deoxy-D-glucose (F-18): nontoxic tracer for rapid tumor detection. J Nucl Med 1980; 21(7):670-

675.

(3) Kelloff GJ, Krohn KA, Larson SM, Weissleder R, Mankoff DA, Hoff man JM et al. The progress and promise of

molecular imaging probes in oncologic drug development. Clin Cancer Res 2005; 11(22):7967-7985.

(4) Bronwell GL, Sweet WH. Localization of brain tumors. Nucleonics 1953; 11:40-45.

(5) Kuhl DE, Edwards RQ. Reorganizing data from transverse section scans of the brain using digital processing.

Radiology 1968; 91(5):975-983.

(6) Ter Pogossian MM, Phelps ME, Hoff man EJ, Mullani NA. A positron-emission transaxial tomograph for

nuclear imaging (PETT). Radiology 1975; 114(1):89-98.

(7) Wagner HN, Jr. A brief history of positron emission tomography (PET). Semin Nucl Med 1998; 28(3):213-220.

(8) van Lingen A, Huijgens PC, Visser FC, Ossenkoppele GJ, Hoekstra OS, Martens HJ et al. Performance

characteristics of a 511-keV collimator for imaging positron emitters with a standard gamma-camera. Eur J

Nucl Med 1992; 19(5):315-321.

(9) Hoekstra OS, Ossenkoppele GJ, Golding R, van Lingen A, Visser GW, Teule GJ et al. Early treatment response

in malignant lymphoma, as determined by planar fl uorine-18-fl uorodeoxyglucose scintigraphy. J Nucl Med

1993; 34(10):1706-1710.

(10) Hoekstra OS, van Lingen A, Ossenkoppele GJ, Golding R, Teule GJ. Early response monitoring in malignant

lymphoma using fl uorine-18 fl uorodeoxyglucose single-photon emission tomography. Eur J Nucl Med

1993; 20(12):1214-1217.

(11) Macfarlane DJ, Cotton L, Ackermann RJ, Minn H, Ficaro EP, Shreve PD et al. Triple-head SPECT with 2-

[fl uorine-18]fl uoro-2-deoxy-D-glucose (FDG): initial evaluation in oncology and comparison with FDG PET.

Radiology 1995; 194(2):425-429.

(12) Martin WH, Delbeke D, Patton JA, Sandler MP. Detection of malignancies with SPECT versus PET, with 2-

[fl uorine-18]fl uoro-2-deoxy-D-glucose. Radiology 1996; 198(1):225-231.

(13) Boren EL, Jr., Delbeke D, Patton JA, Sandler MP. Comparison of FDG PET and positron coincidence detection

imaging using a dual-head gamma camera with 5/8-inch NaI(Tl) crystals in patients with suspected body

malignancies. Eur J Nucl Med 1999; 26(4):379-387.

(14) Delbeke D, Patton JA, Martin WH, Sandler MP. FDG PET and dual-head gamma camera positron coincidence

detection imaging of suspected malignancies and brain disorders. J Nucl Med 1999; 40(1):110-117.

(15) Shreve PD, Steventon RS, Deters EC, Kison PV, Gross MD, Wahl RL. Oncologic diagnosis with 2-[fl uorine-

18]fl uoro-2-deoxy-D-glucose imaging: dual-head coincidence gamma camera versus positron emission

tomographic scanner. Radiology 1998; 207(2):431-437.

(16) Zimny M, Kaiser HJ, Cremerius U, Sabri O, Schreckenberger M, Reinartz P et al. F-18-FDG positron imaging

in oncological patients: gamma camera coincidence detection versus dedicated PET. Nuklearmedizin 1999;

38(4):108-114.



17

(17) Melcher CL, Schweitzer J.S. Cerium-doped lutetium oxyorthosilicate: a fast, effi cient new scintillator. IEEE

Trans Nucl Sci 39, 502-505. 1992.

(18) Holte S, Ostertag H, Kesselberg M. A preliminary evaluation of a dual crystal positron camera. J Comput

Assist Tomogr 1987; 11(4):691-697.

(19) Budinger TF. PET instrumentation: what are the limits? Semin Nucl Med 1998; 28(3):247-267.

(20) Delbeke D, Sandler MP. The role of hybrid cameras in oncology. Semin Nucl Med 2000; 30(4):268-280.

(21) Dahlbom M, MacDonald LR, Eriksson L. Performance of a YSO/LSO phoswich detector for use in a PET/

SPECT system. IEEE Trans Nucl Sci 1997; 44:1114-1119.

(22) Schmand M, Dahlbom M, Eriksson L, Casey ME, Andreaco MS, Vagneur K et al. Performance of a LSO/NaI(Tl)

phoswich detector for a combined PET/SPECT imaging system. J Nucl Med [39], P9. 1998.

(23) Erdi YE, Nehmeh SA, Mulnix T, Humm JL, Watson CC. PET performance measurements for an LSO-based

combined PET/CT scanner using the National Electrical Manufacturers Association NU 2-2001 standard. J

Nucl Med 2004; 45(5):813-821.

(24) Surti S, Karp JS. Imaging characteristics of a 3-dimensional GSO whole-body PET camera. J Nucl Med 2004;

45(6):1040-1049.

(25) Melcher CL. Scintillation crystals for PET. J Nucl Med 2000; 41(6):1051-1055.

(26) Bai C, Kinahan PE, Brasse D, Comtat C, Townsend DW, Meltzer CC et al. An analytic study of the eff ects of

attenuation on tumor detection in whole-body PET oncology imaging. J Nucl Med 2003; 44(11):1855-1861.

(27) Turkington TG. Attenuation correction in hybrid positron emission tomography. Semin Nucl Med 2000;

30(4):255-267.

(28) Wahl RL. To AC or not to AC: that is the question. J Nucl Med 1999; 40(12):2025-2028.

(29) Evaluating positron emission tomography in non-small-cell lung cancer: moving beyond accuracy to

outcome. H. van Tinteren, OS Hoekstra, CA Uyl-De Groot, M. Boers. in: Cancer Imaging: Lung and Breast

Carcinomas. Editor: Hayat-Kean MA. New Jersey.

(30) Herder GJ, Kramer H, Hoekstra OS, Smit EF, Pruim J, van Tinteren H et al. Traditional versus up-front [18F]

fl uorodeoxyglucose-positron emission tomography staging of non-small-cell lung cancer: a Dutch

cooperative randomized study. J Clin Oncol 2006; 24(12):1800-1806.

(31) Lardinois D. New horizons in staging for non-small-cell lung cancer. J Clin Oncol 2006; 24(12):1785-1787.

(32) Gould MK, Maclean CC, Kuschner WG, Rydzak CE, Owens DK. Accuracy of positron emission tomography for

diagnosis of pulmonary nodules and mass lesions: a meta-analysis. JAMA 2001; 285(7):914-924.

(33) Gould MK, Kuschner WG, Rydzak CE, Maclean CC, Demas AN, Shigemitsu H et al. Test performance of

positron emission tomography and computed tomography for mediastinal staging in patients with non-

small-cell lung cancer: a meta-analysis. Ann Intern Med 2003; 139(11):879-892.

(34) Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM et al. The STARD statement for

reporting studies of diagnostic accuracy: explanation and elaboration. Ann Intern Med 2003; 138(1):W1-12.

(35) Haslinghuis-Bajan LM, Hooft L, van Lingen A, van Tulder M, Deville W, Mijnhout GS et al. Rapid evaluation

of FDG imaging alternatives using head-to-head comparisons of full ring and gamma camera based PET

scanners--a systematic review. Nuklearmedizin 2002; 41(5):208-213.



C h a p t e r

2Lesion detectability predicted by a simple phantom study: an observer study of thorax phantom images of fi ve diff erent PET scanners

Urvi JoshiArthur van LingenHugo WAM de JongGerrit JJ TeuleOtto S HoekstraMark Lubberink

Submitted

(orally presented at the 2005 Society of Nuclear Medicine Congress)


Abstract

Purpose: The aim of this study was to relate lesion detectability to image contrast and noise for fi ve diff erent PET scanners, by performing an observer study of thorax phantom images.

Methods: A thorax phantom containing four fi llable spheres with diameters of 10, 13, 17 and 23 mm in the mediastinal region, as well as cold lung, liver and spine inserts, was fi lled with 20 MBq 18F. The activity concentration in the spheres was 2.5, 5 or 10 times the background concentration. Emission and transmission measurements were made at each concentration ratio with each of the scanners using routine clinical acquisition settings. Images were reconstructed using routine clinical protocols both with and without attenuation correction. Images were evaluated by ten observers and either rated positive or negative. Actual image contrasts and image noise were calculated for each scanner and sphere to background ratio and related to lesion detectability.

Results: Total lesion detectability for all sphere sizes and contrasts ranged from 35% to 70%, using attenuation corrected images. Lesion detectability was lower for images without attenuation correction. Lesion detectability for each scanner correlated well with contrast to noise ratio averaged over all lesions and sphere to background ratios (r2 0.90).

Conclusion: Overall, lesion detectability was signifi cantly better for attenuation corrected images and 3D acquisitions. Lesion detectability was predicted well by contrast to noise ratios in the images.


Lesion detectability predicted by a simple phantom study

21

Introduction

There is an increasing variation in the technology behind commercially available PET scanners, considering detector materials, acquisition electronics and reconstruction methods. A few years ago all available scanners consisted of bismuth germanate oxide (BGO) crystals arranged as block detectors, detector rings were separated by removable septa, transmission imaging was done using rotating 68Ge rod sources, and image reconstruction was done primarily using fi ltered back projection. This led to small diff erences between PET systems and made comparison of diff erent PET scanners fairly straightforward. New detector materials such as lutetium oxyorthosilicate (LSO), lutetium-yttrium oxyorthosilicate (LYSO) and (zirconium-doped) gadolinium oxyorthosilicate (GSO) are now commonly used in clinical PET in addition to BGO. Due to their shorter light decay constants and higher relative light emission (1), these crystals have better time and energy resolutions than BGO, leading to better discrimination between true coincidences and randoms and scatter coincidences. The improved timing characteristics of these crystals allow for higher count rates, and thus administration of larger amounts of radioactivity and shorter scanning times, and even time-of-fl ight acquisition.

The changes in scanner characteristics and reconstruction methods may also aff ect the correspondence between technical scanner performance and image quality in terms of lesion detectability. PET scanner performance is most often evaluated using the NEMA NU2-2001 protocol of The National Electrical Manufacturers Association (2). The NEMA NU2-2001 protocol is more adapted to clinical whole-body PET than its predecessor from 1994, but still focuses very much on the technical aspects of the performance of the scanner. The protocol does include an image quality measurement, used by Bergmann and co-workers to compare eight dedicated PET scanners and seven gamma camera systems capable of PET imaging installed in Austria (3). A more clinical comparison evaluating lesion detectability however, using acquisition and reconstruction settings routinely used in clinical practice, is not included in the protocol. While in vivo direct comparison between diff erent cameras is not feasible, it is hoped that an evaluation of lesion detectability using phantom images, acquired under routine clinical settings, can be related to measured image contrasts and will help to predict in vivo performance. The NEMA image quality phantom is not suitable for such lesion detectability observer studies because of its regular shape and predictable sphere locations. Several published studies have compared lesion detectability for diff erent acquisition or reconstruction modes of a single scanner, or compared a number of scanners (4-12), but no systematic comparison between lesion detectability and image contrast of diff erent scanners was made. As such, the purpose of this study was to relate lesion detectability to measured image contrast and noise by performing an independent, blinded lesion detectability comparison between fi ve diff erent PET scanners using an anthropomorphic thorax phantom.


Chapter 2

22

Methods

ScannersFive PET systems installed in The Netherlands at the time of this study were included in this comparison: the ECAT Exact HR+ (13;14), the Biograph LSO PET/CT or ECAT Reveal (15), an experimental LSO NaI(Tl) dual-headed PET/SPECT camera (12;16) (all CTI PET Systems, Knoxville, Tennessee and Siemens Medical Systems, Erlangen, Germany), the Allegro (17), and the C-PET Plus (18) (both Philips Medical Systems, Best, The Netherlands). Details about each of the scanners are given in Table 1. The data in Table 1 was retrieved from Tarantola and co-workers (19) and the other publications mentioned in this paragraph. None of the scanner manufacturers was involved in the present study.

Table 1: Scanner technical specifi cations

Allegro Biograph C-PET HR+ PET/SPECT

Crystal GSO LSO NaI(Tl) BGO LSO/NaI(Tl)

# detectors 17864 9216 6 18432 2*

# rings 29 24 n.a. 32 n.a.

# planes 45/90 47 64/128 63 70

FOV (mm) 576 585 576 585 450

Axial FOV (mm) 180 162 256 155 350

Energy window (keV) 410-665 350-650 435-665 350-650 350-650

Coincidence window

(ns)

8 6 8 12 6

Slice thickness (mm) 4 or 2 3.375 4 or 2 2.46 5

Septa material n.a. n.a. n.a. Lead n.a.

Septa dimensions

(mm)

n.a. n.a. n.a. 0.5x65 n.a.

Transmission source 137Cs, point

source

CT 137Cs, point

source

68Ge, 3 rods 68Ge, 5 point

sources

*2 detectors, each consisting of 10080 detector crystals

Data acquisitionAn anthropomorphic torso phantom (Data Spectrum, Chapel Hill, North Carolina) containing cold lung, liver and spine inserts was used, with four fi llable spheres with inner diameters of approximately 10, 13, 17 and 23 mm placed in the mediastinal region of the phantom (Figure 1). A scan protocol was chosen that could be executed on all systems and in all the participating centres, designed such that the radioactivity concentrations and the scan times correspond to those of a typical study of a lung cancer patient, with sphere to background (S/B) ratios based on those specifi ed in the NEMA NU2-2001 image quality measurement.



23

Figure 1. The anthropomorphic torso phantom. The spheres were connected to the spine insert at 3 cm intervals.

The spheres were fi lled from a 500 ml stock solution containing 10 MBq 18F, resulting in a sphere radioactivity concentration of approximately 20 kBq/ml at the start of the measurement. Initially 20 MBq of 18F was added to the background to give a background activity concentration of 2 kBq/ml, resulting in a 10:1 sphere to background (S/B) ratio, and transmission and emission scans were made. After one (110 min) and two half-lives of 18F, 10 MBq 18F was added to the background to retain a background radioactivity concentration of 2 kBq/ml, yielding S/B ratios of 5 and 2.5, respectively, immediately followed by transmission and emission scans. The total activity within the fi eld of view was thus kept constant at approximately 20 MBq. All radioactivity measurements (stock solution and background solutions) were made with the dose calibrators present at each centre, calibrated using the procedures recommended by their manufacturers.

The phantom was always placed approximately central in the axial FOV. Each imaging sequence consisted of 3D, and for the HR+ also 2D, emission scans of 5 min (HR+, Biograph, C-PET), 3 min (Allegro) or 10 min (PET/SPECT) per bed position, as recommended by the manufacturers and used clinically in the participating centres, followed where possible by 3 min (PET/SPECT: 10 min) transmission scans. One bed position was scanned for the HR+, Biograph and PET/SPECT systems, since these all apply a very limited overlap in clinical whole body studies. The use of two bed positions would not have led to increased measurement times in the planes where the spheres were present. For the Allegro and C-PET two bed positions were scanned, since these scanners both apply an overlap of 50% in clinical whole-body scans, so that each part of the patient is scanned twice.


Chapter 2

24

Image reconstructionAll data were reconstructed using the routine clinical protocols, as recommended by the manufacturers. Except for the Allegro data, images were reconstructed using 2D Attenuation Weighted (HR+, Biograph, PET/SPECT) Ordered Subsets Expectation Maximization (OSEM) with 4 (C-PET) or 2 (other scanners) iterations and 16 (HR+) or 8 (other scanners) subsets, combined with Fourier rebinning (FORE) for 3D data. Allegro data were reconstructed by the fully 3D Row-Action Maximum Likelihood Algorithm (RAMLA) with default settings. A Hanning fi lter with a full width at half maximum (FWHM) of 4 mm (HR+) or 5 mm (Biograph) was used. Images were reconstructed both without (NAC) and with (AC) attenuation correction based on transmission measurements with the sources specifi ed in Table 1. Segmented AC was used on all scanners except the HR+. AC images were corrected for scatter using a simulation based method (HR+ 3D, Biograph) (20;21), a convolution subtraction method (HR+ 2D) (22), or a non-uniform sinogram background subtraction (Allegro, C-PET). NAC images were not corrected for scatter except for HR+ 2D images, for which convolution subtraction scatter correction is standard in the reconstruction software. Randoms correction was based on a delayed measurement for all scanners except the Allegro and C-PET, were it was included in the non-uniform background subtraction.

Data analysis – image contrastSphere contrast and background variability of the reconstructed images were determined separately using an automated algorithm. For each sphere, a circular ROI with similar diameter as the sphere was drawn in the image plane through the centre of that sphere. A second, ring-shaped ROI with an inner diameter of the sphere diameter plus 15 mm and a width of 7 mm was placed concentrically around this fi rst ROI on the same plane. Image contrast was calculated as follows:

background

backgroundsphere

C

CCcontrast

−= (1)

where Csphere

and Cbackground

are the mean counts per pixel in the sphere and in the background ROI. True contrasts were calculated in a similar way, replacing counts per pixel by the known radioactivity concentrations. Relative contrasts were calculated as measured contrast divided by true contrasts.

Background variability in the immediate surroundings of the sphere was determined as relative standard deviation of the mean pixel value in the ring-shaped ROI around the sphere. For each scanner, sphere size and S/B ratio, contrast to noise ratio (CNR) was calculated as the ratio of image contrast and the mean value of the background variability for the three S/B ratios. Background variation and CNR were only calculated for AC images because a uniform image background is required for a meaningful background variation estimate.



25

Data analysis – observer studyThe reconstructed data of all systems were viewed on the same computer under similar conditions. Viewing software was written in Matlab (The Mathworks, Natick, Massachusetts), showing single image planes containing either no lesion or a single lesion. A number of images containing lesions cut from other planes and randomly placed in the mediastinum of the thorax phantom, were included to minimize observer bias due to the predictable location of the actual spheres. These extra images were not included in the fi nal observer score. A typical dataset for each camera (both AC and NAC images) consisted of 24 original images containing a sphere, 18 simulated images in which the sphere was placed in another location in the mediastinum and 18 images without spheres, sorted in a random order that was the same for each observer. An inverse linear grey scale was used that could be adjusted by the observer for each individual image. Observers were instructed to indicate whether or not they visualized a sphere and if so, to indicate its location. All images were assessed by ten independent observers (six nuclear medicine physicians with several years experience in interpreting clinical PET data, four experienced medical physicists). The fi rst shown data set was repeated at the end of the reading to address learning eff ects. The percentage of correctly localized lesions, or detection sensitivity, per sphere size and contrast level as well as the total detection sensitivity for all four spheres and three contrast levels was calculated for each observer and scanner. The signifi cance of possible diff erences in the total detection sensitivity between scanners was determined using one-tailed paired t-tests.

Comparison to clinical dataTo address the clinical relevance of the observer study, detectability and image contrasts as found for the Biograph scanner in the present study were compared to those extracted from a recent publication by Hashimoto and co-workers on the accuracy of PET for diagnosis of CT-confi rmed solid pulmonary lesions with low standardized uptake values (23). To facilitate this comparison, image contrasts were translated to contrast ratios (CR):

backgroundsphere

backgroundsphere

CC

CCCR

+−

= (2)

For comparison with the detectability measures in the clinical study, detectability in the present work was given a score of 2 if 9 or 10 of the observers found a lesion, 0 if 1 or none found it, and 1 in all other cases. In the clinical study, lesions scoring ‘moderate’ were given a score of 2, ‘faint’ a score of 1, and ‘absent’ a score of 0.


Chapter 2

26

Figure 3. Thorax phantom images without attenuation correction; HR+ 2D (a), HR+ 3D (b), Biograph (c), Allegro (d),

C-PET (e) and PET/SPECT (f ).

Results

Data acquisitionThe actual mean (SD) measured S/B ratios were 2.4 (0.3) for the low contrast measurement, 4.9 (0.5) for the medium contrast measurement and 9.8 (0.7) for the high contrast measurement. Figures 2 and 3 show comparable images of a single plane through the centre of the 17 mm lesion with a mean S/B ratio of 4.9, for all scanners.

Figure 2. Thorax phantom images with attenuation correction; HR+ 2D (a), HR+ 3D (b), Biograph (c), Allegro (d),

C-PET (e) and PET/SPECT (f ).



27

Image contrast

Measured contrasts relative to true contrasts are given in Figure 4. Background variability for each scanner is given in Table 2.

Figure 4. Measured contrasts versus true contrasts for all scanners and lesion sizes.

Table 2. Background variability

Scanner Variability (SD)

Allegro 0.10 (0.02)

Biograph 0.16 (0.03)

HR+, 3D 0.23 (0.06)

HR+, 2D 0.23 (0.05)

C-PET 0.10 (0.03)

PET/SPECT 0.15 (0.03)

Lesion detectabilityFigure 5 shows the detection sensitivity in attenuation-corrected images for all four sphere sizes and three S/B ratios. Total sensitivity for both AC and NAC images is shown in Figure 6. No learning eff ect was observed. The highest lesion detection sensitivity was seen for the Allegro camera, followed by but signifi cantly better than the Biograph (p = 0.02, paired t-test). As would be expected, a higher sensitivity was noted for larger spheres and larger concentration ratios for all camera systems.


Chapter 2

28

Image contrast versus lesion detectabilityFigure 7 shows the relationship between lesion detectability for AC images and CNR for the medium S/B ratio and the larger spheres. A CNR above approximately four always led to a detectability of 90% or higher, that is, the lesion was correctly localized by at least 90% of the observers. A CNR below three never led to 100% detectability. Figure 8 shows the total lesion detectability versus the mean contrast to noise ratio for all scanners. A signifi cant correlation between contrast to noise ratio and lesion detectability was found (r2 0.90, p 0.004). Correlation was similar if only the smallest lesions (10-13 mm) were considered. Correlation between image contrast itself and lesion detectability was not signifi cant (r2 0.56, p 0.09).

Figure 5. Lesion detectability per lesion size and S/B ratio with attenuation correction (10 observers).

Figure 6. Total detection sensitivity with and without attenuation correction (10 observers).



29

Figure 7. Contrast to noise ratio versus detectability for larger spheres and medium contrast.

Figure 8. Contrast to noise ratio versus detectability per scanner (A) and normalised for true S/B ratios (B). The dashed

line in fi gure B is a power fi t to the data.

Comparison to clinical dataFigure 9 shows CR values versus detectability score for the clinical study (23) and the present work. The slope of a regression of CR versus detectability for the phantom measurement was 3.2 (standard error 0.6), whereas the slope for the clinical study was 3.4 (0.6). Intercepts were 0.1 (0.2) and 0.3 (0.3), respectively.

Figure 9. Detectability score versus contrast ratio for the study by Hashimoto and co-workers (23) and for the Biograph

data from the present study.


Chapter 2

30

Discussion

The present study compared image contrast and lesion detectability for fi ve models of clinical PET scanners installed in The Netherlands, using a simple standard scan protocol with a single phantom with varying lesion sizes and contrasts. Figures 7 and 8 show that image quality, objectively measured as image contrast to noise ratio in a standard phantom, is a good predictor of clinical performance in terms of lesion detectability. The relation between total lesion detectability LD versus mean relative contrast to noise ratio CNR

rel, as shown in Figure 8b, can be described

accurately (r2 0.99; χ2 0.0024; p<0.001) by the following function:

904.0relCNR542.01LD −⋅−= (3)

Apparently, a simple phantom comparison as performed here can give precise information about the expected lesion detectability of a certain scanner, and/or a certain combination of acquisition and reconstruction protocols, in relation to other scanners and acquisition and reconstruction protocols by looking at the measured CNR only. The image quality measurement included in the NEMA 2001 performance measurement protocol could serve the same purpose of directly relating image contrast to detectability (2). The NEMA image quality measurement, however, calculates background variation based on multiple background ROIs in the uniform background, which gives measure of image uniformity rather than image noise in the immediate surroundings of a lesion. In general, as Figure 7 shows, there is a rather clear limit in CNR at around four, above which the detection sensitivity approaches 100%, which is seemingly independent of the observed lesion size. At a CNR between 2 and 4, detectability improves rapidly from very poor to nearly 100%, so a small improvement in contrast or image noise caused by, for example, a diff erent reconstruction method or fi lter, can have a large eff ect on detectability (7). Instead of doing time-consuming observer studies, CNR could be calculated for a series of phantom studies as performed here, and the expected limits for reliable detection, both in size and contrast, could be derived directly from the phantom measurement.

Tables 3 and 4 (AC and NAC respectively) show the p-values of t-tests between lesion detectability of all scanners. Using a p-value of 0.05 as the threshold for a signifi cant diff erence, the sensitivity (AC) in descending order was: Allegro > Biograph > HR+ in 3D mode > HR+ in 2D mode. The HR+ in 2D mode did not have a signifi cantly higher sensitivity than C-PET, and the lowest sensitivity was found for the PET/SPECT. Without AC, no signifi cant diff erence in lesion detectability between Allegro and Biograph was found. The sensitivity (NAC) in descending order was Allegro and Biograph > HR+ 3D and 2D > C-PET and PET/SPECT. The signifi cantly higher lesion detectability as seen for the Allegro and Biograph scanners, compared to the other scanners, corresponded well with higher contrast to noise ratios, but not with image contrasts themselves as seen by the relatively low contrasts for the Allegro images (Figure 4).

Apart from acquisition and reconstruction settings, two factors may have aff ected the results of the present study: variations in sphere to background radioactivity concentration ratios between the scanners and diff erences in thickness of the image planes, with a larger slice thickness resulting in



31

better count statistics per slice. The poorer detectability of the HR+ compared with the Biograph, and the small diff erences between HR+ and C-PET, may at least partly be explained by the small HR+ slice thickness and the relatively low true sphere to background ratio in the HR+ phantom measurement. As Figure 8a shows, average contrast to noise ratio for the C-PET is higher than that of the HR+ in 2D mode whereas the lesion detectability is lower. Figure 8b, however, shows that this is only caused by slightly higher true S/B ratios used in the C-PET measurement. In the Biograph measurement, the true sphere to background ratios were slightly higher than in the Allegro measurements. On the other hand, the Allegro scanner demonstrated higher detectability than the Biograph despite a lower sphere to background ratio and a lower actual image contrast at the time of measurement (Figure 4). Although Allegro scan durations were longer and slice thickness larger, the number of true counts for the Allegro was considerably lower than for the HR+ and Biograph scanners in 3D mode because of the lower absolute sensitivity (in counts/kBq/ml) of the Allegro. The higher lesion detectability may be related to its diff erent 3D (RAMLA) reconstruction algorithm. Unfortunately, comparable fully 3D reconstruction algorithms were not available for the other scanners.

Table 3. Signifi cance of diff erences in total lesion detectability, in order of decreasing performance, with AC

P-values of one-tailed paired t-tests

Biograph HR+, 3D HR+, 2D C-PET PET/

SPECTAllegro 0.02 <0.001 <0.001 <0.001 <0.001

Biograph 0.009 0.001 <0.001 <0.001

HR+, 3D 0.01 <0.001 <0.001

HR+, 2D 0.1 <0.001

C-PET <0.001

Table 4. Signifi cance of diff erences in total lesion detectability, in order of decreasing performance, without AC

P-values of one-tailed paired t-tests

Biograph HR+, 3D HR+, 2D C-PET PET/

SPECTAllegro 0.5 <0.001 <0.001 <0.001 <0.001

Biograph <0.001 <0.001 <0.001 <0.001

HR+, 3D 0.4 0.002 0.01

HR+, 2D 0.005 0.004

C-PET 0.3

In general, AC images were associated with a higher detection sensitivity as compared to NAC images, which is consistent with other publications for example by Bai and others (24), with the exception of the LSO PET/SPECT camera where the opposite was observed. This last eff ect was


Chapter 2

32

previously found for a BGO-based gamma camera PET system as well (25) and may be caused by the considerably poorer count statistics of these systems. A signifi cantly higher detection sensitivity was also observed for 3D as opposed to 2D acquisition for the HR+ scanner as would be expected given the larger acceptance angle for 3D acquisition. However, the observed results obtained with a simple thorax phantom may not readily compare with an actual whole body acquisition in a patient in which there is signifi cant radioactivity outside of the fi eld of view. This could negatively infl uence the results for 3D scanners, especially PET/CT systems, because of their reduced side shielding to allow for larger patient port diameters. This eff ect could be assessed by a similar measurement as performed in the present study, with an additional radioactive phantom placed immediately outside the FOV of the scanner (8).

Figure 9 shows a good correspondence between the detectability versus CR for the phantom measurement and the clinical evaluation in CT-confi rmed low-SUV solid pulmonary lesions by Hashimoto and co-workers. This shows that the contrast levels used in the phantom measurement are very relevant for low-SUV lesions, and that the results of the observer study can be directly related to the expected performance of each scanner for these low-SUV lesions. The measurements by Hashimoto and co-workers were done with an ECAT Accel scanner (Siemens), which is similar to the Biograph with the exception of the attenuation correction, which is based on 68Ge transmission measurements instead of CT.

Conclusion

A high correlation between image contrast to noise ratio and lesion detectability was found, which allows for assessment of clinical performance by a simple phantom measurement without the need of observer studies. In summary, in the conditions tested with the inherent limitations noted above, the highest lesion detectability was observed with the Allegro scanner. The LSO-based scanner Biograph demonstrated better overall performance than the ECAT Exact HR+, which in turn performed better than the C-PET and PET/SPECT.

AcknowledgementsThe authors would like to thank the participants in the observer study and the staff at the Departments of Nuclear Medicine of the VU University Medical Centre, Amsterdam; Netherlands Cancer Institute / Antoni van Leeuwenhoek Hospital, Amsterdam; St. Antonius Hospital, Nieuwegein; Rijnstate Hospital, Arnhem; and Maastricht University Hospital, for providing access to their scanners and for assistance in performing the measurements.



33

References


(2) Daube-Witherspoon ME, Karp JS, Casey ME, DiFilippo FP, Hines H, Muehllehner G et al. PET performance

measurements using the NEMA NU 2-2001 standard. J Nucl Med 2002; 43(10):1398-1409.

(3) Bergmann H, Dobrozemsky G, Minear G, Nicoletti R, Samal M. An inter-laboratory comparison study of

image quality of PET scanners using the NEMA NU 2-2001 procedure for assessment of image quality. Phys

Med Biol 2005; 50(10):2193-2207.

(4) Adam LE, Karp JS, Smith RJ. PET camera performance measurements: A comparison between three PET

cameras. J Nucl Med 1999; 40(5):76P.

(5) Coleman RE, Laymon CM, Turkington TG. FDG imaging of lung nodules: a phantom study comparing SPECT,

camera-based PET, and dedicated PET. Radiology 1999; 210(3):823-828.

(6) Raylman RR, Kison PV, Wahl RL. Capabilities of two- and three-dimensional FDG-PET for detecting small

lesions and lymph nodes in the upper torso: a dynamic phantom study. Eur J Nucl Med 1999; 26(1):39-45.

(7) Farquhar TH, Llacer J, Sayre J, Tai YC, Hoff man EJ. ROC and LROC analyses of the eff ects of lesion contrast,

size, and signal-to-noise ratio on detectability in PET images. J Nucl Med 2000; 41(4):745-754.

(8) Kadrmas DJ, Christian PE. Comparative evaluation of lesion detectability for 6 PET imaging platforms using

a highly reproducible whole-body phantom with (22)Na lesions and localization ROC analysis. J Nucl Med

2002; 43(11):1545-1554.

(9) Kadrmas DJ, Christian PE, Wollenweber SD, Kohlmyer SG, Stearns CW. Comparative evaluation of 2D and 3D

lesion detectability on a full-ring BGO PET scanner. J Nucl Med 2002; 43(5):204S.

(10) Lartizien C, Comtat C, Kinahan PE, Ferreira N, Bendriem B, Trebossen R. Optimization of injected dose based

on noise equivalent count rates for 2- and 3-dimensional whole-body PET. J Nucl Med 2002; 43(9):1268-

1278.

(11) Visvikis D, Griffi ths D, Costa DC, Bomanji J, Ell PJ. Clinical evaluation of 2D versus 3D whole-body PET image

quality using a dedicated BGO PET scanner. Eur J Nucl Med Mol Imaging 2005; 32:1050-1056.

(12) Joshi U, Raijmakers PG, Van Lingen A, Comans EF, Pijpers R, Teule GJ et al. Evaluation of pulmonary nodules:

comparison of a prototype dual crystal (LSO/NAI) dual head coincidence camera and full ring positron

emission tomography (PET). Eur J Radiol 2005; 55(2):250-254.

(13) Brix G, Zaers J, Adam LE, Bellemann ME, Ostertag H, Trojan H et al. Performance evaluation of a whole-body

PET scanner using the NEMA protocol. J Nucl Med 1997; 38(10):1614-1623.

(14) Adam LE, Zaers J, Ostertag H, Trojan H, Bellemann ME, Brix G. Performance evaluation of the whole-body

PET scanner ECAT EXACT HR+ following the IEC standard. IEEE Trans Nucl Sci 1997; 44(3):1172-1179.



Nucl Med 2004; 45(5):813-821.

(16) Joshi U, Hoekstra OS, Boellaard R, Comans EF, Raijmakers PG, Pijpers RJ et al. Initial experience with a

prototype dual-crystal (LSO/NaI) dual-head coincidence camera in oncology. Eur J Nucl Med Mol Imaging

2004; 31(4):596-598.


Chapter 2

34


45(6):1040-1049.

(18) Adam LE, Karp JS, Daube-Witherspoon ME, Smith RJ. Performance of a whole-body PET scanner using

curve-plate NaI(Tl) detectors. J Nucl Med 2001; 42(12):1821-1830.

(19) Tarantola G, Zito F, Gerundini P. PET instrumentation and reconstruction algorithms in whole-body

applications. J Nucl Med 2003; 44(5):756-769.

(20) Watson CC, Newport D, Casey ME, Dekemp RA, Beanlands RS, Schmand M. Evaluation of simulation-based

scatter correction for 3-D PET cardiac imaging. IEEE Trans Nucl Sci 1997; 44(1):90-97.

(21) Watson CC. New, faster, image-based scatter correction for 3D PET. IEEE Trans Nucl Sci 2000; 47(4):1587-

1594.

(22) Bergström M, Eriksson L, Bohm C, Blomqvist G, Litton J. Correction for scattered radiation in a ring detector

positron camera by integral transformation of the projections. J Comput Assist Tomogr 1983; 7(1):42-50.

(23) Hashimoto Y, Tsujikawa T, Kondo C, Maki M, Momose M, Nagai A et al. Accuracy of PET for diagnosis of solid

pulmonary lesions with 18F-FDG uptake below the standardized uptake value of 2.5. J Nucl Med 2006;

47(3):426-431.

(24) Bai C, Kinahan PE, Brasse D, Comtat C, Townsend DW, Meltzer CC et al. An analytic study of the eff ects of

attenuation on tumor detection in whole-body PET oncology imaging. J Nucl Med 2003; 44(11):1855-1861.

(25) Tocharoenchai C, Tsui BM, Frey EC, Wang WT. Eff ect of attenuation correction on lesion detection using a

hybrid PET system. J Med Assoc Thai 2005; 88(1):96-102.


C h a p t e r

In search of an unknown primary tumour presenting with extracervical metastases: the diagnostic performance of FDG-PET

3Urvi JoshiJacobus JM van der HoevenEmile FI ComansGerarda JM HerderGerrit JJ TeuleOtto S Hoekstra

In: British Journal of Radiology 2004;77(924):1000-1006


Abstract

A retrospective study was carried out to determine the performance of 18F-fl uorodeoxyglucose positron emission tomography (FDG-PET) in patients with unknown primary tumours presenting with metastases external to the neck. All patients referred to an academic PET centre (July, 1997 to December, 2000) presenting with an extracervical metastasis and no prior systemic therapy were eligible. The minimum follow-up period was 11 months. From 63 eligible cases, known metastases were FDG avid in all but one neuroendocrine process. PET scans were retrospectively classifi ed as positive for a primary tumour (n=29), i.e. revealing at least one anatomical site suspected to be the primary tumour. This was confi rmed in 16, either by histology (n=10) or radiological and clinical

follow-up (n=6). There were four false positive cases. In nine patients, the primary tumour was never confi rmed. Of the remaining 33 negative PET scans the primary tumour was clinically not

found in 18. Follow-up and additional pathology investigations demonstrated the primary tumour in 15. A survey on clinical usefulness of PET (response rate 83%) suggested that PET positively

contributed to diagnostic understanding in 29 of 52 evaluable cases. Applied late in the diagnostic trajectory, approximately four patients need to be scanned by PET in order to fi nd one primary tumour. However, in addition to direct demonstration of unknown primaries, there appears to be a positive eff ect on the diagnostic work-up of these patients of a similar magnitude.


In search of an unknown primary tumour presenting with extracervical metastases

Introduction

An unknown primary tumour is defi ned as a biopsy-proven malignancy from unidentifi ed anatomical origin following diagnostic evaluation (1). Investigation often comprises a complete history, thorough physical examination, a blood chemistry profi le, chest radiography,

abdominopelvic CT and mammography in women (1-4). If these tests do not disclose the primary tumour, a myriad of diagnostic tests are available to the clinician. The estimated incidence of unknown primary tumours is 2–7% of all cancers (1;3;5-7). In The Netherlands, the observed incidence is 2500 per year (8).

The medical oncologist must be certain that tumours that are potentially curable when metastatic have been ruled out, such as lymphomas and germ cell tumours in addition to tumours that may have a good response to therapy such as breast cancer, ovarian cancer, prostate cancer and small cell lung cancer. The identifi cation of a primary tumour may off er the possibility of a more specifi c

and effi cacious treatment than if the patient is treated with a general regimen for tumours of unknown primary origin. The dilemma for the oncologist is how much investigation is appropriate.

For example, an extensive endoscopic procedure is expensive and may put the patient at risk for complications of this procedure. Therefore the oncologist needs tools that give a high probability

of fi nding the primary tumour with a minimum of discomfort for the patient.

Positron emission tomography (PET) using 18F-fl uorodeoxyglucose (FDG) is an attractive tool for this indication since most malignancies are FDG avid, its biodistribution is favourable and the whole body is scanned in a single session with minimal patient discomfort. Whereas several studies have indicated that FDG-PET is useful to locate primaries within the head and neck region (9-14), very few data are available on its usefulness in extracervical metastases. After exclusion of melanoma patients, who are considered to be a separate subset (1-4), accumulated data from the

four studies (15-18) providing a specifi c evaluation of the yield of PET in extracervical metastases consists of 48 patients with an overall approximate aggregated yield of 50%. Rades et al (19) have also evaluated the value of PET in cancer of unknown primary. However, it was not possible with the available data in the article to calculate the yield in the subset of patients presenting with extracervical metastases. Therefore, the aim of the present study was to examine the diagnostic performance of FDG-PET scanning after a negative clinical work-up in patients presenting with extracervical metastases from an unknown primary tumour.

Materials and methods

PatientsFrom the FDG-PET scans performed from July 1997 to December 2000, patients presenting with metastases associated with unknown primary tumours were identifi ed using the diagnostic coding

system applied in our institution (20). Within this period, referrals for this indication were accepted only if the indicated standard work-up procedures had been done, according to the referring clinician. Inclusion criteria consisted of the presence of a defi nable suspect metastatic tumour

37


Chapter 3

38

for which the primary tumour was unknown (i.e. patients presenting with only paraneoplastic

phenomena were not included). In addition, the presenting site of the suspected metastasis had to be exterior to the neck region. Patients presenting with cervical metastases are often found

to have primary tumours limited to the head and neck region. For the purposes of this study and for comparison with other similar studies in the literature, supraclavicular lymph nodes were considered to be exterior to the neck and as such, patients presenting with a supraclavicular lymph node metastasis were included in our study. Finally, the patients had undergone no prior systemic anticancer treatment.

The initial search for unknown primary tumours revealed a total of 101 cases. Four patients were excluded because they presented with paraneoplastic phenomena without an identifi able tumour mass. An additional 27 patients were excluded because they presented with a (suspected) metastasis in the neck region. Finally, seven patients were excluded because they had already undergone systemic therapy. The resultant eligible group consisted of 63 cases. The medical records of these patients were then searched to obtain information on the fi nal diagnosis and compared with the PET result. Patients were followed until death or for a minimum period of 11 months.

We used serial questionnaires in an attempt to evaluate the impact of PET results on the diagnostic understanding of the referring clinician and their perception of the eff ect on patient management. For this purpose, questionnaires were sent prior to the PET scan, approximately 1 month after the scan, and about 6 months later. Details of the third questionnaire that specifi cally evaluates the impact on diagnostic understanding and therapy choice can be found in Table 1. Further details regarding the questionnaires have been published elsewhere (21;22).

Table 1. Questionnaire on evaluation of impact PET

Diagnostic understanding (DU)

DU = 1 PET confused my understanding of this patient’s disease and led to investigations that I would not have otherwise have done

DU = 2 PET confused my understanding of this patient’s disease but did not lead to any additional investigations

DU = 3 PET had little or no eff ect on my understanding of this patient’s disease

DU = 4 PET provided information which substantially improved my understanding of this patient’s disease

DU = 5 My understanding of this patient’s disease depended upon diagnostic information provided only by PET (unavailable from any other non-surgical source)

Therapy choice (TC)

TC = 1 PET led me to choose therapy which in retrospect was not in the best interests of the patient

TC = 2 PET was of no infl uence in my choice of therapy

TC = 3 PET did not alter my choice of therapy but did increase my confi dence in the choice

TC= 4 PET contributed to a change in my chosen therapy but other factors (other imaging tests, other diagnostic tests, changes in patient status) were equally or more important

TC =5 PET was very important compared with other factors in leading to a benefi cial change in therapy



39

PET scanPET scans were performed on a Siemens ECAT EXACT HR+ scanner (Siemens, Erlangen, Germany), involving a whole body acquisition in two-dimensional (2D) mode with emission scans of 5 min

to 7 min per bed position starting approximately 1 h after intravenous administration of 370 MBq of FDG. Patients had fasted a minimum of 6 h prior to FDG administration and their serum glucose was <10 mmol l–1. The value of attenuation correction is controversial (23) and our routine whole body FDG-PET protocol does not include transmission scanning.

PET scan result classifi cation Based on the actual PET scan report sent to the referring clinicians, two nuclear medicine physicians who were blinded for clinical outcome retrospectively classifi ed PET scan results into two categories: 1. Positive: the original report mentioned at least one anatomical site suspected to be the

primary tumour. 2. Negative: the original report mentioned no evidence of a primary tumour, with only defi nite

metastatic sites demonstrating FDG uptake.

Results

Patient characteristicsThe study group comprised 63 cases and 62 patients (one patient underwent two PET scans); their mean age was 57 years (standard deviation (SD) 12 years) and 52% were females. A summary of pre-PET pathology can be found in Table 2.. Most (87%) had no prior history of malignancy. Eight patients had previously diagnosed primary tumours (including lung, colon, bladder, brain, cutaneous melanoma and basal cell carcinoma); in none of them was the pathological classifi cation of the current presentation compatible with this original primary. PET was typically requested after an extensive clinical work-up (median duration 3 months) had been fruitless; 81% had undergone at least one CT scan. 35 (56%) of the cases were referred from a tertiary academic

hospital centre, all but one from within our institution, and 27 from community hospitals. Referring medical specialists included oncologists (n=20, 32%), internists (n=16, 25%), pulmonologists

(n=13, 21%), oncological/general surgeons (n=9, 14%), urologists (n=2, 3%), neurologists (n=2, 3%) and a neurosurgeon. Patients were followed until death or for a minimum of 11 months. The

median follow-up in non-deceased patients was 28 months (range 11–51 months). On the last date of follow-up (April 2002), 37% of the patients were still alive. The median follow-up in patients in whom a primary tumour was not found was 29 months (range 17–51 months) in non-deceased patients versus 5 months (range 1–37 months) in deceased ones.

Presenting site(s) and histologyThe largest group of patients (n=17, 27%) presented with lymph node metastases (12 of whom with at least one peripheral involved nodal station (Table 2) followed by metastases in liver (n=7, 11%), lung (n=6, 10%), pleura (n=6, 10%), bone (n=6, 10%), brain (n=4, 6%), skin (n=4, 6%) and other soft tissue sites (n=6, 10%). Seven patients (11%) presented with a combination of the above-described categories. In 59 (94%), the metastases had been pathologically confi rmed.


Chapter 3

40

Table 2. Patient characteristics and PET results

Patient Age* GenderPresenting Site Pre-PET Pathology

PET Scan Result Primary Tumour

1 82 male Supraclav LN*** adenocarcinoma positive** lung carcinoma2 51 male Liver adenocarcinoma negative not confi rmed3 78 male Lung adenocarcinoma negative Lung carcinoma4 57 male Mediast LN adenocarcinoma positive Thymoma5 42 female Axillary LN adenocarcinoma negative breast carcinoma6 71 female Bone, liver adenocarcinoma positive** gastric carcinoma7 59 female Soft tissue adenocarcinoma equivocal not confi rmed8 55 male Bone & liver adenocarcinoma positive** Lung carcinoma9 45 male Bone small cell negative Brain10 67 female Brain not available equivocal Brain11 56 male Mediast LN not available equivocal not confi rmed12 56 male Supraclav/mediast LN adenocarcinoma positive not confi rmed13 71 male Liver adenocarcinoma positive not confi rmed14 37 female Soft tissue squamousadenocarcinoma positive** endometrial carcinoma15 69 male Soft tissue adenocarcinoma equivocal not confi rmed16 49 female Lung adenocarcinoma positive** cholangiocarcinoma17 65 male Liver adenocarcinoma positive not confi rmed18 47 female Supraclav LN large cell negative mediastinal primary19 49 female Axillary LN adenocarcinoma positive** breast carcinoma20 53 male Bone, liver, lung adenocarcinoma negative not confi rmed21 52 female Soft tissue inconclusive negative primitive neuroectodermal tumor22 52 female Soft tissue adenocarcinoma positive not confi rmed23 75 female Liver adenocarcinoma negative cholangiocarcinoma24 64 male Bone neuroendocrine carcinoma no FDG uptake neuroendocrine carcinoma25 54 female Supraclav/retroperit LN adenocarcinoma equivocal not confi rmed26 43 female Lung adeno/large cell carcinoma negative not confi rmed27 52 female Liver adenocarcinoma positive cholangiocarcinoma28 76 female Axillary LN adenocarcinoma positive not confi rmed29 72 male Lung adenocarcinoma negative lung carcinoma30 62 male Bone undiff erentiated positive** lung carcinoma31 50 female Retroperit LN squamous cell carcinoma equivocal not confi rmed32 66 male Soft tissue small cell positive** lung carcinoma33 64 female Soft tissue adenocarcinoma positive breast carcinoma34 38 female Brain adenocarcinoma negative lung carcinoma35 60 male Soft tissue adenocarcinoma negative Mesothelioma36 42 male Liver adenocarcinoma equivocal not confi rmed37 60 female Infraclav LN adenocarcinoma positive not confi rmed38 29 female Soft tissue adenocarcinoma equivocal not confi rmed39 72 female Lung not available negative not confi rmed40 57 female Axillary LN adenocarcinoma equivocal not confi rmed41 48 female Brain adenocarcinoma positive lung carcinoma42 67 female Bone adenocarcinoma positive not confi rmed43 63 male Bone not available equivocal lung carcinoma44 74 female Soft tissue adenocarcinoma negative not confi rmed45 51 female Soft tissue undiff erentiated equivocal not confi rmed46 52 female Bone/liver carcinoid positive not confi rmed47 53 female Soft tissue large cell positive** lung carcinoma48 60 male Mediast LN large cell equivocal not confi rmed 49a 70 male Soft tissue large cell equivocal renal cell carcinoma 49b positive** renal cell carcinoma50 48 female Brain large cell positive** lung carcinoma51 39 female Retroperit LN/liver adenocarcinoma negative cholangiocarcinoma52 60 male Lung adenocarcinoma negative lung carcinoma53 41 female Axillary LN adenocarcinoma equivocal not confi rmed54 75 male Mediast LN/bone inconclusive negative Mesothelioma55 42 male Inguinal LN squamous cell carcinoma equivocal not confi rmed56 48 female Liver large cell equivocal not confi rmed57 43 male Soft tissue adenocarcinoma positive** lung carcinoma

58 64 maleSupraclav/mediast/ retroperit LN adenocarcinoma positive prostate carcinoma

59 42 male Bone adenocarcinoma positive not confi rmed60 50 male Soft tissue adenocarcinoma positive** colorectal carcinoma61 32 female Axillary LN adenocarcinoma positive** breast carcinoma62 72 male Bone adenocarcinoma positive** thyroid carcinoma

Legend: * at time of PET scan ** confi rmed true positive *** LN : lymph node



41

In the others, the metastatic sites were either not readily accessible for biopsy or considered

pathognomonic radiologically (brain metastases, multiple pulmonary metastases on chest CT). The most common histological classifi cation was adenocarcinoma (n=40, 64%) followed by large cell carcinoma (n=7, 11%). Only two patients had squamous cell cancer.

PET scan resultsThe known metastases were FDG avid in all but one patient with bone metastases of a neuroendocrine tumour. In addition, previously unknown metastatic lesions were seen on the PET scan in 27 (43%) of the cases. 29 (46%, 95% CI 33–59%) of the PET scans were classifi ed as positive for a primary tumour (Table 3). Within this group, the PET result was confi rmed to be true positive in 16 cases, either by histology (n=10) or by radiological and clinical follow-up (n=6). The following primary tumours were found: lung carcinoma (n=8; in 2 the initial work-up did not include a chest

CT; both had a prior malignancy of bladder and colon, respectively), breast carcinoma (n=2), colon carcinoma (n=1), gastric carcinoma (n=1), cholangiocarcinoma (n=1) renal cell carcinoma (n=1),

endometrial carcinoma (n=1) and thyroid carcinoma (n=1, Figure 1).

Figure 1. True positive FDG-PET. Patient 62: 70-year-old male with bone metastases. Besides multiple bone metastases

(vertebrae, humerus, chest wall), PET revealed markedly increased uptake in the thyroid (arrow). Thyroidectomy

confi rmed a papillary thyroid carcinoma.

Table 3. Yield of PET and other tests to demonstrate primary tumours

Primary tumour confi rmed, indicated by Primary not found

PET Immunochemistry/histology Other tests

PET + 16 0 4 9

PET - 0 7 8 18


Chapter 3

42

There were four false positive cases: a breast abscess (biopsy proven); thoracotomy later confi rmed the presence of a primary lung tumour (this lesion had been reported at PET but was interpreted

as a metastasis of the presumed breast malignancy, Figure 2). In another patient, who presented with extensive supradiaphragmatic and infradiaphragmatic lymphadenopathy, the kidney was incorrectly interpreted as the primary; the patient was later clinically diagnosed as having prostate cancer. In the third case, the primary site was incorrectly thought to be in the lung (granuloma

appearance on CT) with follow-up revealing the diagnosis of a primary breast tumour (biopsy proven) 11 months after PET. In the fourth case, the primary site was misinterpreted as being

in the stomach (retrospectively likely in the left liver lobe) in a patient clinically diagnosed with cholangiocarcinoma and multiple liver metastases. In nine patients, the primary tumour was not confi rmed.

33 (52%, 95% CI 40–64%) PET scans were negative for a primary tumour, and in 18 the primary tumour was not found. In 12 patients, multiple previously unknown tumour sites were evident at PET but these were assumed to be metastases. Seven cases were solved by pathology revision (including additional immunohistochemical analyses) yielding primary cholangiocarcinoma

(n=2), mesothelioma (n=2), thymoma (n=1), recurrent oligoastrocytoma (n=1) and primary primitive neuroectodermal tumour (n=1). Follow-up revealed the primary in 8 of 33 patients: lung cancer was found in fi ve (in four, a known lung lesion was later considered to be the primary; in another, presenting with a lower lumbar skeletal metastasis, a primary lung tumour was later diagnosed; PET had demonstrated only moderately increased uptake in the left hilar region). The sixth patient presented with what was initially thought to be three metastatic brain lesions. After debulking surgery of the largest lesion a histological diagnosis of glioblastoma multiforme was

Figure 2. False positive FDG-PET. Patient 41: 48-year-old female with cerebral metastases. PET demonstrates increased uptake in the right breast (abscess, misinterpreted as the primary tumour, small arrow) as well as the right lung (primary tumour, misinterpreted as metastasis, large arrow).



43

made. PET had demonstrated defi nite focal increased tracer activity in the largest brain lesion. In the seventh case, PET had been false negative in a patient with axillary lymph node metastasis surgically proven to have a 7 mm primary breast tumour (1 month after PET). The last patient in whom follow-up disclosed the primary tumour after a negative PET scan was the only patient to undergo two PET scans. While his second PET scan was diagnostic for a primary tumour in the left kidney, the initial scan demonstrated several other equally intense/extensive lesions from which assignment of a single site as the primary tumour was diffi cult.

The clinician’s perspectiveIn 52 of 63 (83%) cases, the clinicians returned the survey questionnaires (2 of 52 lacking information on impact on management). PET had a positive impact on diagnostic understanding (DU) in 29 (DU= 4 or 5). Beyond demonstration of the primary, this retrospectively also refl ected the utility of PET to point at accessible sites for biopsy (if no prior histological diagnosis had been feasible), to simplify or redirect the diagnostic thinking guided by extent and pattern of previously unknown metastases, respectively. To a greater or lesser extent, the PET result had favourably contributed to a change in management in 17 patients. This involved diff erent systemic therapy (n=5, e.g. guided by histology obtained after PET, Figure 3), change from local to systemic therapy or vice versa (n=4, e.g. PET showing more extensive disease, or a presumed metastasis was later considered to be the primary, lacking any evidence of other tumour activity, respectively), switch from single to multimodality therapy or vice versa (n=4, e.g. in case of locally advanced breast cancer, and metastasized thyroid cancer), or to start or refrain from systemic treatment (n=3, e.g. prompted by a classifying histological diagnosis or by the extent of dissemination evident at PET, respectively),

and in one patient local therapy was changed from radiotherapy to surgery. Impact of PET was estimated “negative” (therapy choice (TC)=1 or 2) in 3 patients (in 2 of them, additional tests were instigated by PET which would not have been done otherwise), and without impact (TC =3) in 20.

Figure 3. Multiple positive foci. Patient 11: 56-year-old male with mediastinal lymphadenopathy (biopsy diffi cult). PET

demonstrates multiple foci of increased uptake (left pleura, possible separate focus in left lung, mediastinum, ribs,

vertebrae). A rib biopsy based on PET (arrow) revealed small cell lung cancer.


Chapter 3

44

Discussion

The yield (i.e. percentage of confi rmed true-positive results) of FDG-PET scanning in patients with extracervical metastases from an unknown primary tumour in our series was 25% (16/63, 95% CI 15–38%). To our knowledge, our study population comprises the largest series reported to date of FDG-PET scans performed in patients presenting with extracervical metastases related to an unknown primary tumour.

This observed yield appears to be lower than that reported by other investigators (15-19). One possible reason could be diff erences in the pre-PET diagnostic work-up. For example, in the retrospective study by Bohuslavski et al (15), only specifi c mention was made of chest radiography in the pre-PET diagnostic imaging work-up. However, the pre-PET diagnostic work-up appeared

adequate in the prospective studies of Kole et al (16) and Lassen et al (17). The PET scans in our patients were performed late in the diagnostic trajectory, i.e. typically after an extensive diagnostic work-up. There were, however, two patients with prior bladder and colon primary tumours who were found to have a second primary lung tumour on the FDG-PET scan. These patients had

only undergone chest X-rays and abdominal CT scans because of their prior histories and clinical presentation. However, it is controversial whether chest CT is necessary in the routine work-up of unknown primary tumours (1;24) and is not mentioned in the recommendations of Abbruzzese et al (1). Additionally, patient sample size may play a role: in studies published so far, the number of patients with extracervical metastases (excluding melanoma metastases) has ranged from 8 to 22.

The clinicians reported a positive contribution of PET to their diagnostic understanding twice as often as the observed true positive rate for demonstration of a primary tumour. Negative PET fi ndings and the pattern of newly discovered metastatic spread may serve to reduce the number of diff erential diagnoses and better histological classifi cation may result from biopsy of a newly detected lesion. We therefore suggest that the contribution of PET to diagnostic thinking may extend beyond the number of true positive primaries directly demonstrated by PET. However,

robust data can only be obtained with a randomized trial.

Even though whole body FDG-PET has several advantages over conventional imaging, there are limitations. The fi rst is paradoxically related to its high sensitivity. In over 40% of our cases PET disclosed previously unknown lesions. In such cases, it may be diffi cult to diff erentiate metastases from a possible primary tumour. The second is related to the biodistribution of FDG. For example,

FDG is excreted by the kidneys and the ability to detect renal tumours may be limited. Alternatively, FDG uptake is low in several tumours and in our study, as expected, in prostrate cancer and neuroendocrine tumours (25-27). Finally, precise anatomical localization with PET can be diffi cult in comparison with cross-sectional radiological imaging. This may improve with the introduction of combination PET-CT scanners.

It was surprising to learn how often (n=7/63, 11%) histology revision of the biopsy specimen solved the diagnostic dilemma (this may have included re-biopsy and/or case presentation at



45

a specialized panel discussion). We do not know whether histology revision was more often requested in case of a negative PET scan result. It appeared, however, that histology revision more often provided the solution in cases in which the PET scan was classifi ed as negative. In our institution, a pathological review often occurs but is not uniformly applied. In their article,

Abbruzzese et al (5) describe a review of the pathological material as the initial step in a diagnostic strategy for all patients with tumours of unknown origin referred to their institution. Following this observation, it may be advisable to suggest, at least in our institution, that all cases of unknown primary tumours with extracervical metastases undergo routine pathology revision and/or panel discussion prior to FDG-PET scanning. Accordingly, the yield of PET would have increased from 16/63 to 16/56 (29%).

In conclusion, when applied late in the diagnostic trajectory, approximately four patients need to be scanned by PET in order to fi nd one primary tumour. However, in addition to direct demonstration

of unknown primaries, there appears to be a positive impact of PET on the diagnostic work-up of these patients of a similar magnitude.

Prediction of yield in LSO-PS scannersThe median diameter of the true positive primary tumours, as estimated from the HR+ PET images, was 4 cm (interquartile range: 1.8-6 cm). Assuming that for common relative FDG uptakes, tumours > 2cm (see Chapter 2) are likely to be detected with the LSO-PS scanner, we predict that 64% would also have been detected with the latter system.


Chapter 3

46

References

(1) Abbruzzese JL, Lenzi R, Raber MN. Carcinoma of unknown primary. In: Abelhoff , editor. Edinburgh, UK:

Churchill Livingstone, 2000: 2314-2329.

(2) Abbruzzese JL, Abbruzzese MC, Lenzi R, Hess KR, Raber MN. Analysis of a diagnostic strategy for patients

with suspected tumors of unknown origin. J Clin Oncol 1995; 13(8):2094-2103.

(3) Hillen HF. Unknown primary tumours. Postgrad Med J 2000; 76(901):690-693.

(4) Karsell PR, Sheedy PF, O’Connell MJ. Computed tomography in search of cancer of unknown origin. JAMA

1982; 248(3):340-343.

(5) Abbruzzese JL, Abbruzzese MC, Hess KR, Raber MN, Lenzi R, Frost P. Unknown primary carcinoma: natural

history and prognostic factors in 657 consecutive patients. J Clin Oncol 1994; 12(6):1272-1280.

(6) Greco FA, Burris HA, III, Erland JB, Gray JR, Kalman LA, Schreeder MT et al. Carcinoma of unknown primary

site. Cancer 2000; 89(12):2655-2660.

(7) Lortholary A, Abadie-Lacourtoisie S, Guerin O, Mege M, Rauglaudre GD, Gamelin E. [Cancers of unknown

origin: 311 cases]. Bull Cancer 2001; 88(6):619-627.

(8) Visser O CJSL. Incidence of Cancer in The Netherlands. 1993. Utrecht, Netherlands Cancer Registry.

(9) Aassar OS, Fischbein NJ, Caputo GR, Kaplan MJ, Price DC, Singer MI et al. Metastatic head and neck cancer:

role and usefulness of FDG PET in locating occult primary tumors. Radiology 1999; 210(1):177-181.

(10) Braams JW, Pruim J, Kole AC, Nikkels PG, Vaalburg W, Vermey A et al. Detection of unknown primary head

and neck tumors by positron emission tomography. Int J Oral Maxillofac Surg 1997; 26(2):112-115.

(11) Greven KM, Keyes JW, Jr., Williams DW, III, McGuirt WF, Joyce WT, III. Occult primary tumors of the head and

neck: lack of benefi t from positron emission tomography imaging with 2-[F-18]fl uoro-2-deoxy-D- glucose.

Cancer 1999; 86(1):114-118.

(12) Jungehulsing M, Scheidhauer K, Damm M, Pietrzyk U, Eckel H, Schicha H et al. 2[F]-fl uoro-2-deoxy-D-

glucose positron emission tomography is a sensitive tool for the detection of occult primary cancer

(carcinoma of unknown primary syndrome) with head and neck lymph node manifestation. Otolaryngol

Head Neck Surg 2000; 123(3):294-301.

(13) Regelink G, Brouwer J, de Bree R, Pruim J, van der Laan BF, Vaalburg W et al. Detection of unknown primary

tumours and distant metastases in patients with cervical metastases: value of FDG-PET versus conventional

modalities. Eur J Nucl Med Mol Imaging 2002; 29(8):1024-1030.

(14) Safa AA, Tran LM, Rege S, Brown CV, Mandelkern MA, Wang MB et al. The role of positron emission

tomography in occult primary head and neck cancers. Cancer J Sci Am 1999; 5(4):214-218.

(15) Bohuslavizki KH, Klutmann S, Kroger S, Sonnemann U, Buchert R, Werner JA et al. FDG PET detection of

unknown primary tumors. J Nucl Med 2000; 41(5):816-822.

(16) Kole AC, Nieweg OE, Pruim J, Hoekstra HJ, Koops HS, Roodenburg JL et al. Detection of unknown occult

primary tumors using positron emission tomography. Cancer 1998; 82(6):1160-1166.

(17) Lassen U, Daugaard G, Eigtved A, Damgaard K, Friberg L. 18F-FDG whole body positron emission

tomography (PET) in patients with unknown primary tumours (UPT). Eur J Cancer 1999; 35(7):1076-1082.



47

(18) Lonneux M, Reff ad A. Metastases from unknown primary tumor: PET-FDG as initial diagnostic procedure?

Clin Positron Imaging 2000; 3(4):137-141.

(19) Rades D, Kuhnel G, Wildfang I, Borner AR, Schmoll HJ, Knapp W. Localised disease in cancer of unknown

primary (CUP): the value of positron emission tomography (PET) for individual therapeutic management.

Ann Oncol 2001; 12(11):1605-1609.

(20) American College of Radiology Index for Radiological Diagnoses. 4th ed. Reston, Virginia: American College

of Radiology, 1992.

(21) Herder GJ, Van Tinteren H, Comans EF, Hoekstra OS, Teule GJ, Postmus PE et al. Prospective use of serial

questionnaires to evaluate the therapeutic effi cacy of 18F-fl uorodeoxyglucose (FDG) positron emission

tomography (PET) in suspected lung cancer. Thorax 2003; 58(1):47-51.

(22) Mijnhout GS, Comans EF, Raijmakers P, Hoekstra OS, Teule GJ, Boers M et al. Reproducibility and clinical value

of 18F-fl uorodeoxyglucose positron emission tomography in recurrent melanoma. Nucl Med Commun

2002; 23(5):475-481.

(23) Schauwecker DS, Siddiqui AR, Wagner JD, Davidson D, Jung SH, Carlson KA et al. Melanoma patients

evaluated by four diff erent positron emission tomography reconstruction techniques. Nucl Med Commun

2003; 24(3):281-289.

(24) Latief KH, White CS, Protopapas Z, Attar S, Krasna MJ. Search for a primary lung neoplasm in patients with

brain metastasis: is the chest radiograph suffi cient? AJR Am J Roentgenol 1997; 168(5):1339-1344.

(25) Adams S, Baum R, Rink T, Schumm-Drager PM, Usadel KH, Hor G. Limited value of fl uorine-18

fl uorodeoxyglucose positron emission tomography for the imaging of neuroendocrine tumours. Eur J Nucl

Med 1998; 25(1):79-83.

(26) Eff ert PJ, Bares R, Handt S, Wolff JM, Bull U, Jakse G. Metabolic imaging of untreated prostate cancer by

positron emission tomography with 18fl uorine-labeled deoxyglucose. J Urol 1996; 155(3):994-998.

(27) Pasquali C, Rubello D, Sperti C, Gasparoni P, Liessi G, Chierichetti F et al. Neuroendocrine tumor imaging: can

18F-fl uorodeoxyglucose positron emission tomography detect tumors with poor prognosis and aggressive

behavior? World J Surg 1998; 22(6):588-592.



C h a p t e r

Attenuation corrected versus non-attenuation corrected FDG-PET in oncology, a systematic review

4Urvi JoshiPieter GHM RaijmakersIngrid I RiphagenGerrit JJ TeuleArthur van LingenOtto S Hoekstra

In press: Molecular Imaging and Biology


Abstract

Purpose: To perform a systematic review and meta-analysis to determine the diagnostic accuracy of non-attenuation corrected (NAC) vs. attenuation corrected (AC) FDG-PET in oncological patients.

Methods: We performed a comprehensive literature search in Medline and Embase databases without date and language restrictions. The methodological quality of eligible studies was independently assessed by two reviewers. The diagnostic value of attenuation correction was studied through its sensitivity and specifi city compared to histology, and by comparing the relative detection rate reported with NAC versus AC PET, for full ring – as well as for dual head coincidence PET (FR- and DH-PET, respectively). Furthermore, lesion detection rates were compared for diff erent body locations (head/neck, chest, abdomen).

Results: Twelve studies met the inclusion criteria. For FR-PET, the pooled sensitivity and specifi city on a patient basis, were 64 and 97% for AC, and 62% and 99% for NAC, respectively. Pooled lesion detection with NAC - vs. AC FR-PET was 98% (95% CI: 96-99%; n=1012 lesions), and 88% (95%CI: 81-94%) with NAC - vs. AC DH-PET (n=288 lesions), without a signifi cant eff ect of location of lesions in the body.

Conclusion: For FR-PET, the cumulated evidence suggests similar sensitivity/specifi city and lesion detection for AC versus NAC, and for DH-PET a signifi cantly higher lesion detection for AC versus NAC.


Attenuation corrected versus non-attenuation corrected FDG-PET in oncology

51

Introduction

Attenuation of photons originating from the subject before they are detected by the camera is a generic limitation of nuclear medicine imaging. This attenuation can lead to image distortion and impairs adequate quantifi cation. Attenuation correction has been commonly employed in FDG imaging in an attempt to correct for these eff ects. With PET scanners, this is accomplished by transmission scanning using a radionuclide source such as germanium-68 or cesium-137 and with PET-CT using computed tomography. With respect to visual interpretation of the images, the added value of attenuation correction has been controversial. Whereas attenuation correction provides a more realistic image of FDG distribution, its application signifi cantly increases acquisition times on standard full ring PET scanners. In addition, the performance of attenuation correction can introduce noise and even artifact. Paradoxically, even if the nuclear medicine community sees attenuation correction, or the lack of it, as a potential eff ect-modifi er of test accuracy, its impact is rarely accounted for in systematic reviews on the diagnostic accuracy of PET. As a result, the impact of attenuation correction on lesion detectability and interpretation of PET for oncological purposes is not well established. With PET-CT scanning it is customary to evaluate either modality (primarily to account for artifacts) but one needs to know how to deal with discrepancies.

The objective of this systematic review and meta-analysis was to determine the diagnostic accuracy of non-attenuation corrected and attenuation corrected FDG-PET in oncological patients. We studied the eff ects of attenuation correction for both full ring PET (FR-PET) and dual head coincidence PET (DH-PET), and as a function of diff erent body locations (head/neck, chest, abdomen).


Chapter 4

52

Methods

Literature searchA computer-aided literature search was performed in both Medline and Embase databases without time range or language restrictions, applying controlled vocabulary (MeSH and EMTREE keywords, respectively) as well as free text words. The search date was February 10, 2006. The search strategy (Appendix 1) included terms for positron emission tomography with FDG, modifi ed from Mijnhout et al (1) as well as search terms identifying both radionuclide and x-ray transmission, emission, attenuation correction and oncological studies in humans. In addition, the reference lists of the eligible articles were reviewed to ensure that relevant articles had not been missed.

Study selectionFrom the list of retrieved articles, articles were initially evaluated for eligibility on basis of title and abstract by two independent reviewers (UJ, PR). If there was uncertainty as to whether an article was eligible for inclusion, the entire article was reviewed. Inclusion criteria were: 1) clinical studies evaluating FDG imaging with and without attenuation correction in oncology patients, 2) study population of at least ten patients, 3) suffi cient detail to reconstruct a 2 x 2 contingency table expressing FDG imaging results by disease status, or suffi cient detail to reconstruct relative lesion detection measurement of AC versus NAC imaging, 4) studies utilizing FR-PET and/or DH-PET. We excluded abstracts, editorials, reviews, although the latter two categories were used for cross referencing.

Methodological quality assessmentThe methodological quality of each article was independently assessed by each reviewer in terms of internal and external validity (Table 1), based on the Cochrane Methods Group in Screening and Diagnostic Tests, modifi ed for our area of interest (2). Internal validity items focus on whether a valid reference test was used and whether this reference test was uniformly and independently applied and interpreted as well the type of study design. The external validity items evaluate the applicability of the results in terms of the type of patient population and spectrum, demographics, the inclusion/exclusion criteria, the knowledge of previous test/clinical information that could infl uence interpretation, and the index test characteristics. Items were scored as either positive, negative or unclear.



53

A. In

ternal stu

dy valid

ity

Al. Valid reference test

Histology, A

ttenuation-corrected full ring or dual head coincidence PET

A2. Blind m

easurement of reference test(s) w

ithout knowledge of index test

Assessm

ent of reference test independent of index test(s) results

A3. A

voidance of verifi cation biasC

hoice of patients assessed by reference test independent of index test result

A4. Index test(s) interpreted independently of all clinical inform

ationM

entioned in publication

A5. Prospective study

Mentioned in publication

B. Extern

al stud

y validity

B1. Spectrum of diseases

Localization of disease described (selected or general)

B2. Dem

ographic information

Age and sex given

B3. Inclusion criteria describedM


B4. Exclusion criteria describedM


B5. Avoidance of selection bias

Consecutive series of patients

B6. Standardized execution of index test(s)D

escribed technical aspects of index test(s)

C. R

epro

du

cibility d

escribed

Mentioned in publication

Table 1. Methodological assessment of individual diagnostic studies: criteria


Chapter 4

54

Data extraction and quantitative analysisIn addition to methodological quality assessment, data related to the type of camera, the FDG dose, the time interval between injection and imaging, the transmission and emission acquisition protocols, the reconstruction protocol and the interpretation protocol were independently extracted from each study by each reviewer. For studies where it was possible, a contingency 2 x 2 table was constructed. Disagreements were solved by consensus.

For studies using an independent gold standard (histopathology) we determined the sensitivity and specifi city of the index tests using the number of true positive (TP), false positive (FP), true negative (TN), and false negative (FN) results from the 2 x 2 contingency table. Furthermore, we calculated the ‘relative lesion detection’, defi ned as defi ned as the percentage of lesions scoring equally positive or negative with NAC versus AC images. We performed a subgroup analysis for diff erent locations of lesions and analyzed sensitivity, specifi city or relative lesion detection of NAC vs AC for lesion location in the head and neck region, the chest and the abdominopelvic region. In cases of discrepancy of relative lesion detection between NAC and AC, we extracted data to analyze whether this was related to lesion size and/or intensity.

The statistical diagnostic heterogeneity of the sensitivity and specifi city per index test across studies was tested by the chi-square test. In case of statistical heterogeneity of DH- or FR-FDG-PET imaging a random eff ect model for pooling was used, whereas in case of statistical homogeneity a fi xed eff ect model was used. Sensitivity, specifi city and relative lesion detection were pooled independently, all pooled estimates are presented with 95% confi dence intervals (95% CI). The logit transformed sensitivity, specifi city, relative lesion detection and corresponding 95% CI of the index tests were compared using z-test statistics. A p-value of less than 0.05 was considered signifi cant. All statistical analyses were performed with the SPSS 11.0.01 program for Windows (Version 11.0.1., SPSS Inc., Chicago, Illinois).

Results

The search strategy yielded 2202 references, 1477 in Medline and 725 in Embase on February 10th, 2006. Of the Embase references, 370 were also included in Medline, leaving a total of 1832 unduplicated references. On the basis of title and abstract alone, 1806 references were excluded. After review of the full text of the remaining 26 articles an additional 11 studies (3-12) proved to be ineligible because they did not perform a direct comparison of the yield of NAC versus AC images in oncological patients. One study (13) was excluded because it was published in abstract form only. Another study (14) was excluded because it was published in Japanese and was not readily translatable. Finally, the study of Hustinx et al (15), who evaluated the eff ect of attenuation correction in abdominal tumours for a NaI PET scanner, was excluded because no 2 x 2 contingency tables could be constructed. Eventually, we included 12 studies for review (16-27).

A summary of the methodological quality assessment results can be found in Table 2. Methodological quality was scored as negative when quality items were unclear or absent in



55

the original article. In a minority (4/12=33%) of studies, histology served as the reference test. However, 9 of the 12 studies provided a direct comparison of AC and NAC PET. In three studies blind measurement of reference test was performed without knowledge of index test. All but one study avoided verifi cation bias. In four studies, the index test(s) was evaluated independently of all clinical information. All studies provided information about the spectrum of diseases being evaluated and standardized the execution of the index test(s). Almost all studies (11/12=92%) described the demographics of the study population and inclusion criteria. However, only one study mentioned specifi c exclusion criteria. Six studies were prospective. Only two of the twelve studies specifi cally mentioned including consecutive patients and only three studies specifi cally described reproducibility of their results.

Table 2. Quality assessment of included studies

Meta-analysisThree FR-PET studies were eligible for pooling of sensitivity on a patient basis (22;24;25) . The pooled sensitivity for AC and NAC FR-PET was 64% (95% CI 52-74%) and 62% (95% CI 51-73%), respectively (n=182 patients). Two FR-PET studies provided data that allowed pooled analysis of specifi city (22;24). Weber et al (25) could not be included as there were no patients without disease. The pooled specifi city for AC and NAC FR-PET was 97% (95% CI 92-99%) and 99% (95% CI 95-100%), respectively (n=155 patients). For DH-PET, only one study provided data on sensitivity and specifi city.

Relative lesion detection for NAC versus AC PET was pooled for eleven studies, which are demonstrated in Figure 1. Lesion detection of NAC FR-PET versus AC FR-PET was 98% (95%

Study Year A1 A2 A3 A4 A5 B1 B2 B3 B4 B5 B6 C

Bleckmann 1999 + + + + - + - + - - + -

Chan 2001 + - + - - + + + - - + -

Delbeke 2001 + - + - + + + + - - + -

Even-Sapir 2004 + - + - + + + + - + + -

Kotzerke 1999 + - + - - + + + - + + +

Lonneux 1999 + - + - + + + - - + + +

Nakamoto 2002 + + + + + + + + - - + +

Reinhardt 2005 + - + - - + + + + - + -

Schauwecker 2003 + - - - - + + + - - + +

Weber 1999 + - + - + + + + - - + -

Zimny 1999 + + + + - + + + - - + -

Zimny 2003 + - + + + + + + - - + -


Chapter 4

56

confi dence interval: 96-99%) for n=1012 lesions (pooling of n=7 studies); 79% of which were classifi ed FDG positive at AC FR-PET. Lesion detection of NAC DH-PET versus AC DH-PET was 88% (95% CI: 81-94%) for n=288 lesions (pooling of n=4 studies); 74 % of which were classifi ed as FDG positive at AC DH-PET.

Figure 1. Pooled lesion detection of NAC versus AC images for FR-PET and DH-PET

In addition, we evaluated the relative lesion detection depending on body location (head/neck, chest, abdomen) in the four FR-PET - and in the three DH-PET studies that provided suffi ciently detailed information (17;18;26). The relative sensitivity and specifi city based on body location could not be calculated due to an insuffi cient number of studies. For FR-PET, we found similar relative lesion detection for the three body locations: 95% for head/neck (95% CI 84-98%, n=61 lesions), 97% for the chest (95% CI 94-99%, n=396 lesions), and 97% for the abdomen (95% CI 93-0.99%, n=205 lesions). For DH-PET, relative detection rates were not signifi cantly diff erent for the various body sites: 78% in the abdomen (95% CI 65-88%; 53 lesions), 84% in the chest (95% CI 74-91%; 136 lesions) and 90% in the head/neck area (95% CI 73-97%; 38 lesions). However, in chest (p=0.000089) and abdomen (p=0.0037), the relative detection rates with NAC (vs. AC) for DH-PET were signifi cantly lower than those obtained with FR-PET.

A comprehensive analysis of the potential association of relative detection and lesion size and/or intensity, for lesions with discrepant AC and NAC results was not possible due to a lack of detailed information. We summarized the results in Table 3: fi ndings of Bleckmann et al and Reinhardt et al (16;23) suggest that AC and NAC discrepancies may relate to (intrapulmonary) lesion size with more discrepancies occurring with smaller lesions at the subcentimeter level (together 3% of lesions were correctly detected with NAC and not with AC). The single discrepant lesion in the study of Weber et al concerned a < 1 cm lesion in the mediastinum (25). However, the discrepant lesions in the studies of Nakamoto et al, Schauwecker et al, and Delbeke et al included both small and moderate sized lesions (in relation to lesions included each study) (18;22;24). In the study



57

of Schauwecker et al, the discrepant lesions demonstrated SUVmax values ranging from 1.8 to 2.6, while in the study van Delbeke et al, the two discrepant lesions demonstrated only mildly enhanced uptake on AC images and equivocal uptake in NAC images.

Table 3. Evaluation of discrepant lesions between AC and NAC images with respect to lesion size and intensity*

Discussion

The cumulated evidence summarized in this systematic review of oncological FDG imaging studies suggests that the accuracy of attenuation- and non-attenuation corrected full ring PET are similar. However, with dual-head coincidence imaging NAC-corrected images detect 12% less lesions than AC corrected ones, without prominent diff erences between body areas.

Even though in the nuclear medicine fi eld attenuation correction is generally seen as an important issue, it is surprising to fi nd that several large systematic reviews did not thoroughly consider this as a potential eff ect-modifi er. Gould et al. performed systematic reviews on FDG PET in pulmonary lesions (28) and mediastinal lymph node staging in non-small cell lung cancer (29): in the former review the item was not mentioned, and in the latter, attenuation correction was an item of study quality, but no analysis of potential impact was performed.

The choice of the reference test is obviously relevant in studies on test accuracy. In oncology, histopathology is the typical endpoint. Of the 12 eligible studies, four used histology as an independent gold standard. Meta-analysis of sensitivity and specifi city was only possible for FR-

Study Camera

Type

Number of

Discrepant

Lesions

Size

Range

Intensity (Semiquantitative or

Qualitative)

Bleckmann et al FR-PET 5 < 1 cm Not given

Nakamoto et al* FR-PET 1 1.8 cm Not given

Reinhardt et al FR-PET 6 0.5-1.1 cm 79/174 lesions demonstrated

discrepancy in qualitative

lesion intensity: 72/174 lesions

demonstrating higher intensity (i.e.

better visibility) on NAC images

Schauwecker et al* FR-PET 4 0.8-3.9 cm3 1.8-2.6 (SUVmax)

Weber et al FR-PET 1 0.8 cm Not given

Delbeke et al DH-PET 2 1.0-3.0 cm ‘Mild uptake’ at AC, ‘equivocal uptake’

at NAC

*Histology used as gold standard and detailed information given only for true positive lesions


Chapter 4

58

PET, and we found no signifi cant diff erence for either measure. We chose to use AC detection rate as an alternative reference test which defi nes the relative lesion detection of NAC vs. AC images. This choice theoretically biases in favour of AC: Bleckmann et al and Reinhardt et al (16;23) reported about an average of 3% more true positive lesions with NAC FR-PET. However, we expect that the resulting error is small since in the comparison with histopathology, false positive rates were quite low for either modality. Despite this theoretical negative bias towards NAC, similar lesion detection rates were observed with both AC and NAC for FR-PET. Hence, attenuation correction may not contribute to the detection of malignancy using FR-PET. Conversely, with DH-PET AC-images demonstrated a signifi cantly higher detection rate as opposed to NAC images, which is surprising given that AC images are usually signifi cantly noisier than NAC images. We postulate that this may be secondary to diff erences in reconstruction/fi ltering algorithms.

In addition, there are limitations associated with performing a meta-analysis and data pooling such as the homogeneity of the data and the quality of the published studies. Homogeneous data has higher statistical strength than heterogeneous data. The data in our study was heterogeneous so that we used a random eff ect model for pooling. In addition, the statistical strength of the meta-analysis is limited by the quality of the published studies included in it. As mentioned earlier and summarized in Table 3, the studies had several quality limitations. Finally, meta-analyses are limited by publication bias which biases towards publication of favourable results or popular subjects.

We were surprised by the limited number of good comparative studies evaluating the value of attenuation correction. It appears that attenuation correction has been accepted as the standard of practice without sound scientifi c evidence to support it.

The introduction of PET/CT machines has made the time constraints associated with transmission scanning less of an issue. However, PET/CT is not a panacea; x-ray transmission scanning has its own problems and numerous PET/CT publications have demonstrated artifact that can be introduced with x-ray transmission scanning (5;30-45). Furthermore, in the study of Reinhardt et al (23) a signifi cantly improved visibility was demonstrated for 41% of lung metastases with NAC images as opposed to CT-AC images. This higher visibility for NAC images was even more pronounced for lesions smaller than 1 cm. These fi ndings underline that even as PET/CT use becomes more widespread, evaluation of both NAC and AC images should remain an integral part of image interpretation, and not just to recognize image artifacts. At the same time, NAC-AC discrepancies at PET-CT off er an obvious opportunity for further investigation.

Conclusion

In this meta-analysis we found no signifi cant diff erence in sensitivity, specifi city, nor relative lesion detectability between attenuation corrected and non-attenuation corrected full ring FDG PET. However, attenuation correction improved lesion detection for dual head coincidence imaging.



59

References

(1) Mijnhout GS, Riphagen II, Hoekstra OS. Update of the FDG PET search strategy. Nucl Med Commun 2004;

25(12):1187-1189.

(2) The Cochrane Manual. The Cochrane Collaboration. www.cochrane.org/admin/manual.htm.

(3) Bengel FM, Ziegler SI, Avril N, Weber W, Laubenbacher C, Schwaiger M. Whole-body positron emission

tomography in clinical oncology: comparison between attenuation-corrected and uncorrected images. Eur

J Nucl Med 1997; 24(9):1091-1098.

(4) Etchebehere EC, Macapinlac HA, Gonen M, Humm K, Yeung HW, Akhurst T et al. Qualitative and quantitative

comparison between images obtained with fi ltered back projection and iterative reconstruction in prostate

cancer lesions of (18)F-FDG PET. Q J Nucl Med 2002; 46(2):122-130.

(5) Goerres GW, Hany TF, Kamel E, von Schulthess GK, Buck A. Head and neck imaging with PET and PET/CT:

artefacts from dental metallic implants. Eur J Nucl Med Mol Imaging 2002; 29(3):367-370.

(6) Imran MB, Kubota K, Yamada S, Fukuda H, Yamada K, Fujiwara T et al. Lesion-to-background ratio in

nonattenuation-corrected whole-body FDG PET images. J Nucl Med 1998; 39(7):1219-1223.

(7) Kamel E, Hany TF, Burger C, Treyer V, Lonn AH, von Schulthess GK et al. CT vs 68Ge attenuation correction in

a combined PET/CT system: evaluation of the eff ect of lowering the CT tube current. Eur J Nucl Med Mol

Imaging 2002; 29(3):346-350.

(8) Laubenbacher C, Saumweber D, Wagner-Manslau C, Kau RJ, Herz M, Avril N et al. Comparison of fl uorine-18-

fl uorodeoxyglucose PET, MRI and endoscopy for staging head and neck squamous-cell carcinomas. J Nucl

Med 1995; 36(10):1747-1757.

(9) Pitman AG, Hicks RJ, Binns DS, Ware RE, Kalff V, McKenzie AF et al. Performance of sodium iodide based

(18)F-fl uorodeoxyglucose positron emission tomography in the characterization of indeterminate

pulmonary nodules or masses. Br J Radiol 2002; 75(890):114-121.

(10) Stevens H, Bakker PF, Schlosser NJ, van Rijk PP, de Klerk JM. Use of a dual-head coincidence camera and

18F-FDG for detection and nodal staging of non-small cell lung cancer: accuracy as determined by 2

independent observers. J Nucl Med 2003; 44(3):336-340.

(11) Turlakow A, Larson SM, Coakley F, Akhurst T, Gonen M, Macapinlac HA et al. Local detection of prostate

cancer by positron emission tomography with 2-fl uorodeoxyglucose: comparison of fi ltered back projection

and iterative reconstruction with segmented attenuation correction. Q J Nucl Med 2001; 45(3):235-244.

(12) Zasadny KR, Kison PV, Quint LE, Wahl RL. Untreated lung cancer: quantifi cation of systematic distortion of

tumor size and shape on non-attenuation-corrected 2-[fl uorine-18]fl uoro-2-deoxy-D-glucose PET scans.

Radiology 1996; 201(3):873-876.

(13) Wong T, Coleman R, Hagge R, Borges-Neto S, Hanson M. 27. PET Image Interpretation. Attenuation-

Corrected (ATN) Vs Non-Attenuation Corrected (NATN) Images. Clin Positron Imaging 2000; 3(4):181.

(14) Yasuda S, Ide M, Takagi S, Shohtsu A, Mitomi T, Kobayashi S et al. [Cancer detection with whole-body FDG

PET images without attenuation correction]. Kaku Igaku 1996; 33(4):367-373.

(15) Hustinx R, Dolin RJ, Benard F, Bhatnagar A, Chakraborty D, Smith RJ et al. Impact of attenuation correction

on the accuracy of FDG-PET in patients with abdominal tumors: a free-response ROC analysis. Eur J Nucl

Med 2000; 27(9):1365-1371.


Chapter 4

60

(16) Bleckmann C, Dose J, Bohuslavizki KH, Buchert R, Klutmann S, Mester J et al. Eff ect of attenuation correction

on lesion detectability in FDG PET of breast cancer. J Nucl Med 1999; 40(12):2021-2024.

(17) Chan WL, Freund J, Pocock NA, Szeto E, Chan F, Sorensen BJ et al. Coincidence detection FDG PET in the

management of oncological patients: attenuation correction versus non-attenuation correction. Nucl Med

Commun 2001; 22(11):1185-1192.

(18) Delbeke D, Martin WH, Patton JA, Sandler MP. Value of iterative reconstruction, attenuation correction, and

image fusion in the interpretation of FDG PET images with an integrated dual-head coincidence camera

and X-ray-based attenuation maps. Radiology 2001; 218(1):163-171.

(19) Even-Sapir E, Yuzefovich B, Miller E, Bouhnik JP, Zak O, Lerman H et al. Coincidence imaging using 2

dual-head gamma-camera systems, with and without attenuation correction. J Nucl Med Technol 2004;

32(4):190-197.

(20) Kotzerke J, Guhlmann A, Moog F, Frickhofen N, Reske SN. Role of attenuation correction for fl uorine-18

fl uorodeoxyglucose positron emission tomography in the primary staging of malignant lymphoma. Eur J

Nucl Med 1999; 26(1):31-38.

(21) Lonneux M, Borbath I, Bol A, Coppens A, Sibomana M, Bausart R et al. Attenuation correction in whole-body

FDG oncological studies: the role of statistical reconstruction. Eur J Nucl Med 1999; 26(6):591-598.

(22) Nakamoto Y, Chang AE, Zasadny KR, Wahl RL. Comparison of attenuation-corrected and non-corrected FDG-

PET images for axillary nodal staging in newly diagnosed breast cancer. Mol Imaging Biol 2002; 4(2):161-169.

(23) Reinhardt MJ, Wiethoelter N, Matthies A, Joe AY, Strunk H, Jaeger U et al. PET recognition of pulmonary

metastases on PET/CT imaging: impact of attenuation-corrected and non-attenuation-corrected PET

images. Eur J Nucl Med Mol Imaging 2006; 33(2):134-139.



2003; 24(3):281-289.

(25) Weber WA, Neverve J, Sklarek J, Ziegler SI, Bartenstein P, King B et al. Imaging of lung cancer with fl uorine-

18 fl uorodeoxyglucose: comparison of a dual-head gamma camera in coincidence mode with a full-ring

positron emission tomography system. Eur J Nucl Med 1999; 26(4):388-395.

(26) Zimny M, Kaiser HJ, Cremerius U, Reinartz P, Schreckenberger M, Sabri O et al. Dual-head gamma camera

2-[fl uorine-18]-fl uoro-2-deoxy-D-glucose positron emission tomography in oncological patients: eff ects of

non-uniform attenuation correction on lesion detection. Eur J Nucl Med 1999; 26(8):818-823.

(27) Zimny M, Hochstenbag M, Lamers R, Reinartz P, Cremerius U, ten Velde G et al. Mediastinal staging of lung

cancer with 2-[fl uorine-18]-fl uoro-2-deoxy-D-glucose positron emission tomography and a dual-head

coincidence gamma camera. Eur Radiol 2003; 13(4):740-747.



(29) Gould MK, Kuschner WG, Rydzak CE, Maclean CC, Demas AN, Shigemitsu H et al. Test performance of

positron emission tomography and computed tomography for mediastinal staging in patients with non-

small-cell lung cancer: a meta-analysis. Ann Intern Med 2003; 139(11):879-892.

(30) Antoch G, Freudenberg LS, Egelhof T, Stattaus J, Jentzen W, Debatin JF et al. Focal tracer uptake: a potential

artifact in contrast-enhanced dual-modality PET/CT scans. J Nucl Med 2002; 43(10):1339-1342.



61

(31) Beyer T, Antoch G, Blodgett T, Freudenberg LF, Akhurst T, Mueller S. Dual-modality PET/CT imaging: the

eff ect of respiratory motion on combined image quality in clinical oncology. Eur J Nucl Med Mol Imaging

2003; 30(4):588-596.

(32) Bujenovic S, Mannting F, Chakrabarti R, Ladnier D. Artifactual 2-deoxy-2-[(18)F]fl uoro-D-glucose localization

surrounding metallic objects in a PET/CT scanner using CT-based attenuation correction. Mol Imaging Biol

2003; 5(1):20-22.

(33) Cohade C, Wahl RL. Applications of positron emission tomography/computed tomography image fusion

in clinical positron emission tomography-clinical use, interpretation methods, diagnostic improvements.

Semin Nucl Med 2003; 33(3):228-237.

(34) DiFilippo FP, Brunken RC. Do implanted pacemaker leads and ICD leads cause metal-related artifact in

cardiac PET/CT? J Nucl Med 2005; 46(3):436-443.

(35) Goerres GW, Ziegler SI, Burger C, Berthold T, von Schulthess GK, Buck A. Artifacts at PET and PET/CT caused

by metallic hip prosthetic material. Radiology 2003; 226(2):577-584.

(36) Goerres GW, Burger C, Kamel E, Seifert B, Kaim AH, Buck A et al. Respiration-induced attenuation artifact at

PET/CT: technical considerations. Radiology 2003; 226(3):906-910.

(37) Goerres GW, Schmid DT, Eyrich GK. Do hardware artefacts infl uence the performance of head and neck PET

scans in patients with oral cavity squamous cell cancer? Dentomaxillofac Radiol 2003; 32(6):365-371.

(38) Gorospe L, Raman S, Echeveste J, Avril N, Herrero Y, Herna NS. Whole-body PET/CT: spectrum of

physiological variants, artifacts and interpretative pitfalls in cancer patients. Nucl Med Commun 2005;

26(8):671-687.

(39) Kamel EM, Burger C, Buck A, von Schulthess GK, Goerres GW. Impact of metallic dental implants on CT-

based attenuation correction in a combined PET/CT scanner. Eur Radiol 2003; 13(4):724-728.

(40) Nehmeh SA, Erdi YE, Kalaigian H, Kolbert KS, Pan T, Yeung H et al. Correction for oral contrast artifacts in CT

attenuation-corrected PET images obtained by combined PET/CT. J Nucl Med 2003; 44(12):1940-1944.

(41) Osman MM, Cohade C, Nakamoto Y, Wahl RL. Respiratory motion artifacts on PET emission images obtained

using CT attenuation correction on PET-CT. Eur J Nucl Med Mol Imaging 2003; 30(4):603-606.

(42) Otsuka H, Graham MM, Kubo A, Nishitani H. The eff ect of oral contrast on large bowel activity in FDG-PET/

CT. Ann Nucl Med 2005; 19(2):101-108.

(43) Papathanassiou D, Liehn JC, Bourgeot B, Amir R, Marcus C. Cesium attenuation correction of the liver

dome revealing hepatic lesion missed with computed tomography attenuation correction because of the

respiratory motion artifact. Clin Nucl Med 2005; 30(2):120-121.

(44) Sarikaya I, Yeung HW, Erdi Y, Larson SM. Respiratory artefact causing malpositioning of liver dome lesion in

right lower lung. Clin Nucl Med 2003; 28(11):943-944.

(45) Sureshbabu W, Mawlawi O. PET/CT Imaging Artifacts. J Nucl Med Technol 2005; 33(3):156-161.


Chapter 4

62

Appendix 1. Detailed search strategy(“x ray” OR x-ray OR cine-ct OR “Tomography, X-Ray Computed”[mesh] OR transmission OR at-tenuat* OR nonattenuat* OR ac[tw] OR nac[tw] OR nonac OR germanium OR ge[tw] OR gallium OR ga[tw] OR cesium OR cs[tw]) AND (oncolog* OR cancer* OR neoplas* OR neoplasms[mesh] OR tumour* OR tumor OR tumors OR carcinom* OR melanom* OR lymphom* OR malig-nan*) AND (Deoxyglucose[mesh] OR Deoxyglucose[tw] OR Desoxyglucose[tw] OR Desoxy-glucose[tw] OR deoxy-glucose[tw] OR Deoxy-d-glucose[tw] OR Desoxy-d-glucose[tw] OR fl uorodeoxyglucose[tw] OR Fluorodesoxyglucose[tw] OR fl uorodeoxy-glucose[tw] OR Fluorode-oxy-d-glucose[tw] OR Fluoro-d-glucose[tw] OR Fludeoxyglucose[tw] OR Fluordeoxyglucose[tw] OR Fluordesoxyglucose[tw] OR 18fl uorodeoxyglucose[tw] OR 18fl uorodeoxy-glucose OR 18fl uorodesoxyglucose[tw] OR 18Fluordeoxyglucose[tw] OR fdg*[tw] OR 18fdg*[tw] OR 18f-dg*[tw] OR 2deoxyglucose[tw] OR 2deoxy-d-glucose[tw] OR ((fl uor[tw] OR fl uoro[tw] OR 18fl uor[tw] OR 18fl uoro[tw]) AND glucose[tw])) AND (pet[tw] OR pet/* OR petscan* OR “Tomography, emission-computed”[mesh] OR (positron[tw] AND emission[tw] AND tomograph*[tw]) OR (emission[tw] AND computed[tw] AND tomograph*[tw])) NOT (animal[mesh] NOT human[mesh])


C h a p t e r

Initial experience with a prototype dual-crystal

(LSO/NaI) dual head coincidence camera in oncology

5Urvi JoshiOtto S HoekstraRonald BoellaardEmile FI ComansPieter GHM RaijmakersRik J PijpersSteven D MillerGerrit JJ TeuleArthur van Lingen

In: European Journal of Nuclear Medicine and Molecular Imaging2004;31(4):596-598


Abstract

The aim of this study was to evaluate the in vivo performance of a prototype dual-crystal [lutetium oxyorthosilicate (LSO)/sodium iodide (NaI)] dual-head coincidence camera (DHC) for PET and SPET (LSO-PS), in comparison to BGO-PET with fl uorine-18 fl uorodeoxyglucose (FDG) in oncology. This follows earlier reports that LSO-PS has noise-equivalent counting (NEC) rates comparable to partial ring BGO-PET, i.e. clearly higher than standard NaI DHCs. Twenty-four randomly selected oncological patients referred for whole-body FDG-PET underwent BGO-PET followed by LSO-PS. Four nuclear medicine physicians were randomized to read a single scan modality, in terms of lesion intensity, location and likelihood of malignancy. BGO-PET was considered the gold standard. Forty-eight lesions were classifi ed as positive with BGO-PET, of which LSO-PS identifi ed 73% (95% CI 60–86%). There was good observer agreement for both modalities in terms of intensity, location and interpretation. Lesions were missed by LSO-PS in 13 patients in the chest (n=6), neck (n=3) and abdomen (n=4). The diameter of these lesions was estimated to be 0.5–1 cm. Initial results justify further evaluation of LSO-PS in specifi c clinical situations, especially if a role as an instrument of triage for PET is foreseen.


Initial experience with a prototype dual-crystal (LSO/NaI) dual head coincidence camera in oncology

65

Introduction

In vitro data indicated that a prototype dual-crystal [lutetium oxyorthosilicate (LSO)/sodium iodide (NaI)] dual-head coincidence camera (DHC) for PET and SPET (LSO-PS) might outperform NaI DHC (1). Furthermore, LSO-PS allows for simultaneous fl uorodeoxyglucose (FDG)/technetium-99m cardiac SPET (2) and relatively fast (30 min) whole-body acquisitions (including simultaneous transmission scanning).

Since DHC has a lower sensitivity than full-ring PET (3), DHC might be used for triage of patients: after decisively positive fi ndings at DHC, e.g. a distant metastasis in a patient with non-small cell lung cancer, full-ring PET is unnecessary. In this context, head-to-head comparisons using PET as the reference standard allow rapid assessment of relative performance. However, the NaI DHC literature with this design suff ers from lack of blinding and standardisation (4).

Our aim was to perform an initial evaluation of the potential of LSO-PS in oncology, using a prospective and properly blinded direct comparison with full-ring BGO-PET.


Within 1 month, 30 randomly selected patients referred for FDG-PET were included, after informed consent had been obtained. Whole-body BGO-PET scans were acquired in 2D mode using 5-min emission scanning/bed position, yielding non-attenuation-corrected (NAC) images. LSO-PS scans were done in (the standard) 3D mode, comprising two bed positions (eff ective scanned trajectory 64 cm), scanned with simultaneous emission-transmission scans at 10 min per bed position, yielding NAC as well as attenuation-corrected (AC) images.

Figure 1. Noise-equivalent count rate (NEC) of LSO-PS (3D mode), NaI DHC (3D mode) and full-ring BGO-PET in 2D

mode, measured according to NEMA NU-2 2001


Chapter 5

66

Acquisitions with BGO-PET (ECAT HR+) and LSO-PS (CTI PET) were done 60 and 120 min after injection of 370 MBq FDG, respectively. LSO-PS was timed so that activity concentrations corresponded with expected peak NEC conditions (Figure 1). Data were reconstructed with OSEM (2 iterations, 16 subsets) and a 5-mm full-width at half-maximum Gaussian smoothing fi lter. Scans were viewed on a PC running Matlab 5.3.

Data analysis Datasets of six patients were randomly chosen to obtain experience with LSO-PS, leaving 24 evaluable cases. Four experienced BGO-PET observers were randomised to independently read a single NAC scan modality per patient. To assess the impact of AC in LSO-PS, two readers independently compared detection at AC with NAC using the same scoring system as described below.

Focally enhanced uptake at either modality outside the normal FDG biodistribution was considered abnormal. The clinical information as provided on the PET request form was available (but CT scans were not available for direct comparison). Foci of uptake were classifi ed as benign, equivocal or malignant, with BGO-PET serving as the reference standard for malignancy. Moreover, the presumed anatomical localisation was recorded, reasoning that this is also a function of image quality. Discrepancies were resolved by consensus, if necessary using a third reader.

Relative intensity versus background was assessed subjectively (0= absent, 1= faint, 2= moderate, 3= intense), and in lesions <30 mm [estimated from the number of involved axial (4.5 mm) slices on BGO-PET] also quantitatively using an automated algorithm to search for the pixel with the highest contrast (maximum vs minimum) within 15 mm from the centre of the lesion.

Results

The 24 evaluated patients (mean age 59 years, range 29–81 years; 15 males) had neoplasms of the lung (10), colon (4), breast (3), thymus (1) and salivary gland (1), as well as gynaecological cancers (2), unknown primary tumours (2) and malignant lymphoma (1).

At BGO-PET, 48 lesions were positive, situated in the neck (9), chest (26) and abdomen (13). Their median size was 20 mm [interquartile range (IQR): 9–31, range 5–135]. In two patients with multiple abdominal lesions, foci could not be analysed separately, so they were classifi ed as para-aortic, para-iliac and pre-aortic. LSO-PS identifi ed 73% of BGO-PET-positive foci [95% confi dence interval (CI): 60–86%, Table 1]: 33 were interpreted as malignant, one as equivocal and one as benign. There were no false positives with LSO-PS. Lesions were missed in 13 patients in the chest (n=6, Fig. 2), neck (n=3) and abdomen (n=4), corresponding to a relative regional sensitivity of LSO-PS versus BGO-PET of 77% in the chest (20/26), 67% in the neck (6/9, Fig. 3) and 69% in the abdomen (9/13). A direct comparison of NAC and AC LSO-PS revealed identical detection rates.



67

Table 1. Concordance between BGO-PET and LSO-PS

Figure 2. Spatial resolution limitations: LSO-PS (upper row) demonstrates FDG uptake centrally in the left lung

(primary tumour). However, it is diffi cult to decide whether there is adjacent mediastinal lymph node involvement.

BGO-PET (lower row) clearly demonstrates a separate focus in a paratracheal node (arrow)

Figure 3. BGO-PET (lower row) demonstrates focally increased FDG uptake in a malignant right cervical lymph node with a small focus ≤ 1 cm (arrow) in the pharyngeal region suggesting a primary tumour which was also seen with LSO-PS (upper row)

LSO:Total

Not seen Benign Equivocal Malignant

PET

Benign 13 2 15

Equivocal 6 3 2 11

Malignant 13 1 1 33 48

Total 32 3 4 35 74


Chapter 5

68

Concordance of LSO-PS and BGO-PET was inversely related to size: in 40 measurable BGO-PET foci, LSO-PS was positive in 17/20 lesions >20 mm, in 6/8 lesions 10–20 mm and in 5/12 lesions <10 mm. Subjectively, contrast (intensity vs background) was highest for BGO-PET (Wilcoxon’s signed rank test, P=0.05). Measured contrast in lesions <30 mm with BGO-PET was 28.1 (IQR 11.8–88.6), versus 4.7 (IQR 3.2–6.8) and 4.7 (IQR 2.7–7.6) for NAC and AC LSO-PS, respectively. The high contrasts measured for BGO-PET were due to the algorithm searching for the maximum and minimum in an area, rather than maximum and average in a surrounding area.

Observer variability of intensity classifi cation tended to be higher with LSO-PS [intraclass correlation coeffi cient (ICC): 0.56 (95%CI 0.29–0.74) vs 0.77 (95%CI 0.63–0.86) with BGO-PET], but

interpretation demonstrated a similar level of agreement (ICC BGO-PET 0.67, 95%CI 0.53–0.77 vs 0.58, 95% CI 0.39–0.72, for LSO-PS); major diff erences (two categories within the three-point scoring system) were found in 13% and 15% with BGO-PET and LSO-PS, respectively. Observers localized foci identically in 93% and 91% with BGO-PET and LSO-PS, respectively, with discrepancies at the hilar and mediastinal level, skin/breast and larynx/tongue base.

Discussion

NEC curves of LSO-PS and NaI DHC suggest a better performance of LSO-PS since its peak NEC value is about three times higher. In vivo, this might translate into improved signal-to-noise ratios and lesion detectability. Our initial in vivo results demonstrate suffi cient image quality and detection effi ciency to justify further investigation.

At fi rst glance, concordance appears to be similar to that between NaI-DHC and PET studies (74–79% (4). However, comparison is impaired by diff erent patient spectra (lesion sizes, tumour types) and methodology (image acquisition, study design). Especially inappropriate blinding induces signifi cant bias in favour of DHC (up to 25%) (5). Even though we avoided this problem, our study has several limitations. First, based on NEC rates, we chose to perform LSO-PS after BGO-PET, so that rising lesion-to-background ratios over time may have biased results in favour of LSO-PS. Second, NAC rather than AC BGO-PET was the gold standard. Our observers had extensive experience with NAC PET but not with AC PET. We anticipated that using AC PET as the gold standard would induce signifi cant observer variation, which might dominate the evaluation instead of the actual diff erence between devices. Finally, LSO-PS acquisitions were relatively short compared with other DHC studies (10 vs. 17–30 min, respectively).

We conclude that these initial results with the LSO-PS camera justify further research on its diagnostic accuracy and clinical potential, at least in settings where its application as triage for full-ring PET is considered.



69

References



(2) Pichler BJ, Miller SM, Hamill JJ, Gremillion T, Weber WA, Bendriem B et al. Evaluation of the NaI-LSO-hybrid

detector PET-SPECT system: dual isotope scans and fi rst patient studies. Eur J Nucl Med 29[S1], S109. 2002.

(3) Ak I, Blokland JA, Pauwels EK, Stokkel MP. The clinical value of 18F-FDG detection with a dual-head

coincidence camera: a review. Eur J Nucl Med 2001; 28(6):763-778.

(4) Haslinghuis-Bajan LM, Hooft L, van Lingen A, van Tulder M, Deville W, Mijnhout GS et al. Rapid evaluation

of FDG imaging alternatives using head-to-head comparisons of full ring and gamma camera based PET

scanners--a systematic review. Nuklearmedizin 2002; 41(5):208-213.



38(4):108-114.



C h a p t e r

Evaluation of pulmonary nodules: comparison of a prototype dual crystal (LSO/NaI) dual head coincidence camera and full ring positron emission tomography (PET)

6Urvi JoshiPieter GHM RaijmakersArthur van LingenEmile FI ComansRik PijpersGerrit JJ TeuleOtto S Hoekstra

In: European Journal of Radiology 2005;55(2):250-254


Abstract

Purpose: To determine the concordance of a prototype dual head coincidence camera (LSO-PS) and full ring PET (BGO-PET) using 18F-fl uorodeoxyglucose (FDG) in the evaluation of pulmonary nodules (PNs).

Materials and methods: Patients referred for evaluation of ≤3 PNs (≤3 cm diameter) were prospectively studied on the same day with both BGO-PET and LSO-PS. Imaging was performed at 60 and 120 min after injection of 370 MBq FDG, respectively. Images were independently interpreted by four observers with each observer blinded to the other modality for the same patient. Lesions were scored in terms of relative intensity versus background. Non-attenuation corrected (nonAC) BGO-PET was used as the reference test.

Results: Forty-seven patients with 54 PNs (mean diameter 1.7 cm, S.D. 0.7) were included. Twelve nodules were in the ≤1.0 cm – 27 in the 1.1–2.0 cm – and 15 in the 2.1–3.0 cm range. Interobserver agreement was similar for both FDG imaging modalities. Using a sensitive assessment strategy with LSO-PS (≥ faint intensity deemed positive), there was a 97% (38/39, 95%CI 87–100%) concordance with BGO-PET and one false positive case with LSO-PS. Conservative reading (moderate or intense intensity deemed positive) resulted in a 92% (36/39, 95%CI 80–97%) concordance with BGO-PET, without false positives. The only lesion missed by LSO-PS using both assessment strategies involved a nodule 1.5 cm diameter that demonstrated moderate increased FDG uptake on BGO-PET.

Conclusion: Depending on the test positivity criteria, LSO-PS demonstrates a high concordance (92–97%) with nonAC BGO-PET for the characterization of pulmonary nodules.


Evaluation of pulmonary nodules

73

Introduction

A pulmonary nodule is a well-defi ned parenchymal lung lesion with a diameter of ≤3–4 cm (1). The probability of malignancy varies widely (10–70%) (2;3) and depends on nodule size, age, smoking status and radiologic characteristics (4). Full ring positron emission tomography (PET) with 18F-fl uorodeoxyglucose (FDG) in such nodules has demonstrated suffi cient accuracy (5) for PET to obtain a prominent role in cost-eff ectiveness modeling (6), and to be included in recent guidelines for the management of pulmonary nodules (7;8). The bottomline is that absence of FDG uptake in a radiologically indeterminate nodule is a strong argument against malignancy.

Dual head coincidence cameras (DHC) have also been used to evaluate pulmonary lesions (9;10). However, no study evaluating the performance of DHC cameras has, to the best of our knowledge, specifi cally focused on the characterization of smaller pulmonary nodules (i.e. nodules ≤3 cm diameter). From the literature, we know that the main shortcoming of DHC systems is their diminished sensitivity for smaller lesions (11). At the same time, a new type of DHC camera was developed consisting of a dual layer of phoswhich crystals: LSO and NaI, designed for both PET and SPECT imaging (LSO-PS). Initial in vitro evaluation has demonstrated physical performance parameters (noise equivalent count rate, spatial resolution) comparable to a partial ring BGO-PET system (12;13).

It is expected that FDG PET will have an increasingly important role in evaluation of solitary pulmonary nodules (SPN), in part amplifi ed by CT lung cancer screening programs. However, PET capacity is still limited and costs are considerable. Alternatively, DHC might be used as an instrument of triage for FDG PET: if the sensitivity of DHC is comparable to that of PET, SPN’s negative at DHC would justify a policy similar to that following a negative FDG PET. Following this line of thought, the FDG PET scan should serve as the reference technique for comparison. Therefore, we performed a prospective, blinded comparison of patients referred for the evaluation of pulmonary nodules (≤3 cm) utilizing the prototype dual crystal LSO/NaI DHC camera for PET and SPECT (LSO-PS) as compared to full ring PET (BGO-PET).


Between March 2002 and April 2003, we performed a prospective study in which consecutive patients referred for evaluation of maximally three pulmonary nodules ≤3 cm in diameter on CT, not previously proven to be malignant, were included. All patients gave informed consent and the study was approved by the ethics committee of our institution.

Patients underwent scanning with both full ring BGO-PET and LSO-PS on the same day at fi xed time points (60 and 120 min, respectively, after a single dose of circa 370 MBq FDG). Patients had fasted a minimum of 6 h prior to FDG administration and their serum glucose was <10 mmol/l. Lesion size (short axis diameter) was measured using chest CT scans performed around the time of PET. If the chest CT scan was not available at the time of patient inclusion, we relied upon


Chapter 6

74

the information provided at the referral forms to judge eligibility. In such cases, lesion size could only be verifi ed after the patient had already enrolled in the study and FDG imaging had been performed. Patients with pulmonary lesions larger than 3 cm or with lesions that could not be precisely measured were excluded.

Image acquisition and reconstructionBGO-PET scans were performed on an ECAT HR+ scanner (CTI PET systems), involving a whole body acquisition in 2D mode with an emission scan of 5 min/bed position. LSO-PS scans were performed on a prototype dual crystal, dual head coincidence system (CTI PET systems). This scanner has camera detector heads consisting of a dual layer of phoswich crystals, lutetium oxyorthosilicate (LSO) and sodium iodide (NaI), each 10 mm thick. The useable camera fi eld of view is 35 cm × 50 cm. The system has a coincidence timing window of 6 ns. Transmission scanning is performed with seven collimated germanium-68 point sources, fi tted to the yokes of the detector heads. Cross talk between the transmission sources and the emission window may theoretically degrade image quality, but this has been reported to be minimal (13). Further technical details of the LSO-PS camera have been published elsewhere (12;14). The LSO-PS acquisition consisted of simultaneous 3D emission and transmission scans extending inferiorly from the neck (two bed positions of 10 min each). Typically, the scan trajectory extended from the thorax to the level of the urinary bladder.

The use of attenuation correction (AC) is controversial and our clinical routine whole body protocol with the BGO-PET system does not include attenuation correction (15). As such, we decided to focus the methodologically most stringent analysis (with respect to blinding of observers) on a comparison of non-attenuation corrected images. However, we performed an additional evaluation of AC LSO PET-PS images (visual and semiquantitative). All scans were reconstructed using an OSEM algorithm (ordered subsets expectation maximum with two iterations and 16 ordered subsets) with a 5 mm slice thickness.

Image analysisQualitative analysisAll images (5 mm slices in axial, coronal and sagittal planes, were viewed with the same software on a single viewing station by four experienced nuclear medicine physicians. Each physician was randomly assigned to independently interpret a single scan modality of each patient, so that they were blinded to the other scan modality of the same patient. Each scan was read by two independent observers and interobserver discrepancy was resolved through a third observer. The observers had access to information about lesion size and its localization.

Each relevant focal pulmonary abnormality was interpreted as visually related to mediastinal background activity and scored using a four-point scale (absent, faint, moderate or intense) (16). Lesions demonstrating at least moderate increased FDG uptake on BGO-PET were considered as suspect for malignancy. For the purpose of this study, this threshold was considered to be the reference test. Concordance between full ring PET and LSO-PS was calculated using a ‘conservative’ as well as a ‘sensitive’ assessment strategy, in which lesions with faint enhanced uptake at LSO-PS were classifi ed as normal and positive, respectively (17).



75

Semiquantitative analysisImage contrast was determined using an algorithm utilized by Weber et al. (18) in which circular regions of interest (ROI) were placed over the primary tumor in the plane with the highest FDG uptake. The reference (background) ROI was determined by drawing a large irregular ROI comprising the contralateral lung at the same axial level. Image contrast was determined using the ratio between the mean counts in the tumor versus background ROIs.

Statistical analysisInterobserver agreement was assessed with intraclass correlation coeffi cients (ICCs). Values of the ICC range from 0 to 1, with values close to 0 indicating poor agreement. General linear model analysis for repeated measures was used to analyze diff erences in semiquantitative measures, and Wilcoxon’s signed rank test to analyze potential diff erences of lesion intensity using the summed scores of the two observers assigned to either modality. Signifi cance was set at 5%.

Results

Patient populationPaired full ring PET and LSO-PS scans were obtained in 62 patients. After reading the CT scans, 15 patients did not fulfi l the eligibility criteria (pulmonary lesions >3 cm, size not measurable), leaving a total of 47 patients with 54 evaluable pulmonary nodules. Three patients had presented with two pulmonary nodules, two with three nodules, and the remaining ones with a solitary lesion.

The mean patient age was 65 years (S.D. ± 13) and 35% were females. Six patients (13%) had a prior history of malignancy (not primary lung cancer). The mean nodule size on CT scanning was 1.7 cm (S.D. ± 0.7 cm). Twelve nodules (22%) were ≤1.0 cm (median 0.9 cm, range 0.4–1 cm), 27

Table 1. Detection at LSO-PS vs. BGO-PET as a function of lesion size

Size category (cm) N

Semiquantitative analysis contrast vs. background (mean ± S.D.)

Visual analysis percent and number of patients scored positive

BGO-PET LSO-PS BGO LSO-PSsensitivea

LSO-PSconservativeb

≤1.0 12 2.8 ± 0.9 1.8 ± 0.8 33% (4) 33% (4) 33% (4)

1.1–1.5 14 4.0 ± 1.6 2.3 ± 1.1 86% (12) 79% (11) 71% (10)

1.6–2.0 13 5.1 ± 4.0 3.1 ± 2.2 77% (10) 77% (10) 77% (10)

2.1–2.5 7 4.0 ± 2.7 3.1 ± 2.1 86% (6) 86% (6) 86% (6)

2.6–3.0 8 4.0 ± 1.7 2.6 ± 0.7 88% (7) 100% (8) 75% (6)

a Sensitive = faint increased uptake or greater deemed positive.b Conservative = moderate or intense increased uptake deemed positive.


Chapter 6

76

(50%), in the 1.1–2.0 cm range (median 1.5 cm, range 1.1–2 cm), and 15 (28%) in the 2.1–3.0 cm range (median 2.6 cm, range 2.2–3.0 cm) (Table 1).

Qualitative analysisBased on the relative lesion intensity criteria described for lesion positivity for malignancy with BGO-PET, 39/54 (72%, 95%CI 59–82%) pulmonary nodules were considered suspect for malignancy. Using sensitive assessment criteria with LSO-PS, (i.e. faint lesion intensity or greater considered positive for malignancy), 38 (97%, 95%CI 87–100%) were also positive at LSO-PS. However, such an assessment strategy resulted in one false positive with LSO-PS. At BGO-PET, this nodule had faint increased uptake and was therefore classifi ed ‘negative’. Using a conservative assessment strategy with LSO-PS (i.e. moderate or intense increased lesion intensity defi ned as positive), 36/39 (92%, 95%CI 80–97%) were positive with LSO-PS. The one lesion missed by LSO-PS in either assessment strategy involved a 1.5 cm diameter lesion that demonstrated moderate increased uptake on BGO-PET. The smallest positive lesion visualized by LSO-PS was 0.7 cm (Fig. 1).

Fig. 1. BGO-PET (right) as well as LSO-PS (left) demonstrate uptake in a 0.7 cm lesion in the lower lobe of the left lung.

Both AC and nonAC LSO-PS images were reviewed by two observers in the cases in which there was discrepancy between LSO-PS and BGO-PET. No diff erence in lesion detectability was demonstrated in the AC images as opposed to the nonAC images. Interobserver agreement with respect to the relative lesion intensity reading was similar between BGO-PET and LSO-PS (ICC: 0.93 and 0.89 for BGO-PET and LSO-PS, respectively).

Semiquantitative analysisLesion contrast was determined using the algorithm described previously. The mean lesion contrast with BGO-PET was 4.2 (S.D. 2.6). As compared to BGO-PET, a signifi cantly lower contrast was demonstrated with LSO-PS for both nonAC (mean 2.7, S.D. 1.6, p < 0.05) and AC images (mean 3.1, S.D. 1.6, p < 0.05). The slight diff erence of contrast between AC and nonAC images was statistically signifi cant (p = 0.04).



77

Discussion

FDG-PET has an established role in the diagnostic work-up of solitary pulmonary nodules. The demand for PET scanning may further increase through new potential indications such as screening for lung cancer (19). Therefore, the performance of cheaper and more widely available alternatives for PET should be explored. The main concern when considering the use of gamma cameras for this purpose is their expected lower sensitivity (11). We found that the prototype LSO-PS scanner had a detection rate which was highly comparable to that of a full ring BGO-PET scanner.

Meta-analysis (5) indicates that the median reported sensitivity of FDG PET in pulmonary nodules (1–3 or 4 cm) is 97% (n = 450, median specifi city 83%). Earlier studies investigating gamma cameras with high energy collimators for pulmonary nodules (lesion size range 1.2–3.8 cm) (20), and masses (lesion size range 0.5–11.0 cm) (21), demonstrated sensitivities of 80 and 77%, respectively. Studies using DHC technology in indeterminate pulmonary lesions (9;10) have shown better sensitivity versus histology (97–100%), but these data were derived from a broad spectrum of lesion sizes (1.0–9.5 cm), the apparent majority (75% in Weber et al. (10)) being larger than 2.0 cm. Therefore, the data on accuracy of DHC in smaller pulmonary nodules is limited. The study design of Weber et al. (10) included the use of controls in order to compensate for potential selection bias due to the high prevalence of malignancy [90% prevalence of malignancy among the cases versus about 70% in the accumulated literature (5)]. Weber et al. (10) reported a DHC sensitivity versus histology of 88% (95%CI 69–96%, n = 24) in solitary pulmonary nodules ≤2 cm. In our prospective study, we attempted to limit selection bias by entering consecutive patients, and in fact, the distribution of lesion size and the PET positivity rate were similar to that observed in our clinical practice (22). Even though the role of FDG PET for characterisation of subcentimeter pulmonary lesions is not established (23), we included these cases in the present study since our previous experience with full ring BGO-PET was positive (16).

We intentionally chose for a study design that involved a head-to-head comparison with full ring PET using the test result of the latter as the reference test. Traditionally, the performance of DHC is compared to a clinical gold standard. However, if one foresees a role for DHC as an instrument of triage for full ring PET, the alternative study design is to evaluate the relative performance using the results of full ring PET as the reference test. In retrospect, this was indeed justifi ed since lower sensitivity of DHC is the key discriminant, and false positivity did not prevail. It should be noted that, even though lesion detectability was almost similar with LSO-PS and BGO-PET, contrast with LSO-PS was lower than with full ring PET. If confi rmed in a larger group of patients, these results suggest that negative test result at LSO-PS obviates the use of full ring PET. Such validation study should also account for limitation of the present study that full ring PET was not performed with attenuation correction, which currently is the state-of-the-art method.

The LSO-PS camera evaluated in the present study incorporates the scintillator LSO, physical properties of which make it especially suited to imaging the 511 keV emission of FDG. LSO has a similar density as BGO but signifi cantly higher than NaI which is an important factor for detecting


Chapter 6

78

the higher energy emission of FDG. In addition, LSO has a shorter decay constant than both BGO and NaI which improves its count rate capability. In vitro analysis of noise equivalent count rates has demonstrated a performance similar to a partial ring BGO-PET camera (14). This higher count rate as compared to conventional DHC cameras incorporating only NaI enables a signifi cantly shorter acquisition time (13;24;25). In theory, performing LSO-PS imaging after BGO-PET might favour the former since FDG uptake in tumors increases over time (26). Finally, LSO-PS images had to be acquired in 3D mode as compared to 2D mode with BGO-PET with a known resultant increased amount of scatter beyond geometry limitations that can degrade image quality.

Conclusion

LSO-PS demonstrates a high concordance (92–97%, depending on whether a conservative or sensitive assessment strategy is used) with non-attenuation corrected BGO-PET for the detection of pulmonary nodules with enhanced FDG uptake. If these results are confi rmed in a larger study, LSO-PS is suited as an instrument of triage for PET in patients with clinically indeterminate solitary pulmonary nodules.

References (1) Gupta NC, Maloof J, Gunel E. Probability of malignancy in solitary pulmonary nodules using fl uorine-18-FDG

and PET. J Nucl Med 1996; 37(6):943-948.

(2) Gurney JW. Determining the likelihood of malignancy in solitary pulmonary nodules with Bayesian analysis.

Part I. Theory. Radiology 1993; 186(2):405-413.

(3) Khouri NF, Meziane MA, Zerhouni EA, Fishman EK, Siegelman SS. The solitary pulmonary nodule.

Assessment, diagnosis, and management. Chest 1987; 91(1):128-133.

(4) Swensen SJ, Silverstein MD, Ilstrup DM, Schleck CD, Edell ES. The probability of malignancy in solitary

pulmonary nodules. Application to small radiologically indeterminate nodules. Arch Intern Med 1997;

157(8):849-855.



(6) Gould MK, Sanders GD, Barnett PG, Rydzak CE, Maclean CC, McClellan MB et al. Cost-eff ectiveness of alternative

management strategies for patients with solitary pulmonary nodules. Ann Intern Med 2003; 138(9):724-735.

(7) Ost D, Fein AM, Feinsilver SH. Clinical practice. The solitary pulmonary nodule. N Engl J Med 2003;

348(25):2535-2542.

(8) Tan BB, Flaherty KR, Kazerooni EA, Iannettoni MD. The solitary pulmonary nodule. Chest 2003; 123(1

Suppl):89S-96S.

(9) Kim S, Park CH, Han M, Hwang S, Lee C, Pai M. The clinical usefulness of F-18 FDG coincidence PET without

attenuation correction and without whole-body scanning mode in pulmonary lesions comparison with CT,

MRI, and clinical fi ndings. Clin Nucl Med 1999; 24(12):945-949.



79

(10) Weber W, Young C, Abdel-Dayem HM, Sfakianakis G, Weir GJ, Swaney CM et al. Assessment of pulmonary

lesions with 18F-fl uorodeoxyglucose positron imaging using coincidence mode gamma cameras. J Nucl

Med 1999; 40(4):574-578.





(13) Watson CC, Eriksson L, Casey ME, Jones WF, Moyers JC, van Lingen A. Design and performance of

collimated coincidence point sources for simultaneous transmission measurements in 3D PET. IEEE Nucl Sci

2001;(48):673-679.

(14) van Lingen A, Bendriem B, Luk P, Jones WF, Young J, Nutt R. Initial characterization of a new hybrid

tomograph: LSO/NaI PET/SPECT. J Nucl Med 41[5], 20P. 2001.



2003; 24(3):281-289.

(16) Herder GJ, Golding RP, Hoekstra OS, Comans EF, Teule GJ, Postmus PE et al. The performance of( 18)F-

fl uorodeoxyglucose positron emission tomography in small solitary pulmonary nodules. Eur J Nucl Med Mol

Imaging 2004; 31(9):1231-1236.

(17) Avril N, Rose CA, Schelling M, Dose J, Kuhn W, Bense S et al. Breast imaging with positron emission

tomography and fl uorine-18 fl uorodeoxyglucose: use and limitations. J Clin Oncol 2000; 18(20):3495-3502.




(19) Pastorino U, Bellomi M, Landoni C, De Fiori E, Arnaldi P, Picchio M et al. Early lung-cancer detection with

spiral CT and positron emission tomography in heavy smokers: 2-year results. Lancet 2003; 362(9384):593-

597.

(20) Worsley DF, Celler A, Adam MJ, Kwong JS, Muller NL, Coupland DB et al. Pulmonary nodules: diff erential

diagnosis using 18F-fl uorodeoxyglucose single-photon emission computed tomography. AJR Am J

Roentgenol 1997; 168(3):771-774.



(22) Herder GJ, van Tinteren H, Comans EF, Hoekstra OS, Teule GJ, Postmus PE et al. Prospective use of serial

questionnaires to evaluate the therapeutic effi cacy of 18F-fl uorodeoxyglucose (FDG) positron emission

tomography (PET) in suspected lung cancer. Thorax 2003; 58(1):47-51.

(23) Nomori H, Watanabe K, Ohtsuka T, Naruke T, Suemasu K, Uno K. Evaluation of F-18 fl uorodeoxyglucose (FDG)

PET scanning for pulmonary nodules less than 3 cm in diameter, with special reference to the CT images.

Lung Cancer 2004; 45(1):19-27.



2004; 31(4):596-598.


Chapter 6

80


(26) Kubota K, Itoh M, Ozaki K, Ono S, Tashiro M, Yamaguchi K et al. Advantage of delayed whole-body FDG-PET

imaging for tumour detection. Eur J Nucl Med 2001; 28(6):696-703.


C h a p t e r

Evaluation of new imaging tests suited for triage in oncology: the case of a prototype dual crystal (LSO/NaI) dual head coincidence camera and mediastinal staging of non-small cell lung cancer (NSCLC)

7Urvi JoshiArthur van LingenEmile FI ComansRik PijpersPieter GHM RaijmakersGerrit JJ TeulePieter E PostmusOtto S Hoekstra

Submitted


Abstract

Objective: Technological advances of dual head coincidence cameras (DHC) would benefi t from faster evaluation than is feasible with standard accuracy studies. If DHC should serve as triage for FDG PET, head-to-head comparison with PET as gold standard is an effi cient design. We compared the performance of a prototype LSO/NaI DHC camera (LSO-PS) with full ring BGO-PET in staging the mediastinum of non-small cell lung carcinoma (NSCLC).

Methods: Prospectively, 55 consecutive patients with (suspected) operable NSCLC and repre-sentative nodal sizes at CT underwent BGO-PET and LSO-PS, using a whole body survey (5 min/bed) for BGO-PET and the thorax and abdomen (10 min/bed) for LSO-PS. BGO-PET was acquired in 2D without attenuation correction and LSO-PS in 3D with simultaneous emission/transmission acquisition. Four randomly assigned observers independently interpreted one modality per pa-tient, judging nodal stations, N-stage and management.

Results: In 44 evaluable mediastinal datasets, a lower N-stage was found with LSO-PS than with BGO-PET in 6/44 (14%), leading to diff erent management advice in 9%. Contrast was subjec-tively higher with BGO-PET (p=0.001). Of PET positive nodes, 64% were identifi ed at LSO-PS; 53% of PET positive stations measuring ≤ 10 mm at CT were also positive at LSO-PS.

Conclusion: LSO-PS and BGO-PET correlate well with respect to the N-stage in NSCLC. At about half the acquisition time of a typical NaI DHC scanner, LSO-PS can visualize small tumour deposits. It remains to be shown whether confi rmation of mediastinal involvement at LSO-PS suff ers from its relative lack of sensitivity on a nodal station level.


Evaluation of new imaging tests suited for triage in oncology

83

Introduction

The major shortcoming of DHC systems is diminished sensitivity vs. full ring PET (1). Therefore, a potentially useful application of DHC systems is to serve as a triage for full ring PET. Evaluation of technological advances in DHC scanners is typically done with accuracy studies using clinical gold standards. However, these are time-consuming, diffi cult to conduct and therefore less suited to rapidly evolving technologies. A more effi cient study design to evaluate advances in DHC systems would be a head-to-head comparison with a standard full ring PET system using the latter as the gold standard, assuming that observers are strictly blinded to the alternative modality.

In non-small cell lung cancer (NSCLC), FDG-PET improves staging and management as compared to computed tomography (CT) (2-5). However, the evidence is largely based on full ring bismuth germanate (BGO) PET systems; sodium iodide (NaI) DHC scanners have been compared to pathological staging (6-12) with variable success.

In the meantime, new crystals such as lutetium oxyorthosilicate (LSO) have been introduced (13). In 1997, a new type of DHC camera was developed consisting of a dual layer of phoswhich crystals: LSO and NaI, designed for both PET and SPECT imaging (LSO-PS). Initial in vitro evaluation has demonstrated physical performance parameters (noise equivalent count rate, spatial resolution) comparable to a partial ring BGO-PET system (14;15). The major diff erence between LSO-PS and other coincidence cameras on the market relates to its higher count rate. This is due in part to the higher density of LSO which is approximately twice that of NaI and similar to that of BGO. The higher count rate is also related to the shorter decay constant of LSO vs. NaI and BGO. Typical peak noise equivalent count rates (NEC) for LSO-PS are approximately 50kps vs. ca. 10 kps for NaI (16). This facilitates a better image quality with LSO-PS for the same amount of time.

This higher count rate as compared to conventional DHC cameras incorporating only NaI enables a signifi cantly shorter acquisition time (10 min of simultaneous emission and transmission scan-ning per bed position, vs. emission scan times per bed from 21-47 min (6-12), respectively). In vivo imaging in clinical oncology patients (17) justifi ed further evaluation of the performance of this prototype camera.

The goal of the present study was to perform a direct, prospective comparison with a dedicated full ring BGO-PET system focused on mediastinal lymph node staging in patients with (suspected) NSCLC.


Chapter 7

84

Material and methods

Study DesignA prospective study was performed whereby patients referred to our PET centre for pre-opera-tive/pre-mediastinoscopy staging of proven/suspected NSCLC with FDG-PET were eligible. All pa-tients were considered to be potentially operable prior to PET scanning (i.e. no evidence of distant metastases). To obtain an adequate case mix, i.e. to avoid overrepresentation of patients with en-larged nodes, we set out to include consecutive patients according to the size of their mediastinal and hilar lymph nodes (LN) at CT scanning: 1) ≤10 mm short axis diameter (n=20), or at least a sin-gle LN of 2) 10.1-15 mm (n=10), 3) > 15 mm (n= 10). The size of LN’s at CT were measured prior to the PET scan (if available). Otherwise, the external CT reading was used to decide upon eligibility of individual patients. In retrospect, some of these external readings proved incorrect, explaining why 25 instead of 20 patients were included in the subset of patients with LN <10 mm.

All patients underwent scanning with both full ring BGO-PET and LSO-PS on the same day at fi xed time points (60 and 120 min respectively) after a single dose of FDG (average dose 370 MBq/70kg). Patients had fasted a minimum of six hours prior to FDG administration and their serum glucose was < 10 mmol/l. The study protocol was approved by our institutional ethics committee and all patients gave written informed consent.

Image acquisition and reconstructionBGO-PET scans were performed on an ECAT HR+ scanner (CTI PET systems), involving a whole body acquisition in 2D mode with an emission scan of 5 minutes/bed position. Data was recon-structed using OSEM (Ordered subsets expectation maximum with 4 iterations, 16 subsets. The use of attenuation correction (AC) is controversial (18) and our clinical routine whole body proto-col with the BGO-PET system does not include this. As such, we decided to focus the methodolo-gically most stringent analysis (with respect to having two blinded observers for each modality) on a comparison of non-attenuation corrected images. However, we performed an additional, separate evaluation of AC LSO PET-PS images by two independent observers (EC, OH) in cases in which there was a discrepancy in lymph node station obtained by BGO-PET as compared to LSO-PS. Finally, we performed a separate analysis to determine the level of observer bias which would have been present without appropriate blinding of observers. To this end, LSO-PS scans were reinterpreted by a single observer (EC) with knowledge of the BGO-PET images.

LSO-PS scans were performed on a prototype dual crystal, DHC system (CTI PET systems) which has detector heads consisting of a dual layer of phoswich crystals, LSO and NaI, each 10 mm thick. The useable camera fi eld of view is 35 x 50 cm. The system has a coincidence timing window of 6 nsec. Transmission scanning is performed with 7 collimated germanium-68 point sources, fi tted to the yokes of the detector heads (further technical details of the LSO-PS camera have been published elsewhere (14;15;19)). The LSO-PS acquisition consisted of simultaneous 3D emission and transmission scans extending inferiorly from the neck (2 bed positions of 10 minutes each). Data were reconstructed with OSEM (2 iterations, 8 subsets), and 5 mm FWHM Gaussian smoothing fi lter.



85

Data AnalysisAll images (5mm slices in axial, coronal and sagittal planes, were viewed with the same software on a single viewing station running Matlab 5.3, by four experienced nuclear medicine physicians (RP, EC, OH, PR) blinded to the other scan modality of the same patient. Each scan was independently read by two randomly assigned observers. Interobserver discrepancy was resolved through a third observer. The scans were read without knowledge of CT scan fi ndings nor of specifi c clinical information other than the suspicion of NSCLC.

Each relevant focal abnormality (defi ned as enhanced uptake versus background) was interpreted in terms of intensity (4 point scale: decreased, mildly increased, moderately increased and markedly increased) and likelihood of malignancy (5 point scale: defi nitely or probably benign, equivocal, probably or defi nitely malignant). In the fi nal data analysis, the fi nal results (of both observers +/- a third observer if necessary) were scored as positive (probably or defi nitely malignant), equivocal (equivocal) or negative (probably or defi nitely benign) for malignancy. However, all fi nal results were either positive or negative and, as such, the equivocal category was omitted from the table of results (Table 1). Positive scores with BGO-PET were considered as the gold standard.

If the primary tumour accumulated FDG, the observers assigned an N-stage (N0, N1, N2, N3), and recorded the localization of abnormal foci (ipsi- or contralateral cervical, upper pretracheal, upper paratracheal, pre/retrotracheal, tracheobronchial, aortopulmonary, paraaortic, carinal, para-oesophageal or hilar). Finally, each observer formulated an advice for further patient management (biopsy, thoracotomy or expectative policy in case of an FDG negative primary lung lesion). Observers were instructed to recommend biopsy (usually mediastinoscopy) in case of suspected lymph node involvement. However, mediastinoscopy was also to be advised in case of tumours adjacent to the mediastinum or hilus, and in case of presumed ipsilateral hilar adenopathy. In the remaining cases, the next step should be thoracotomy. The presence of distant metastases was not evaluated since BGO-PET and LSO-PS scan trajects were not identical.

Statistical analysisWilcoxon’s signed rank test was used to analyse potential diff erences of lesion intensity using the summed scores of the 2 observers assigned to either modality. Signifi cance was set at 5%.

Results

Patients were referred for PET imaging according to prevailing guidelines for the use of FDG-PET in staging lung cancer in our area, by community and university hospitals. Between March and October, 2002, 55 consecutive eligible patients were included until our requirements for distribution of LN sizes at CT were met. The mean age was 65 yrs (SD 12), and 65% were male. In three patients, technical problems with the LSO-PS scanner precluded image analysis, and eight had an FDG negative primary lesion at BGO-PET. Therefore, for comparison of the mediastinal staging with the two scanners, we obtained 132 datasets in 44 patients (nonattenuation corrected/attenuation corrected LSO-PS, and nonattenuation corrected BGO-PET in each patient). Twenty-


Chapter 7

86

fi ve of these patients had normal sized LN at CT (< 1cm: cN0), 10 had LNs of 10-15mm and 9 of >15mm, respectively.

The N-stage at BGO-PET of the remaining 44 patients was: 23 N0, 4 N1, 11 N2, 6 N3. With LSO-PS, a lower N stage was observed in 6/44 as compared to BGO-PET. Since LSO-PS showed no overstaging vs. BGO-PET, a 86% concordance was measured [95% Confi dence Interval (CI): 73-94%]. Table 1 compares the N-stage obtained with CT, BGO-PET and LSO-PS as well BGO-PET and LSO-PS positivity as a function of nodal size.

Table 1. Lymph node (LN) location, size, BGO-PET and LSO-PS scores in patients with FDG avid primary tumours,

excluding patients who had normal nodes at CT as well as negative PET/LSO-PS mediastina.

P a t i e n t no.

T site* LN station** CT short axis LN diameter (mm)

BGO-PET****

LSO-PS****

N stage

CT BGO PET

LSO PS

1 R Cervical (R) Not scanned + - N0 N3 N3Paratracheal (R) Not enlarged + -Paratracheal (L) Not enlarged + +Mediast. Ant. Not enlarged + -Pretracheal Not enlarged + +Tracheobronchial (R) Not enlarged + +Tracheobronchial (L) Not enlarged + -Hilar (R) Not enlarged + +Hilar (L) Not enlarged + +

2 L Cervical (L) Not scanned + - N2 N3 N2Retrotracheal (L) 14 + +Para-aortic 25 + -APV *** 15 + -Carinal 18 + +Hilar (R) Not enlarged + -Para-aortic 22 - -

3 R Cervical Not scanned + - N3 N3 N2Paratracheal (R) 15 + +Tracheobronchial (R) 22 + +Carinal 22 + +Hilar (R) Not enlarged + +Hilar (L) Not enlarged + -Para-aortic 15 - -

4 R Hilar (R) 32 + + N2 N1 N1Tracheobronchial (R) 11 - -

5 R Tracheobronchial (R) 15 + + N2 N2 N28 L Tracheobronchial (L) 16 + + N2 N2 N29 L APV 11 - - N2 N0 N0

Carinal 11 - -

11 Carinal 15 - - N2 N0 N012 L Hilar (L) Not enlarged + + N0 N1 N116 R Hilar (R) Not enlarged + - N0 N1 N017 L APV 11 - - N2 N0 N0



87

19 R Cervical (R) Not scanned + + N2 N3 N3Cervical (L) Not scanned + +

Mediast. Sup. Not enlarged + +

Parasternal/retrosternal Not enlarged + -

Tracheobronchial (R) 22 + +

APV Not enlarged + -

Carinal 20 + +

Para-esophageal Not enlarged + -

Hilar (L) Not enlarged + +

21 L Hilar (L) Not enlarged + - N0 N1 N022 R Cervical (R) Not scanned + + N2 N3 N3

Paratracheal (R) Not enlarged + +

Tracheobronchial (R) 12 - -

Hilar (R) 12 - -

Mediast. Sup. 12 - -

Mediast. Ant. 22 - -

23 L Para-aortic Not enlarged + + N0 N2 N228 R Paratracheal (R) Not enlarged + + N0 N2 N229 Carinal Not enlarged + - N0 N2 N031 R Tracheobronchial (R) 12 + + N2 N2 N2

Carinal 11 - -

32 Carinal 25 + + N2 N2 N236 L APV Not enlarged + - N2 N2 N2

Carinal 11 + +

37 L Cervical (L) Not scanned + + N0 N3 N3Paratracheal (L) Not enlarged + +

Paratracheal (R) Not enlarged + +

Mediast. Ant. Not enlarged + -

Carinal Not enlarged + +

Hilar (R) Not enlarged + +

39 Carinal (7) 12 - - N2 N0 N044 R Tracheobronchial (R) 15 - - N2 N0 N050 R Tracheobronchial (R) 15 + + N2 N2 N2

Carinal 12 + -

51 L Carinal Not enlarged + - N2 N2 N0APV 11 - -

52 R Tracheobronchial (R) 33 + + N2 N2 N2Hilar (R) 26 + +

Carinal 11 - -

55 Carinal 20 - - N2 N0 N0

*location of primary tumour if relevant for N-classifi cation (R = Right, L=Left)

** Naruke station; ***APV aortopulmonary window; ****Positive (+)/Negative (-)

P a t i e n t no.

T site* LN station** CT short axis LN diameter (mm)

BGO-PET****

LSO-PS****

N stage

CT BGO PET

LSO PS


Chapter 7

88

Analysis by nodal station revealed that 64% (36/56, 95%CI: 51-76%) of the nodal stations visualized with BGO-PET were also detected with LSO-PS. Of the 20 positive LN stations at BGO-PET not vi-sualized by LSO-PS, 14 were ≤ 10 mm at CT, two were in the 10-15 mm range, one was larger than 15 mm and the size of the remaining three was not evaluable (in the cervical region, i.e. outside the chest CT scan trajectory); 16/30 of PET positive stations sized ≤ 10 mm were also positive at LSO-PS (53%, 95%CI: 36-70%, Fig 1.) vs. 6/8 (75%, 85%CI: 41-93%) of PET positive stations sized 10-15 mm.

Figure 1. Right: normal sized para-aortic lymph node positive on BGO-PET (bottom row) and LSO-PS (top row).

Primary tumour (left) measured 7x7 mm, and was located peripherally in the left lung.

There were no false positives with LSO-PS. Based on the approach described in the methods section, discrepancies in patient management advice were found in 4/44 patients (9%, 95%CI: 4-21%). These patients received an N0 stage with LSO-PS (i.e. advice to proceed with thoracotomy) compared to an N1 (n=2) or N2 (n=2) stage with BGO-PET (i.e. advice to proceed with mediastinoscopy). Two patients received an N2 stage with LSO-PS while BGO-PET suggested N3 stage, but either classifi cation led to mediastinoscopy. Figures 2 and 3 demonstrate two examples in which a lymph node was missed by LSO-PS, resulting in a diff erent N staging as compared to BGO-PET. In Figure 2, there is no change in management advice while in Figure 3 thoracotomy is suggested based on LSO-PS as opposed to mediastinoscopy by BGO-PET.



89

Figure 2. Discrepant reading of nodal stations but similar management advice. Left hilar lymph node not seen with

LSO-PS (top row) but visible with BGO-PET (bottom row, arrow). However, with either modality, other mediastinal

lymph nodes were positive, resulting in congruent N-stage and a similar advice for mediastinoscopy.

Figure 3. Diff erent management advice. Pathological aortopulmonary lymph node (not enlarged on CT) with BGO

PET (right, arrow), not seen with LSO-PS (left). Both modalities show uptake in the 15 x 20 mm primary lesion in the

left lung. Staged as N0 with LSO-PS with advice for thoracotomy; staged as N2 with BGO-PET with advice for biopsy.

The relative lesion intensity was signifi cantly lower with LSO-PS (Wilcoxon’s signed rank test, p =0.001). Interobserver agreement to classify ‘intensity’ using the Likert scale was similar between the modalities: complete agreement in 72% and 60% for PET and LSO-PS, respectively, with only in the LSO-PS group disagreement of 2 scaling points (in 6%). With respect to the interpretation, observers agreed in 84% and 90% of nodal stations for BGO-PET and LSO-PS, respectively.

Several months after the initial readings, and independently, two observers (EC, OH) evaluated the LSO-PS images with and without attenuation correction (AC and nonAC, respectively) vs. BGO-PET. AC provided no improvement in detection of nodal stations that had been missed with nonAC. Image quality was considered less optimal with AC LSO-PS. Finally, interpretation of LSO-PS images with knowledge of BGO-PET fi ndings resulted in a 18% increase (46/56 vs. 36/56) in the number of nodal stations visualized with LSO-PS.


Chapter 7

90

Discussion

Finding the appropriate place for new technology is a cumbersome process. Even after the diagnostic performance of full ring PET in mediastinal staging of NSCLC has been studied in more than 1000 patients, its role is still not well-defi ned (20). In part this relates to the formidable demands that are put upon such accuracy studies (21). In the end, very few studies meet all methodological requirements. Not surprisingly, the same is true for DHC studies. Diff erences in patient spectra, reference tests, blinding procedures and other methodological issues probably contribute to the observed heterogeneity of the results. However, if one foresees a role for DHC as an instrument of triage for full ring PET (i.e. each patient should undergo FDG imaging prior to surgery), the alternative design is to study the relative performance using the results of full ring PET as the gold standard.

To our knowledge, four studies used head-to-head comparisons of PET and NaI-DHC to study the mediastinum in NSCLC, with reported relative sensitivities with DHC vs. PET readings between 77% and 100%. Such results are a function of the level of analysis: we found 64% concordance at the nodal station-, 86% at the N-stage-, and 91% at the management level. In the earlier studies, these analysis levels were heterogeneous: Zimny et al (10) reported N-stages, while Tatsumi et al (9) evaluated the entire mediastinum or hilus as being positive or negative. Weber et al (11) appear to have used a nodal station analysis and Delahaye et al (12) a lesional and an N-stage evaluation. The accumulated data suggest that positive concordance was dependent of the level of analysis, with the lowest concordance of positive scores between DHC and PET in the nodal station analysis (10/13, 77%, 95%CI: 50-92%) (11). In the subset of lymph nodes smaller than 1 cm, we found a 53% positive concordance of LSO-PS and BGO-PET. This compares favourably with Delahaye et al (12) who reported 17% positive concordance with BGO-PET for such lesions.

Inappropriate blinding of observers for the alternative imaging modality may induce considerable bias in head-to-head comparisons, in favour of the DHC method. Our 18% higher nodal detection rate with combined LSO-PS and BGO-PET reading corroborates and extends the fi ndings of Zimny et al (22) who reported an increased detection rate up to 25% with small lesions. Among the abovementioned DHC-PET studies, strictness of blinding was variable, using a single observer for the entire study (23) or multiple observers. Tatsumi et al (9) and Weber et al (11) used two independent observers but it was unclear whether they were blinded to the other modality. Finally, Delahaye et al (12) used two diff erent observers for each modality, but did not specify whether the observers read the images independently. Our design to randomly assign two observers to a single scan modality of each patient ensures an unbiased estimation of DHC-PET performance together with an assessment of interobserver variation.

Finally, measured DHC-PET concordance may be aff ected by the patient spectrum. In the present setting this refers to the distribution of nodal sizes at CT. This is also relevant for external validity. We could not always extract these data from the DHC-PET studies, but in two of them cN0 prevalences were variable (37% and 62% (10;12)). We had aimed to obtain a similar cN0 prevalence (about 50%) as in the study of Pieterman et al (24).



91

In summary, even though diff erences in methodology impair a direct comparison with other existing DHC scanners, it appears that LSO-PS was able to detect metastases in small lymph nodes, which so far has been the main limitation of DHC. It remains to be shown whether the observed diff erences between LSO-PS and BGO-PET are clinically relevant. This may indeed be the case if the yield of mediastinoscopy or other invasive procedures increases by exact localization of involved LN at PET (25).

We did not fi nd improved sensitivity with attenuation correction in DHC imaging; others have reported likewise (7) or found a 12% better relative performance with AC (10). It should be noted, however that our results obtained in the chest without AC may not translate to other more diffi cult regions such as the abdomen. It could be argued that with the present scanner, emission source scatter into the transmission acquisition may have degraded transmission scan image quality, but such cross-talk is reported to be minimal (19). While the capability to perform simultaneous transmission and emission acquisition is attractive, the simultaneous AC mechanism on the LSO-PS prototype may require further refi nement. BGO-PET studies were obtained in 2D mode whereas LSO-PS studies were obtained in 3D mode, resulting in an increased scatter fraction beyond geometry limitations. Finally, we applied LSO-PS and BGO-PET at time points suggested to be optimal for either scanner by the in vitro data (i.e. the NEC-curves). This refl ects the expected clinical setting but obviously precludes direct comparison of simultaneously acquired lesion to background ratios (26). Since LSO-PS imaging was performed after BGO-PET, we cannot exclude that increasing signal-background contrast over time may have biased the comparison in favour of LSO-PS. This is a common limitation of such head-to-head comparative studies, and ours is no exception. However, our standard oncological FDG PET protocol entails whole body imaging commencing 60 min post injection of 370 MBq of FDG. Such a dose does not enable earlier imaging with LSO-PS than 120 min, due to count rate limitations. Alternatively, we should have asked the patients to be scanned with LSO-PS on another occasion after renewed injection of a lower dose of FDG. However, since we reasoned that our LSO-PS acquisition protocol might also classify for clinical implementation, we decided to avoid the ethical dilemma of subjecting patients to the additional radiation and the inconvenience associated with an extra FDG injection. Therefore, it remains to be shown whether modifi ed LSO-PS acquisition protocols also compatible with the NEC curves (lower dose, scanning e.g. 60 minutes after FDG administration, longer acquisition time) would provide the same results.

It could be argued that mediastinal staging is only part of the yield of PET in NSCLC. Even though the LSO-PS scanner allows true whole body imaging, we deliberately chose to focus on the mediastinum. Even though this implies that the ability of the scanner to pick up distant metastases still needs to be established, the choice for the mediastinum was prompted by our experience that such distant metastases typically are not very small.

We conclude that evaluation of scanners considered for triage of oncology patients benefi ts from using the reference scan technology as the gold standard and rigorous blinding of observers to the alternative modality. For LSO-PS, the overall agreement with BGO-PET on fi nal N-stage was high, but the superiority of PET to visualize the smaller deposits was evident. It remains to be


Chapter 7

92

shown to which extent this diff erence negatively aff ects the yield of confi rmatory biopsies when using LSO-PS.

AcknowledgementsThe authors thank the referring pulmonologists for their cooperation: CR Apap, BTJ van den Berg, J Berkovits, B Biesma, WG Boersma, CD Colder, PC Dekker, CL van Felius, T Haitjema, AHM van der Heijden, PM Hooghiemstra, FJM. Jacobs, PJH Janssen, AM Janssens, J Kodde, FH Krouwels, CF Melissant, JL van Opstal, PE Postmus, J Prins, CH Rikers, PI van Spiegel, GHA Staaks, RALM Stallaert, WFM Strankinga, JP Teengs, JAALM Thönissen, FMJ Tobin, G Visschers, A Vonk Noordegraaf, A Welling, as well as A Kalwij and C Karga for their secretarial assistance.

References



(2) Dwamena BA, Sonnad SS, Angobaldo JO, Wahl RL. Metastases from non-small cell lung cancer: mediastinal

staging in the 1990s--meta-analytic comparison of PET and CT. Radiology 1999; 213(2):530-536.

(3) Fischer BM, Mortensen J, Hojgaard L. Positron emission tomography in the diagnosis and staging of lung

cancer: a systematic, quantitative review. Lancet Oncol 2001; 2(11):659-666.

(4) Toloza EM, Harpole L, McCrory DC. Noninvasive staging of non-small cell lung cancer: a review of the

current evidence. Chest 2003; 123(1 Suppl):137S-146S.

(5) van Tinteren H, Hoekstra OS, Smit EF, van den Bergh JH, Schreurs AJ, Stallaert RA et al. Eff ectiveness of

positron emission tomography in the preoperative assessment of patients with suspected non-small-cell

lung cancer: the PLUS multicentre randomised trial. Lancet 2002; 359(9315):1388-1393.

(6) Roman MR, Rossleigh MA, Angelides S, Walker BM, Dixon J. Staging and managing lung tumors using F-18

FDG coincidence detection. Clin Nucl Med 2001; 26(5):383-388.

(7) Stevens H, Bakker PF, Schlosser NJ, Van Rijk PP, de Klerk JM. Use of a dual-head coincidence camera and

18F-FDG for detection and nodal staging of non-small cell lung cancer: accuracy as determined by 2

independent observers. J Nucl Med 2003; 44(3):336-340.

(8) Stokkel MP, Bakker PF, Heine R, Schlosser NJ, Lammers JW, Van Rijk PP. Staging of lymph nodes with FDG

dual-headed PET in patients with non- small-cell lung cancer. Nucl Med Commun 1999; 20(11):1001-1007.

(9) Tatsumi M, Yutani K, Watanabe Y, Miyoshi S, Tomiyama N, Johkoh T et al. Feasibility of fl uorodeoxyglucose

dual-head gamma camera coincidence imaging in the evaluation of lung cancer: comparison with FDG

PET. J Nucl Med 1999; 40(4):566-573.

(10) Zimny M, Hochstenbag M, Lamers R, Reinartz P, Cremerius U, Ten Velde G et al. Mediastinal staging of lung

cancer with 2-[fl uorine-18]-fl uoro-2-deoxy- D-glucose positron emission tomography and a dual-head

coincidence gamma camera. Eur Radiol 2003; 13(4):740-747.






93

(12) Delahaye N, Crestani B, Rakotonirina H, Lebtahi R, Sarda L, Girard P et al. Comparative impact of standard

approach, FDG PET and FDG dual-head coincidence gamma camera imaging in preoperative staging of

patients with non-small-cell lung cancer. Nucl Med Commun 2003; 24(12):1215-1224.


Trans Nucl Sci 39, 502-505. 1992.



(15) van Lingen A, Bendriem B, Luk P, Jones WF, Young J, Nutt R. Initial characterization of a new hybrid

tomograph: LSO/NaI PET/SPECT. J Nucl Med 41[5], 20P. 2001.




2004; 31(4):596-598.



2003; 24(3):281-289.

(19) Watson CC, Eriksson L, Casey ME, Jones WF, Moyers JC, van Lingen A. Design and performance of

collimated coincidence point sources for simultaneous transmission measurements in 3D PET. IEEE Nucl Sci

2001;(48):673-679.

(20) Silvestri GA, Tanoue LT, Margolis ML, Barker J, Detterbeck F. The noninvasive staging of non-small cell lung

cancer: the guidelines. Chest 2003; 123(1 Suppl):147S-156S.

(21) Bossuyt PM, Reitsma JB, Bruns DE, Gatsonis CA, Glasziou PP, Irwig LM et al. The STARD statement for

reporting studies of diagnostic accuracy: explanation and elaboration. Ann Intern Med 2003; 138(1):W1-12.



38(4):108-114.

(23) Zimny M, Kaiser HJ, Cremerius U, Reinartz P, Schreckenberger M, Sabri O et al. Dual-head gamma camera

2-[fl uorine-18]-fl uoro-2-deoxy-D-glucose positron emission tomography in oncological patients: eff ects of

non- uniform attenuation correction on lesion detection. Eur J Nucl Med 1999; 26(8):818-823.

(24) Pieterman RM, van Putten JW, Meuzelaar JJ, Mooyaart EL, Vaalburg W, Koeter GH et al. Preoperative staging

of non-small-cell lung cancer with positron- emission tomography. N Engl J Med 2000; 343(4):254-261.

(25) Gonzalez-Stawinski GV, Lemaire A, Merchant F, O’Halloran E, Coleman RE, Harpole DH et al. A comparative

analysis of positron emission tomography and mediastinoscopy in staging non-small cell lung cancer. J

Thorac Cardiovasc Surg 2003; 126(6):1900-1905.

(26) Hamberg LM, Hunter GJ, Alpert NM, Choi NC, Babich JW, Fischman AJ. The dose uptake ratio as an index of

glucose metabolism: useful parameter or oversimplifi cation? J Nucl Med 1994; 35(8):1308-1312.



C h a p t e r

Summary and future directions

8


In Chapter 2, the lesion detectability of fi ve diff erent PET scanners incorporating four diff erent detector crystals was compared utilizing a thorax phantom and an observer study. In order to obtain a prediction of the relative clinical performance data acquisition and reconstructions as used in the clinical settings were chosen. Sphere size was selected to represent relevant clinical situations. A range of radioactivity concentrations in the spheres and background was used that was expected to demonstrate potential diff erences between scanners. Image presentation was specially designed to minimize any observer bias and ten independent observers were used. The high correlation demonstrated between contrast to noise ratio and lesion detectability suggests that this experimental set-up allows for estimation of in vivo scanner performance without the need for observer studies. Moreover, these in vitro results correlated well with those ones obtained in vivo as reported by others, using the same Biograph PET scanner (1). The highest lesion detectability was observed with the Philips Allegro camera and block detector scanners demonstrated overall better performance than non-block detector scanners.

This in vitro work focussed on the thoracic region. However, an advantage of the LSO-PS camera lies in its ability to perform whole body imaging in less than half the time of other dual head PET cameras. To this end, attempts were made to predict LSO-PS performance for the whole body imaging indication of an unknown primary tumour from observed BGO-PET performance.

In Chapter 3, the yield of full ring BGO-PET in the detection of unknown primary tumours presenting with extracervical metastases was found to be 23% (14/63, 95% CI 13-35%). The mean size of the true positive lesions was 4.1 cm (SD ± 2.7 cm). Nine of the true positive lesions were located in the thorax while fi ve were located in the abdominopelvic region. Based on the performance of the LSO-PS camera in the phantom study, we estimate that the LSO-PS camera would also have detected approximately 69% of these true positive lesions based on lesion size.

In Chapter 4, a systematic review was performed to evaluate the accuracy of attenuation corrected versus non-attenuation corrected FDG-PET. The added value of images with attenuation correction (AC) for routine whole body clinical imaging has been controversial. Even with the recent introduction of relatively rapid X-ray computed tomography transmission for attenuation correction, evaluation of both images with and without attenuation correction remains an integral part of image interpretation due to the possibility of artifact in certain situations with attenuation corrected images. No signifi cant diff erence in lesion detectability for images with and without attenuation correction for full ring PET was demonstrated. However, improved lesion detectability with attenuation correction was demonstrated for dual head PET imaging.

In Chapter 5, a fi rst in vivo comparison of the prototype LSO-PS camera versus full ring BGO-PET was performed. Randomly selected oncological patients with various malignancies referred for whole body PET underwent imaging with both full ring BGO-PET and LSO-PS on the same day. With full ring BGO-PET deemed as the gold standard, a true blinded comparison utilizing four independent observers demonstrated a sensitivity of 73% for LSO-PS as compared to BGO-PET. It was felt that these initial results justifi ed further evaluation of the LSO-PS camera. The next step involved evaluating the performance of LSO-PS in commonly accepted indications for PET



97

scanning such as the evaluation of indeterminate pulmonary nodules and the mediastinal staging of non-small cell lung carcinoma.

In Chapter 6, the performance of LSO-PS in the evaluation of pulmonary nodules (not larger than 3 cm) was examined. By excluding larger pulmonary lesions, smaller ones were again targeted where dual head PET cameras have demonstrated lower sensitivity. Patients referred for PET imaging with not more than 3 pulmonary nodules were prospectively included and underwent imaging with both BGO-PET and LSO-PS utilizing a similar same day protocol. Image analysis was performed utilizing a rigorous observer blinding system. The average tumour diameter was 1.7 cm (SD 0.7). LSO-PS demonstrated a high positive concordance with BGO-PET, both for sensitive (97%) and conservative (92%) assessment strategies.

In Chapter 7, the performance of LSO-PS with respect to the mediastinal staging of NSCLC was examined. A prospective study was performed in which consecutive patients with proven or suspected potentially operable non-small cell lung carcinoma (NSCLC) were included according to hilar and mediastinal lymph node size. Lymph node size was used as an inclusion criterion to allow better evaluation of smaller lymph nodes, for which a signifi cantly lower sensitivity had already been demonstrated with conventional dual head PET cameras. Another head-to head comparison with full ring BGO-PET and LSO-PS was performed. Not surprisingly, a higher lesion contrast with BGO-PET as compared to LSO-PS was found. LSO-PS identifi ed 64% of BGO-PET positive lymph node stations, resulting in a lower nodal stage in 14% of the patients and diff erent management advice in 9% of the patients.

Unlike the results of the systematic review (which suggested improved lesion detection with attenuation correction for dual-head PET cameras), there was no advantage of attenuation corrected versus non-attenuation corrected images for pulmonary nodules nor for mediastinal lymph node staging. This may suggest that the transmission technology of the prototype camera needs to be improved. Attenuation correction with radionuclide transmission source scanning has limitations because of the low photon fl ux, which increases statistical noise that can result in “errors” in the attenuation correction. In this regard, attenuation correction with X-ray transmission scanning in CT is a welcome alternative, even when performed with a low dose and a limited slice number due to the higher photon fl ux. The higher photon fl ux and the higher spatial resolution of the CT results in more accurate estimates of the densities of the diff erent tissues with less noise. Therefore, attenuation correction using the CT technique may perform better than using radionuclide transmission sources.

Looking back at the phantom experiments described in Chapter 1, it appears that the in vivo performance of the LSO-PS scanner was better than expected (at least for non-attenuation corrected images): in the phantom study only 40% of the 13 mm spheres detected with 2D HR+ were detected with the LSO-PS scanner, versus 80% (95% CI 58-92%) in the in vivo setting (lesions of 11-15 mm, solitary pulmonary nodule and mediastinal node data taken together). Apart from chance (small clinical samples, hence large confi dence intervals) it is hypothesized that this apparent discrepancy resulted from the relatively low activity ratios of the spheres to background


Chapter 8

98

in the phantom experiment. These ratios were derived from the NEMA protocols. This range of ratios indeed provided contrast between scanners and allowed comparison with diffi cult clinical applications such as low SUV pulmonary coin lesions. However, in retrospect, and after examining data from our non-small cell lung cancer database [data of CJ Hoekstra (2)], it was determined that the mean contrast measured in proven non-small cell lung cancer (NSCLC) was twice as high as the maximal sphere to background ratio used in the phantom study. Hence, the phantom data pertain to the clinical setting of the average breast cancer patient (SUV’s about 50% of NSCLC) rather than NSCLC. Unfortunately, when this became apparent, the experiments could not be repeated at those levels. The simple reason for this was that the manufacturer decided to halt the development of the LSO-PS scanner thus abolishing access.

Clinical scenarios

In the clinical domain, the relevant question is whether and how such technically inferior but markedly cheaper scanner technology should be implemented. One potential role is that of triage. PET is no longer a new research imaging modality under investigation; it is rapidly becoming the standard of clinical care in oncology (3-5). However, it remains relatively expensive and this is a signifi cant concern for governments and health care insurers who must provide the funding for it. Despite the growing number of full ring PET cameras available for clinical imaging, the number of

Figure 1. Probability of malignancy in a solitary pulmonary nodule (SPN) with negative and positive full ring BGO-PET

and dual head LSO-PS scans

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

Prior probability of malignant SPN (prevalence)

etsoP

rp roi

rytili

babo

fo t

nangila

m N

PS PET positive

PET negative

sensitive LSO+

sensitive LSO -



99

indications for PET imaging also continues to increase. In addition, geographical problems remain a concern in large countries such as Canada where the population density in more remote regions does not justify the placement of a full ring PET camera. Furthermore, in many poorer second and third world countries, the availability of PET imaging remains scarce or nonexistent.

Dual head PET cameras could potentially be used to triage patients requiring full ring PET. For example, suppose guidelines state that FDG-PET is indicated for radiologically indeterminate solitary lung nodules larger than 1cm in diameter with a pretest likelihood of malignancy of 10-50% (6). Assuming that observations about the performance of the prototype LSO-PS camera are valid, a prediction can be made that the negative predictive values of either test result are nearly identical, so that a negative LSO-PS scan, just as a negative full ring PET scan, would lead to an expectative watchful waiting policy, as modelled in Figure 1.

In the event of a FDG positive primary lesion at LSO-PS, mediastinal staging comes into play. It was found that the overall nodal stage classifi cation by LSO-PS was similar to full ring PET. However, at the level of identifying individual nodal stations LSO-PS was clearly inferior to full ring PET. In the past few years, less invasive but regionally complementary techniques for mediastinoscopy have emerged (esophageal and endobronchial ultrasound fi ne needle aspiration). Nowadays, patients are directed for further nodal evaluation depending on the localization of the involved nodes, and this is an argument in favour of techniques with higher sensitivity and localizing precision. Therefore, patients with a positive primary lesion without evidence of distant metastases or extrathoracic lymphatic involvement at LSO-PS should be referred for full ring PET imaging.

Future Directions

Diagnostic imaging is an area of medicine providing unique noninvasive information and, similar to other technology, is evolving at a rapid pace. PET imaging was fi rst performed in 1953 and PET imaging with FDG in humans in 1976. The last decade has seen a international acceptance of the benefi ts of FDG imaging associated with an exponential increase in the number of PET scanners worldwide. New crystals for PET imaging such as LSO and GSO have been developed and incorporated into PET cameras, the majority of which have come on the market within the past decade. The speed of attenuation correction has dramatically improved beginning with faster acquisition sequences for radionuclide based attenuation correction and most recently with the introduction of X-ray computed tomography attenuation correction. Even the manner in how we reconstruct PET images has fundamentally changed. Further advances will undoubtedly be developed and need to be eff ectively evaluated in a cost-effi cient manner. It has become clear that the standard procedure of test evaluation, i.e. defi ning test accuracy versus clinical standards followed by randomized trials, has diffi culty keeping pace with technological advances. This is not a conceptual limitation but a practical one (randomized trials would need to be performed earlier in the limited window of opportunity).


Chapter 8

100

This thesis explores an eff ective alternative to evaluate new technology. By performing an in vitro phantom comparison as well as utilizing a head-to-head in vivo comparison between the new technology (the prototype dual crystal dual head coincidence camera LSO-PS) and the established technology (full ring BGO-PET), the two were eff ectively compared in an effi cient manner, with an emphasis on intrathoracic oncological imaging. This analysis determined that the new technology had an acceptable performance as compared to other dual head PET imaging devices. In fact, the acquisition times were signifi cantly more rapid than other dual head PET cameras on the market. Normally, these results might have justifi ed the development of additional technical refi nements (e.g. with respect to the attenuation correction technology) to improve the prototype camera performance. However, no further development was performed on this prototype camera for a combination of political and economic reasons. The fi rst concerned the decision of Medicare (the funding body for public health care in the United States) to refuse reimbursement for all FDG-PET studies performed on dual head PET cameras, despite the fact that much of this evidence was based on early models with poorer performance. The second factor was the popularity and immediate acceptance of PET/CT cameras, especially in the United States, which were being introduced at approximately the same time that technical refi nements were needed on the prototype LSO-PS camera. Faced with limited manpower resources and the potentially higher marketability of the PET/CT cameras, CTI PET Systems chose to focus on the development of PET/CT cameras and further development of the LSO-PS prototype camera was halted. More recently, a similar process was observed after the introduction of PET-CT: even though the incremental cost-eff ectiveness of PET-CT versus PET was not proven, the scanner industry stopped production of whole body PET scanners. Again, commercial rather than scientifi c arguments prevailed.

Despite ongoing improvements in dual head PET technology, it appears that performance of dual head coincidence cameras will remain inferior to that of full ring PET cameras given the geometry limitations of two parallel detectors versus the presence of a full ring of detectors.

With the introduction of new crystals and computed tomography attenuation correction, a new generation of PET imaging devices has already arrived, and as stated before, the ‘standard full ring PET scanner’ has been replaced by PET-CT scanners and manufacturers are no longer selling full ring PET scanners without CT attenuation. Other new imaging devices or signifi cant upgrades of an established imaging device will undoubtedly be developed in the future. At the same time attempts to reduce cost, radiation exposure and patient delay will continue. Therefore, rapid but eff ective evaluation of these technological developments is imperative in providing patients with the best clinical care possible while at the same time eff ectively utilizing limited health care resources.

The principles used to evaluate the prototype camera in this thesis could be used in the evaluation of other new imaging technologies, with decisions to proceed or not to proceed (“go or no go”) possible at any phase of the evaluation. After initial measurements in standardized NEMA (National Electrical Manufacturers Association) phantoms, the next step would be to design an in vitro phantom experiment with properly blinded observers to predict potential in vivo performance. Only if in vitro results would suggest similar or improved performance of the new technology



101

versus the old standard test, would a limited, prospective head-to-head in vivo comparison be performed. Again, adequate blinding of the observers and a relevant patient spectrum (e.g. lesion size and distribution) are key issues. If the new technology would likely improve the clinical standard, an accuracy study versus the clinical standard or a randomized trial could be considered. If the new technology would be unlikely to improve the standard, or even likely to underperform but with major advantages at other levels (e.g. cost, radiation exposure, etc.), the setting of triage should be considered, and such an evaluation could largely follow the experimental design outlined in this thesis. Such an approach is justifi ed as health care resources become progressively more scarce in the face of continuing technological advancement.

References



47(3):426-431.

(2) Hoekstra CJ, Stroobants SG, Hoekstra OS, Vansteenkiste J, Biesma B, Schramel FJ et al. The value of

[18F]fl uoro-2-deoxy-D-glucose positron emission tomography in the selection of patients with stage IIIA-N2

non-small cell lung cancer for combined modality treatment. Lung Cancer 2003; 39(2):151-157.

(3) Gould MK, Sanders GD, Barnett PG, Rydzak CE, Maclean CC, McClellan MB et al. Cost-eff ectiveness of

alternative management strategies for patients with solitary pulmonary nodules. Ann Intern Med 2003;

138(9):724-735.

(4) Juweid ME, Cheson BD. Positron-emission tomography and assessment of cancer therapy. N Engl J Med

2006; 354(5):496-507.








C h a p t e r

Met andere woorden (samenvatting in het Nederlands)

9


Inleiding

Het eerste apparaat dat positron-emitterende radionucliden gebruikte (in dit geval om hersentumoren te lokaliseren) werd in 1953 beschreven door Bronwell en Sweet (1). Dertien jaar later had men een apparaat ontwikkeld dat regionale hersenbloeddoorstroming kon meten. Dat gebeurde op het Brookhaven National Laboratory, dat later verhuisde naar het Neurologische Instituut van Montreal.

Het idee van tomografi sche, beeldvormende diagnostiek werd eind jaren zestig van de vorige eeuw geintroduceerd, op basis van plaatsing of rotatie van detectoren rond het lichaam(sdeel). Mathematische algoritmen werden gebruikt om de data in bepaalde vlakken te reconstrueren. De eerste rudimentaire tomografi sche PET- beelden werden gemaakt door Kuhl en Edwards in 1968 (2). In 1975 verbeterde men de beeldkwaliteit door de ontwikkeling van de ‘fi ltered back projection’ reconstructie techniek door Ter-Pogossian (3;4). In Figuur 1 ziet u een voorbeeld van een van de eerste tomografi sche PET scanners. De eerste apparaten konden alleen beelden van de hersenen maken vanwege de kleine diameter van de detectorring. De beperking lag in de hoge kosten van de detectorkristallen en van de benodigde elektronica.

Figuur 1. voorbeeld van een van de eerste tomografi sche PET scanners

(uit: Wagner HN, Semin Nucl Med 1998; 28: 213-220)


Met andere woorden

105

Pas rond eind jaren ’80 leidde een meer effi ciënte vervaardiging van kristallen en snellere elektronica (om lichtsignalen in elektronische signalen om te zetten alsook voor het verwerken van data) tot de ontwikkeling van scanners die tegen acceptabele kosten whole body diagnostiek mogelijk maakten. Daar was vraag naar omdat de radiologische CT technologie (die ca. 10 jaar eerder geintroduceerd was) gaandeweg toch ook beperkingen bleek te hebben.

De thans meest gebruikte tracer voor oncologische diagnostiek is een radioactieve vorm van glucose gekoppeld aan radioactief fl uoratoom: 18F-Fluoro-deoxyglucose (FDG). Het gebruik van FDG is gebaseerd op de observaties van Warburg in de jaren 1930: hij vond dat tumorcellen een verhoogde glucose stofwisseling (glycolyse) hadden in vergelijking met normale cellen (5). In tegenstelling tot glucose wordt deoxyglucose (zoals in FDG) door de cel opgenomen zonder verdere stofwisseling te ondergaan. FDG is de meeste gebruikte tracer voor oncologische beeldvormende diagnostiek vanwege zijn hoge affi niteit tot meerdere soorten tumorcellen en zijn gunstige biodistributie. De meest gebruikte toepassingen in de oncologie hebben betrekking op stadiëring en evaluatie van de respons op behandeling. Dat veranderingen van FDG opname tijdens behandeling prognostische waarde hebben, staat allang vast, maar hoe individuele geneesmiddelen ingrijpen op cellulaire mechanismen die uiteindelijk leiden tot verminderde FDG opname in geval van eff ectieve behandeling, is kennis die meer recent verkregen is (6).

FDG diagnostiek met gamma camera’sFDG-PET werd sinds begin jaren ‘90 internationaal geleidelijk erkend als een belangrijk instrument voor de diagnostiek van verschillende kwaadaardige tumoren. Echter de hoge kosten (van scanners en FDG) verhinderden een meer wijdverspreid gebruik van PET. Dat bleef vooralsnog een groot probleem: zo waren er zelfs in 1997 in Nederland nog maar 2 scanners (Groningen, VU Amsterdam). Daarom werd intensief gezocht naar goedkopere alternatieven, zoals het gebruik van gewone gamma-camera’s die toch al in elke afdeling nucleaire geneeskunde stonden. In eerste instantie werden deze camera’s uitgerust met speciale collimatoren om de hoge energetische fotonen van positron-emitters te verwerken. In het begin werden planaire beelden gemaakt gevolgd bij tomografi sche beelden waarbij maar een van de twee 511 keV fotonen werd gebruikt; kort daarna volgde tomografi sche beeldvorming met SPECT camera’s (7-10). Deze technologie leek vooral in de cardiologie veelbelovend en wel om vitaliteit van het myocard te evalueren (11). Rond 1993 werden hele sessies bij internationale nucleairgeneeskundige congressen gewijd aan beeldvormende diagnostiek met FDG een hoge energie collimatoren, en het VU-ziekenhuis (in een samenwerking van de afdelingen cardiologie en nucleaire geneeskunde) had daarbij een leidende positie. In de (haemato-)oncologie leek de technologie toepasbaar bij visuele beoordeling van responsiviteit op behandeling van lymfklierkanker (7;8). Echter de sensitiviteit voor kleinere lesies was te laag en dat maakte de technologie minder geschikt voor stadiëringsonderzoek.Vanwege de nog steeds beperkte beschikbaarheid en de hoge kosten van volle ring PET camera’s bleef men zoeken naar goedkopere alternatieven. Halverwege de jaren ’90 leidde dit tot een nieuwe generatie dubbelkops ‘PET camera’s’: men kon de camera zonder collimatoren in ‘coïncidentie mode’ laten werken om de twee simultane fotonen te detecteren (12-15). Het grote voordeel van deze tweede generatie dubbelkops ‘PET camera’s’ was hun hogere sensitiviteit in vergelijking met de eerste generatie scanners met hoge energie collimatoren.


Chapter 9

106

Kristal technologieTegelijkertijd onderzocht men de toepasbaarheid van verschillende kristallen: aanvankelijk werden in de meeste volle ring PET camera’s bismuth germanaat (BGO) en natrium jodide (NaI) kristallen toegepast. Eind jaren ‘80 en begin jaren ‘90 werden nieuwe kristallen zoals lutetium - en gadolinium oxyorthosilicaat [resp. LSO en GSO (16;17)] geïntroduceerd. NaI, het standaard kristal van de gamma-camera, is minder geschikt voor positron-emissie tomografi e door zijn lagere dichtheid (18). In eerste instantie gebruikte men daarom dikkere kristallen [5/8 - in plaats van 3/8 inch (19)]. Een andere oplossing was de ontwikkeling van een nieuw type dubbelkops PET camera die twee aan elkaar gekoppelde kristallen bevat (‘phoswich’ kristal genoemd). Het eerste model bevatte LSO en yttrium orthosilicate (YSO) kristallen (20). Dit werd opgevolgd door een LSO/NaI phoswich detector systeem [LSO PET-SPECT (21)], waarop het onderzoek van dit proefschrift is gebaseerd. Een potentieel voordeel van de koppeling van het LSO kristal aan het NaI kristal is de grotere dichtheid in vergelijking met andere bestaande dubbelkops PET camera’s, waardoor snellere scans kunnen worden gemaakt. Bovendien werden intussen de nieuwe kristallen zoals LSO en GSO ook in volle ring PET camera’s gebruikt (22;23).

Beeldreconstructie ontwikkelingenEen algemeen probleem van PET beeldvorming is de verspreiding of absorptie van fotonen voordat ze door de camera geregistreerd worden. Die fotonen worden niet of op een andere locatie gedetecteerd, zoals in Figuur 2 is te zien. Het eindresultaat is vervorming van de ware verdeling van de straling in het lichaam. Attenuatie correctie wordt toegepast om voor absorptie van straling in het lichaam te corrigeren, voor zowel volle ring en dubbelkops PET camera’s (24). De toevoeging van attenuatie correctie heeft voor- (zoals ook de mogelijkheid om de radioactiviteit te kwantifi ceren) en nadelen (zoals de introductie van statistische ruis). In een tijd van schaarste is een belangrijk nadeel van attenuatie correctie dat de transmissie-scan waarmee per beeldpunt een “verzwakkings-kaart” van de patient wordt opgemaakt, veel tijd kost: in een routine klinische praktijk van een volle ring BGO-PET scanner leidt toevoeging van dergelijke transmissie-scans aan de emissie-scans waarmee de verdeling van radio-aktiviteit in de patient wordt gemeten,

Figuur 2


Met andere woorden

107

tot 30% productie-verlies. De toegevoegde waarde van attenuatie correctie voor wat betreft lesie-detectie is verder controversieel (25). Bij dubbelkops NaI gamma-camera’s wordt bovendien attenuatie-correctie nog bemoeilijkt door de relatief lage dichtheid van de NaI detectorkristallen en de niet-optimale plaatsing van transmissie-bronnen op de twee camerakoppen versus de roterende ring van transmissiebronnen zoals die gebruikt wordt bij volle ring PET scanners. De transmissie scans voor attenuatie correctie zijn ook wel gebruikt om voor scatter te corrigeren waarbij de transmissie scan wordt gebruikt om de hoeveelheid verspreiding te schatten. Echter, deze technieken zijn tijdrovend en niet algemeen gebruikelijk.

De uitdagingSinds 1990 is er een enorme toename in de aanvraag naar positron emissie tomografi e zowel voor klinische als wetenschappelijke toepassingen. Tegelijkertijd is er ongerustheid over de kosten voor de gezondheidszorg. Middelen zijn beperkt en moeten verdeeld worden tussen de diagnostiek en behandeling. Recent heeeft men nieuwe, dure en soms bijzonder eff ectieve ‘targeted’ behandelingen ontwikkeld, die geïntroduceerd moeten worden in dagelijkse patiëntenzorg. In de komende decade is de uitdaging voor beeldvormende diagnostiek om methoden te ontwikkelen die de behandelend specialist kunnen helpen om de juiste behandeling met dergelijke middelen voor een individuele patient te vinden. Het is inmiddels allang duidelijk dat vooruitgang stagneert indien er geen technieken komen die voorbij de huidige, op anatomie en histologie gebaseerde classifi catie-systemen gaan. Daarnaast wordt de gezondheidszorg onbetaalbaar indien dure nieuwe geneesmiddelen via ‘trial and error’ methodiek, vertrekkend van de ‘gemiddelde’ patient moeten worden toegepast.

Net als met nieuwe behandelingen moeten nieuwe testen in de beeldvormende diagnostiek ook zorgvuldige validatie ondergaan. In dit opzicht hebben wij en anderen hiërarchische schema’s ontwikkeld om de kosteneff ectiviteit van beeldvormend onderzoek te evalueren (26). Wij hebben intussen ook ondervonden, dat doorlopen van de hele evaluatie-cyclus voor iedere nieuwe ontwikkeling ondoenlijk is, en teveel tijd kost in vergelijking met de evolutie in de klinische praktijk: bij de afronding van onze tweede gerandomiseerde trial in longkanker was inmiddels de PET-CT technologie op de markt gekomen, zodat de uitkomsten van ons onderzoek (met klassieke volle ring PET scanners) minder relevant werden voor de praktijk (27;28). Zoals eerder beschreven is het verfi jnen van de uitkomsten van beeldvormende diagnostiek (dat wil zeggen camera hardware en beeld reconstructie software) een doorlopend proces. Meestal gaat het daarbij om kleine verbeteringen en niet om grote, fundamentele stappen. De uitdaging voor statistici en clinici is om snelle, effi ciente strategieën te ontwikkelen voor de evaluatie van bv. nieuwe scanners. In het kader van dit proefschrift onderzochten wij zo’n strategie aan de hand van een prototype LSO/NaI PET/SPECT camera (verder aangeduid als LSO-PS).

In Hoofdstuk 2 vergeleken we het lesie detectievermogen van vijf verschillende PET scanners (4 ring scanners en het prototype LSO-PS, met vier verschillende kristallen) in een thoraxfantoom. Dit fantoom bevatte bolletjes van verschillende grootten, gevuld met vastgelegde concentraties radio-aktiviteit ingebed in een achtergrond (ook gevuld met radio-aktiviteit) die het mediastinum


Chapter 9

108

en longen moest voorstellen. De beelden werden verkregen in verschillende ziekenhuizen, en gebruikmakend van de standaard klinische procedures met betrekking tot data-acquisitie en –reconstructie. Vervolgens hebben 10 waarnemers onafhankelijk van elkaar en geblindeerd voor de waarheid de beelden beoordeeld. We vonden een goede correlatie tussen lesie detectie en de contrast-ruis verhouding. De verkregen gegevens suggereerden dat de contrast-ruis ratio zou kunnen worden gebruikt om detectie te voorspellen zodat dergelijke tijdrovende observerstudies niet langer nodig zouden zijn. Bovendien was er een goede correlatie tussen onze in vitro resultaten en de in vivo resultaten, zoals gepubliceerd door anderen die onderzoek deden naar longhaarden met een relatief lage FDG opname (29). De hoogste detectie sensitiviteit zagen we met de Allegro scanner. De waarnemingen bij de ons klinisch bekende HR+ volle ring BGO scanner suggereerden dat we met het fantoom aan de ondergrens van klinische relevantie hadden gemeten (relatief lage contrasten). Dat werd bevestigd in de vergelijking met de eerder genoemde studie van longhaarden met een lage FDG opname. Zoals verwacht, presteerde de LSO-PS scanner relatief het minst. Lesie-grootte was daarbij een belangrijke variabele.

Deze fantoom-evaluatie betrof dus de thoraxregio. Echter, in vergelijking met de bestaande dubbelkops coincidentie systemen, had de LSO-PS scanner als voordeel dat ‘whole body scans’ gemaakt konden worden in minder dan de helft van de tijd. Om die reden werd een poging gedaan om de LSO-PS opbrengst te voorspellen voor whole body scans voor patiënten met onbekende primaire tumoren ten opzichte van bekende BGO-PET performance, waarbij we in eerste instantie het accent legden op lesie-grootte.

In Hoofdstuk 3 onderzochten we de opbrengst van volle ring BGO-PET bij patiënten met een onbekende primaire tumor die zich presenteert met metastasen buiten het hoofd-hals gebied. Die opbrengst was 1 gevonden tumor op 4 met PET onderzochte patienten (rendement 23%, 95% CI 13-35%). De mediane grootte van de terecht-positieve lesies was 4 cm. Negen van deze lesies bevonden zich in de thorax en 5 in het abdomen of bekken. Gecombineerd met de fantoombevindingen van hoofdstuk 2, verwachtten wij op basis hiervan, dat de grootte van dergelijke lesies niet de beperkende factor zou moeten zijn voor toepassing van LSO-PS bij deze indicatie, al was het maar als eerste triage voor PET.

De eerder geschetste problematiek rond het al dan niet toepassen van attenuatie-correctie bij klinische PET onderzoeken onderzochten wij in een systematisch review (hoofdstuk 4). Daarin vergeleken wij de accuratesse van FDG-PET (volle ring scanners en gamma-camera’s) met en zonder attenuatie correctie. Zelfs met de introductie van relatief snelle röntgen transmissie computer tomografi e voor attenuatie correctie blijft evaluatie van beide beelden (met en zonder attenuatie correctie) een belangrijk deel van beeld interpretatie, al was het maar om eventuele door CT geinduceerde artefacten te herkennen in de voor verzwakking gecorrigeerde beelden. Dit literatuur onderzoek toonde geen signifi cant verschil in lesie detectie tussen beelden met en zonder attenuatie correctie voor volle ring PET; bij gamma-camera’s was er echter wel een voordeel voor beelden met attenuatie correctie.


Met andere woorden

109

In Hoofdstuk 5 beschrijven wij de eerste in vivo vergelijking van de LSO-PS camera versus volle ring BGO-PET. Willekeurig gekozen oncologische patiënten ondergingen op dezelfde dag een whole body scan met zowel de BGO-PET als de LSO-PS; elke patient is dus zijn/haar eigen controle (‘head-to’head’ vergelijking). Wij beschouwden de BGO-PET als de referentie techniek (de gouden standaard); in vergelijking met andere studies hebben wij een meer stringente methode van blindering toegepast, met als kenmerk dat ieder van een groep van 4 onafhankelijke waarnemers slechts 1 scanmodaliteit per patient beoordeelde. We vonden een relatieve sensitiviteit voor de LSO-PS scanner van 73% (ten opzichte van BGO-PET). Op basis van deze resultaten vonden wij nadere evaluatie van de LSO-PS camera gerechtvaardigd voor gebruikelijke PET scan indicaties zoals de evaluatie van radiologisch onduidelijke long lesies en de mediastinale stadering van niet-kleincellig longcarcinoom.

In Hoofdstuk 6 onderzochten we de accuratesse van LSO-PS bij radiologisch onduidelijke longlesies (niet groter dan 3 cm). Patiënten met niet meer dan 3 longlesies werden prospectief geïncludeerd, en er werden bij hen op dezelfde dag scans gemaakt met zowel BGO-PET als LSO-PS. De gemiddelde lesie-grootte was 1.7 cm (SD 0.7). Met hetzelfde studie-ontwerp als in hoofdstuk 5, vonden wij een hoge mate van overeenstemming tussen BGO-PET en LSO-PS. Wij hadden nu ook ruimte gemaakt voor beoordelingen die ‘onzeker’ waren, zoals gebruikelijk in de klinische praktijk. Opvallend was dat de accuratesse bij LSO-PS slechts in geringe mate afhankelijk bleek van deze variabele.

In Hoofdstuk 7 werd de accuratesse van LSO-PS ten opzichte van BGO-PET bij stadiëring van het mediastinum bij niet-kleincellig longcarcinoom onderzocht. Wij verrichten daartoe opnieuw een prospectief onderzoek waarbij opeenvolgende patiënten met mogelijk operabele ziekte werden geïncludeerd. Om een goed beeld te krijgen van de opbrengst van het prototype, werden patientengroepen geselecteerd met tevoren gedefi nieerde grootten van hilaire en mediastinale lymfklieren op de CT scan. Dit, omdat bekend was uit de fantoomstudies (en eerdere ervaringen van anderen met dubbelkops coincidentie systemen) dat grootte een belangrijke beperkende factor voor gamma-camera’s was. Opnieuw pasten wij het studie-ontwerp toe dat ook bij de longhaarden werd gebruikt (strikte blindering van waarnemers, BGO-PET als referentie-techniek). Daarnaast onderzochten wij niet alleen op lesie-detectie niveau maar ook de kwaliteit van conclusies over de globale lymfklierstadiering (het ‘N-stadium’) en de invloed op beleidsadviezen.

Wij vonden in 44 evalueerbare patiënten een te laag N-stadium bij 14% van de patienten met LSO-PS, hetgeen leidde tot een ander beleidsadvies bij 1 op de 11 patiënten. Niet onverwacht bleek het subjectief waargenomen lesie-contrast bij BGO-PET ook hoger dan bij LSO-PS. In totaal bleek 64% van de BGO-PET positieve lymfklieren ook zichtbaar met LSO-PS. Er was echter een duidelijke, evenredige relatie tussen lymfkliergrootte en concordantie tussen BGO-PET en LSO-PS. Van de 20 met BGO-PET positieve lymfklierstations, waren 14 ≤ 10 mm (op de CT scan), 2 waren tussen 10-15 mm groot, en 1 was groter dan 15 mm; echter, 16/30 van de BGO-PET positieve stations ≤ 10 mm werden ook gedetecteerd met LSO-PS (53%, 95%CI: 36-70%). Er was geen fout-positiviteit bij de beoordeling van de LSO-PS beelden (in vergelijking met BGO-PET).


Chapter 9

110

Het viel ons op dat wij, in tegenstelling tot de eerder gepubliceerde resultaten over attenuatie correctie, geen enkel voordeel vonden voor attenuatie correctie bij LSO-PS. Dit gold voor beide klinische longkankerstudies. Deze uitkomst zou kunnen betekenen dat de transmissie technologie van het prototype camera verbeterd moet worden.

Klinische scenario’s

De relevante vraag in het klinische domein is of en hoe een technologisch inferieure maar duidelijk goedkopere scanner technologie gebruikt moet worden. Een mogelijkheid is die van triage. PET is geen nieuwe technologie meer die alleen voor wetenschappelijk onderzoek geschikt is; het wordt snel de standaard van klinische zorg in de oncologie (30-32). Toepassing blijft echter tamelijk kostbaar en dat is van belang voor overheden en verzekeraars die het moeten betalen. Ondanks het toenemend aantal volle ring PET camera’s dat beschikbaar is voor klinisch gebruik, neemt de hoeveelheid klinische indicaties voor PET toe. Met name in landen als Canada met grote afstanden en een lage bevolkingsdichtheid is de aanschaf van een volle ring PET camera in veraf gelegen gebieden niet verantwoord. Ook blijft de beschikbaarheid van PET beeldvormende diagnostiek in tweede- en derdewereldlanden schaars of helemaal afwezig. Dubbelkops PET camera’s zouden daar gebruikt kunnen worden als triage: te besluiten welke patienten nog aanvullend, in een centrum, een volle ring PET nodig hebben.

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00

Prior probability of malignant SPN (prevalence)

etsoP

rp roi

rytili

babo

fo t

nangila

m N

PS PET positive

PET negative

sensitive LSO+

sensitive LSO -

Figuur 3. Kans op maligniteit in solitaire longhaarden (solitary pulmonary nodules, SPN) als functie van voorafkans

op kanker (verwachte prevalentie bij bevolkingsonderzoek of klinische inschatting door de longarts) en de test

resultaten met volle ring BGO-PET (aangeduid als PET) en LSO-PS (aangeduid als LSO).


Met andere woorden

111

Laten we aannemen dat PET geindiceerd is voor radiologisch onduidelijke long lesies groter dan 1 cm (33). Als wij er van uit gaan dat onze waarnemingen met de LSO-PS scanner bij die patienten valide zijn, dan kunnen wij voorspellen dat de negatief voorspellende waarde van BGO-PET en LSO-PS scan resultaten vrijwel gelijk zijn (Figuur 3). Dus zou een negatieve LSO-PS scan, net als een negatieve volle ring PET scan, leiden tot een expectatief, afwachtend (‘watchful waiting’) beleid, met name in gevallen met een lage klinische voorafkans op kanker (de meest voorkomende toestand).

Als een primaire lesie op LSO-PS positief is, dan wordt mediastinale stadiëring belangrijk. Wij hebben aangetoond dat het globale lymfklierstadium bij LSO-PS niet bijzonder sterk verschilt van volle ring PET. Echter, het vermogen van LSO-PS om individuele FDG positieve, en dus voor maligniteit verdachte, klieren te identifi ceren is duidelijk minder, en dat is een probleem. In de laatste jaren zijn minder invasieve alternatieven voor mediastinoscopie als slokdarm- en endobronchiale echogeleide naaldbiopsie geintroduceerd. Omdat er dus voorkeur is voor technieken met een hoge sensitiviteit en nauwkeurigheid per station, is een volle ring PET(-CT) scan aan te bevelen voor patiënten die op een LSO-PS scan een FDG-positieve primaire tumor hebben zonder aanwijzingen voor metastasen op afstand.

Blik op de toekomst

Beeldvormende diagnostiek is een gebied van de geneeskunde dat unieke niet-invasieve informatie geeft en dat net als andere technologie “à tempo” verandert. Beeldvormende diagnostiek met PET werd het eerste toegepast in 1953 en PET beelden in mensen met FDG werden voor het eerst gemaakt in 1976. In het voorbije decennium hebben wij internationaal erkenning gezien van de voordelen van beeldvormende diagnostiek met FDG. Dat leidde tot sterke toename van het aantal PET scanners, maar niet overal ter wereld. Nieuwe kristallen voor PET diagnostiek zoals LSO en GSO zijn ontwikkeld en geïntegreerd in PET camera’s. De meeste camera’s zijn de laaste tien jaar op de markt gekomen. Het tijdsverlies door attenuatie correctie scans is sterk afgenomen sinds de introductie van röntgen computer tomografi e voor attenuatie correctie. Zelfs de manier waarop wij PET beelden reconstueren is fundamenteel veranderd. Het is te verwachten dat de eerste PET-MRI scanners binnenkort geintroduceerd worden. Deze nieuwe ontwikkelingen moeten snel op hun kosteneff ectiviteit beoordeeld kunnen worden, idealiter voordat ze de kliniek instromen.

Het is duidelijk dat de huidige manier van het evalueren van technieken, dat wil zeggen het bepalen van de accuratesse van een techniek versus een klinische gouden standaard gevolgd door gerandomiseerd onderzoek, het tempo van de technologische verbeteringen niet altijd kan bijbenen. Overigens, dit is niet automatisch een conceptuele beperking van het eerder opgestelde hierarchische evaluatie-systeem, maar eerder een praktische beperking (gerandomiseerd onderzoek zou eerder uitgevoerd moeten worden, gegeven de beperkte tijd).

In dit proefschrift hebben wij zo’n alternatief om nieuwe technologie te evalueren onderzocht, en ditmaal in een voorziene toepassing als triage-instrument, waarbij tevoren zeer aannemelijk


Chapter 9

112

was dat de nieuwe technologie geen verbetering in de zin van diagnostische accuratesse zou meebrengen. De nieuwe gammacamera technologie (LSO-PS) echter was veel sneller dan bestaande dubbelkopssystemen. Door het gebruik van een in vitro fantoom vergelijking zowel als een directe ‘head-to-head’ in vivo vergelijking tussen de nieuwe technologie (prototype dubbelkops LSO-PS scanner) en de gevestigde technologie (volle ring BGO-PET) was het mogelijk de twee technologieën effi ciënt te vergelijken met accent op oncologische beeldvormende diagnostiek in de thorax. Wij hebben aangetoond dat de nieuwe technologie een redelijke performance had in vergelijking met andere dubbelkops PET camera’s. Achteraf bleek dat onze keuze voor standaard aktiviteits-verschillen tussen bollen en achtergrond ook een beperking had: longkanker lesies bleken in werkelijkheid 2x zoveel aktiviteit op te nemen als toegepast in het fantoom.

Normaal gesproken zouden de uitkomsten hebben gerechtvaardigd verdere technische verbeteringen van de LSO-PS scanner (bv de techniek voor attenuatie correctie) te ontwikkelen. Paradoxaal genoeg was de ontwikkeling van de LSO-PS camera tegen die tijd echter al door de fabrikant gestopt, zodat wij ook de hypothese dat de in vitro opstelling ook in dat domein de accuratesse goed kon voorspellen, niet eens konden toetsen. De ontwikkeling is gestopt om politieke en economische redenen. De eerste categorie betreft de beslissing van Medicare, de publieke gezondsheidszorgverzekeraar in de Verenigde Staten, om geen FDG-PET scan met dubbelkops PET camera’s meer te vergoeden. Deze beslissing was gebaseerd op een analyse van de inferieure accuratesse van de bestaande gammacamera’s. De tweede factor was de populariteit en onmiddellijke aanvaarding van PET-CT camera’s, vooral in the Verenigde Staten, die tegelijkertijd werden geïntroduceerd toen technische verbeteringen op de LSO-PS camera nodig waren. Door een tekort aan arbeidskrachten en de potentiëel veel grotere markt voor PET-CT scanners (aantrekkelijk voor kleinere ziekenhuizen omdat die scanners immers ook volwaardige CT scans kunnen maken en niet per se alleen voor de dure PET scans gebruikt hoeven te worden), heeft CTI PET Systems ervoor gekozen om zich te concentreren op de ontwikkeling van PET-CT cameras en de verdere ontwikkeling van de LSO-PS camera te stoppen. Meer recent hebben wij een vergelijkbaar proces gezien met de introductie van PET/CT: zelfs toen de kosten-eff ectiviteit van PET-CT nog niet bewezen was, was de productie van volle ring PET scanners zonder CT al geschrapt. Commerciële belangen hebben duidelijk de overhand gekregen boven wetenschappelijk bewijs.

Met de introductie van nieuwe kristallen en attenuatie correctie met computer tomografi e, is er alweer een nieuwe generatie van PET camera’s. Zoals eerder beschreven is de ‘standaard volle ring PET camera’ al vervangen door de PET-CT scanners en PET scanners zonder CT attenuatie correctie worden niet meer verkocht. In de toekomst zullen er altijd nieuwe scanners of reconstructie methoden ontwikkeld worden. Tegelijk zullen ook de pogingen om kosten te verminderen, de patiëntenstralingsdosis te verminderen en de wachttijd te verkorten belangrijk blijven. Daarom zal snelle en eff ectieve evaluatie van nieuwe beeldvormende diagnostiek van groot belang blijven om enerzijds de beste klinische zorg te leveren en anderzijds de kosten te beperken.

De beginselen die wij in dit proefschrift gebruikten om het prototype te evalueren, kunnen bij


Met andere woorden

113

andere scanners dienen om op enig punt van de evaluatie te beslissen al dan niet door te gaan (‘go’ of ‘no go’). Na de eerste metingen met de gestandaardiseerde NEMA (National Electrical Manufacturers Association ) fantomen, zou de volgende stap een in vitro fantoom studie moeten zijn. Daarbij kan een eerste selectie worden gemaakt op basis van uit de beelden berekende contrast-ruis verhoudingen, eventueel aangevuld met detecteerbaarheidsvergelijkingen met menselijke waarnemers, ten einde de in vivo performance te voorspellen. Afhankelijk van de voorziene toepassing (triage voor -, of verbeterde accuratesse ten opzichte van bestaande scanners) kan men dan verder. In het laatste geval: ‘no go’ als de in vitro resultaten geen betere performance van de nieuwe technologie versus de oude standaard suggereren. In het eerste geval: ‘go’ als de resultaten een acceptabele opbrengst suggereren, gevolgd door een prospectieve, directe ‘head-to-head’ in vivo vergelijking, in een relevant patientenspectrum met adequate blindering van waarnemers en de standaard methode als referentie techniek in plaats van de gangbare complexe accuratesse studies met klinische of histologische uitkomstmaten. Alleen als deze eerste stappen gunstig uitpakken, en er voldoende twijfel resteert met betrekking tot het klinisch rendement, zou evaluatie van accuratesse versus een klinische gouden standaard of een gerandomiseerd onderzoek overwogen kunnen worden.

Op het eerste gezicht staat de gang van zaken rond de op- en ondergang van de LSO-PS scanner op gespannen voet met het hoofdthema van dit proefschrift – onderzoek naar alternatieven om snelle evaluatie van nieuwe technologie mogelijk maken –. Stopzetting van het programma belemmerde immers patientenstudies, en zelfs verbeterde fantoom-experimenten. Bij nadere beschouwing toont ze ons inziens echter vooral ook de juistheid van de vraag naar dergelijke methodologie aan.

Referenties

(1) Bronwell GL, Sweet WH. Localization of brain tumors. Nucleonics 1953; 11:40-45.

(2) Kuhl DE, Edwards RQ. Reorganizing data from transverse section scans of the brain using digital processing.

Radiology 1968; 91(5):975-983.

(3) Ter Pogossian MM, Phelps ME, Hoff man EJ, Mullani NA. A positron-emission transaxial tomograph for

nuclear imaging (PETT). Radiology 1975; 114(1):89-98.

(4) Wagner HN, Jr. A brief history of positron emission tomography (PET). Semin Nucl Med 1998; 28(3):213-220.

(5) Warburg O. The Metabolism of Tumors. New York, NY: Smith Press, 1931.

(6) Kelloff GJ, Krohn KA, Larson SM, Weissleder R, Mankoff DA, Hoff man JM et al. The progress and promise of

molecular imaging probes in oncologic drug development. Clin Cancer Res 2005; 11(22):7967-7985.

(7) Hoekstra OS, Ossenkoppele GJ, Golding R, van Lingen A, Visser GW, Teule GJ et al. Early treatment response

in malignant lymphoma, as determined by planar fl uorine-18-fl uorodeoxyglucose scintigraphy. J Nucl Med

1993; 34(10):1706-1710.

(8) Hoekstra OS, van Lingen A, Ossenkoppele GJ, Golding R, Teule GJ. Early response monitoring in malignant

lymphoma using fl uorine-18 fl uorodeoxyglucose single-photon emission tomography. Eur J Nucl Med

1993; 20(12):1214-1217.


Chapter 9

114

(9) Macfarlane DJ, Cotton L, Ackermann RJ, Minn H, Ficaro EP, Shreve PD et al. Triple-head SPECT with 2-

[fl uorine-18]fl uoro-2-deoxy-D-glucose (FDG): initial evaluation in oncology and comparison with FDG PET.

Radiology 1995; 194(2):425-429.



(11) Huitink JM, Visser FC, van Lingen A, Groeneveld AB, Bax JJ, van Leeuwen GR et al. Feasibility of planar

fl uorine-18-FDG imaging after recent myocardial infarction to assess myocardial viability. J Nucl Med 1995;

36(6):975-981.

(12) Boren EL, Jr., Delbeke D, Patton JA, Sandler MP. Comparison of FDG PET and positron coincidence detection

imaging using a dual-head gamma camera with 5/8-inch NaI(Tl) crystals in patients with suspected body

malignancies. Eur J Nucl Med 1999; 26(4):379-387.

(13) Delbeke D, Patton JA, Martin WH, Sandler MP. FDG PET and dual-head gamma camera positron coincidence

detection imaging of suspected malignancies and brain disorders. J Nucl Med 1999; 40(1):110-117.

(14) Shreve PD, Steventon RS, Deters EC, Kison PV, Gross MD, Wahl RL. Oncologic diagnosis with 2-[fl uorine-

18]fl uoro-2-deoxy-D-glucose imaging: dual-head coincidence gamma camera versus positron emission

tomographic scanner. Radiology 1998; 207(2):431-437.



38(4):108-114.

(16) Holte S, Ostertag H, Kesselberg M. A preliminary evaluation of a dual crystal positron camera. J Comput

Assist Tomogr 1987; 11(4):691-697.


Trans.Nucl.Sci. 39, 502-505. 1992.

(18) Budinger TF. PET instrumentation: what are the limits? Semin Nucl Med 1998; 28(3):247-267.

(19) Delbeke D, Sandler MP. The role of hybrid cameras in oncology. Semin Nucl Med 2000; 30(4):268-280.

(20) Dahlbom M, MacDonald LR, Eriksson L. Performance of a YSO/LSO phoswich detector for use in a PET/

SPECT system. IEEE Trans Nucl Sci 1997; 44:1114-1119.


phoswich detector for a combined PET/SPECT imaging system. J.Nucl.Med. [39], P9. 1998.



Nucl Med 2004; 45(5):813-821.


45(6):1040-1049.

(24) Turkington TG. Attenuation correction in hybrid positron emission tomography. Semin Nucl Med 2000;

30(4):255-267.

(25) Wahl RL. To AC or not to AC: that is the question. J Nucl Med 1999; 40(12):2025-2028.

(26) Hayat-Kean MA. Cancer Imaging. New Jersey: 2006.


Met andere woorden

115

(27) Herder GJ, Kramer H, Hoekstra OS, Smit EF, Pruim J, van Tinteren H et al. Traditional versus up-front [18F]

fl uorodeoxyglucose-positron emission tomography staging of non-small-cell lung cancer: a Dutch

cooperative randomized study. J Clin Oncol 2006; 24(12):1800-1806.

(28) Lardinois D. New horizons in staging for non-small-cell lung cancer. J Clin Oncol 2006; 24(12):1785-1787.



47(3):426-431.

(30) Gould MK, Sanders GD, Barnett PG, Rydzak CE, Maclean CC, McClellan MB et al. Cost-eff ectiveness of

alternative management strategies for patients with solitary pulmonary nodules. Ann Intern Med 2003;

138(9):724-735.

(31) Juweid ME, Cheson BD. Positron-emission tomography and assessment of cancer therapy. N Engl J Med

2006; 354(5):496-507.








Dankwoord/Acknowledgements/Remerciements



119

Dankwoord/Acknowledgements/Remerciements

Promoveren doe je niet alleen en dit proefschrift zou niet tot stand gekomen zijn zonder de inspanning, medewerking en de steun van meerdere mensen.

Als eerste wil ik drs. Clemens Ticheler bedanken omdat hij mij destijds heeft voorgesteld aan Jaap Teule die toen professor aan de VU was. Beste Clemens, zonder geluk vaart niemand wel. Onze ontmoeting op het European Association of Nuclear Medicine (EANM) congres in Glasgow was voor mij zo’n gelukkig moment. Door jou ben ik voor het eerst op de hoogte gebracht van de praktijk van de nucleaire geneeskunde in Nederland. Je bent ook een goede vriend geworden en ik hoop dat wij regelmatig contact met elkaar zullen blijven houden en elkaar in de toekomst nog op meerdere congressen zullen ontmoeten.

Prof.dr. G.J. Teule, beste Jaap, ik wil je enorm bedanken voor de moeite die je genomen hebt om een Canadese afgestudeerde nucleaire geneeskundige op de afdeling aan te nemen en om haar de mogelijkheid te bieden promotieonderzoek te doen. Jouw visie op de toekomst van PET in Nederland was haarscherp. Ik heb veel bewondering voor je bescheidenheid, je oprechte interesse in je medewerkers en hun leven en je toewijding in het verbeteren van het leerproces van de nucleaire geneeskunde.

Prof.dr. O.S. Hoekstra, beste Otto, je hebt grote invloed gehad op dit proefschrift. Ik heb veel bewondering voor je originele en pragmatische manier van denken en analyseren alsmede ook je uitgebreide vakkennis. Ik ben je erg dankbaar voor het geduld en het begrip voor de ‘ups and downs’ van een promovendus, vooral in laatste fase toen ik weer terug was in Canada. Ondanks je drukke agenda als afdelingshoofd heb je mij toch geholpen en gestimuleerd. Meer kon ik van een promotor niet verwachten. Ik ben je uiteraard hiervoor zeer erkentelijk.

Dr.ir. A. van Lingen, beste Arthur, je bent het brein geweest achter het fantoom experiment. Je bent altijd bereid geweest mij te helpen, zowel met je technische hulp bij de LSO-PS camera, als ook het scannen van de patiënten, de fantoom proef, de computerproblemen, de natuurkundevragen en voor het vinden van de sponsors. Ik ben je zeer dankbaar voor alles.

Beste leden van de leescommissie en andere leden van de promotiecommissie: Dr. G.W. Sloof, Dr.ir. J.A.K. Blokland, Prof.dr. E.F. Smit, Prof.dr. C.R. Leemans, Prof.dr. J.W.R. Twisk, Prof.dr. R. de Bree en Prof.dr. G.J. Teule. Ik wil u allen hartelijk danken voor de tijd die u genomen heeft voor het bestuderen van mijn proefschrift en voor de door u gegeven kritiek hierop.

Dr. M. Lubberink, beste Mark, ik wil je bedanken voor het feit dat je mijn paranimf wil zijn en voor alle energie die je gestoken hebt in het fantoom experiment dat essentieel was voor dit proefschrift. Zonder jouw inspanningen zou het fantoom hoofdstuk niet tot stand gekomen zijn.

Dr. G.J.M. Herder, beste Judith, ik ben je dankbaar voor al je adviezen, met name met betrekking tot het afdrukken van dit proefschrift. Ik prijs me gelukkig jouw als collega te hebben getroff en:

Dankwoord


120

je bent altijd vriendelijk en bereid te helpen. Ik hoop dat het ons nog eens zal lukken samen een bergwandeling te maken.

Dr. P.G.H.M. Raijmakers, beste Pieter, het is altijd een plezier geweest om met je samen te werken. Ik heb vooral goede herinneringen aan onze discussies over het leven. Bedankt voor je begrip toen ik weer in Montreal zat.

Dr E.F.I. Comans, beste Emile, jij was altijd bereid de PET en CT scans te beoordelen. Ik heb veel over beeldinterpretatie van jou geleerd. Hiervoor heel erg bedankt.

Dr. R. Pijpers, beste Rik, ik wil je bedanken voor het beoordelen van meerdere PET scans. Ook heb ik veel over lymfoscintigrafi e van jou geleerd.

Dr. Ronald Boelaard en Dr. Hugo W.A.M. de Jong wil ik bedanken voor hun medewerking bij het fantoomexperiment.

Dr. C.J. Hoekstra, beste Corneline, het was leuk om met jou samen te werken toen ik stage liep op het PET centrum. Zonder jouw data zou een link tussen het fantoom-experiment en de klinische werklijkheid niet mogelijk zijn geweest. Hartelijk dank.

Drs. Ingrid Riphagen, Dr. Lotty Hooft en Dr. Sophie Mijnhout wil ik alle drie bedanken voor hun hulp bij het voorstellen van zoekstrategieën voor systematische reviews.

Beste voormalige assistenten van de afdeling nucleaire geneeskunde en PET op de VU: drs. Philip Baars, dr. Bart van Berckel, drs. Annelies van Schie, drs. Arjan van Dijk , drs. Livia Haslinghuis-Bajan en ook diegenen die ik hier met naam vergeet te noemen: ik wil jullie bedanken voor het vragen van toestemming aan de patiënten voor dit onderzoek. Het was leuk en leerzaam om met jullie samen te werken.

Ik wil de laboranten van het PET centrum, mw. S. van Balen, dhr. R. Koopmans en destijds dhr. B. Hoving, bedanken voor hun moeite om talloze patiëntenscans op het systeem terug te zetten en hun zorgzame aandacht voor de patiënten.

De administratieve medewerkers van het PET centrum: Mw. Amanda Kroonenberg-Kalwij, Mw. Cemile Karga en Dhr. Jaap van der Kuij wil ik bedanken. Beste Amanda, je hebt mij als buitenlander altijd goed advies geven. Ik ben jou en alle medewerkers dankbaar voor je administratieve ondersteuning, vooral toen ik weer in Canada zat.

Alle andere medewerkers van de afdeling nucleaire geneeskunde en PET research wil ik bedanken voor hun professionele en prettige samenwerking.

Dankwoord


121

Ik wil de longartsen bedanken die patiënten voor dit onderzoek naar ons verwezen hebben en de collega’s van de nabijgelegen ziekenhuizen die aan het fantoom-onderzoek hebben deelgenomen.

Ik wil tevens alle patiënten die hebben meegedaan aan het onderzoek bedanken.

J’aimerais aussi remercier tous mes anciens collègues au département de médecine nucléaire à l’Hôpital Lakeshore, notamment Raymond Lemieux et Diane Ste-Hilaire. Sans votre support et compréhension, il m’aurait pas été possible de fi nir cette thèse.

I would also like to thank my current collegues, Jeff rey Ross, Ken Ting, Walter Ammann and John Olekshy who supported me in in the fi nal phases of this endeavor.

Mijn paranimf drs. V. de Bot, beste Vera, jouw passie maakte de Nederlandse les elke keer weer bijzonder leuk. Ik heb altijd genoten van de Nederlandse romans die je mij hebt aanbevolen te lezen. Ondertussen ben je een heel goede vriendin geworden. Het was altijd gezellig bij jou te logeren als ik weer naar Amsterdam kwam om aan het proefschrift te werken.

Mijn schoonfamilie in Nederland (ainsi que ma belle soeur Isabelle) wil ik bedanken en met name mijn schoonvader Jan Timmers. Lieve Papa Jan, je bent mijn tweede Nederlandse docent naast Vera. Ik wil je bedanken voor het corrigeren van mijn eerste Nederlandse teksten alsmede ook de Nederlandse samenvatting van het proefschrift.

Euvella Codrington. Dear Euvella, you have been my second Mom and I appreciate your help with the title page.

Dr. Manish Joshi. Dear Manish, I’m so lucky that my brother is also one of my best friends and mentors. Your advice has always been helpful, especially with regards to the writing style in the introduction and summary.

My dear parents. Dear Mummy and Papa, you have always been there to support me and to encourage me with your unconditional love. I owe you so much.

Lieve Jories, zonder jouw steun en liefde zou dit proefschrift nooit tot stand gekomen zijn. Je bent mijn echte ‘soul-mate’. Dit proefschrift is hopelijk een van de vele uitdagingen die wij zullen overwinnen op onze gezamenlijk reis door het leven.

Dankwoord



List of abbreviations



125

AC attenuation correctionBGO bismuth germanateCI confi dence intervalCNR contrast to noise ratioCT computed tomographyDH dual headDHC dual head coincidenceDU diagnostic understandingF-18 fl uorine-18FDG fl uorodeoxyglucoseFN false negativeFORE Fourier rebinningFOV fi eld of viewFP false positiveFR full ringFWHM full width at half maximumGSO gadolinium oxyorthosilicateICC intraclass correlation coeffi cientIQR inter quartile rangeLN lymph nodeLSO lutetium oxyorthosilicateLUL left upper lobeLYSO lutetium-yttrium oxyorthosilicate MRI magnetic resonance imagingNAC nonattenuation correctedNaI sodium iodideNEC noise equivalent countNEMA National Electrical Manufacturers AssociationNSCLC non-small cell lung cancerOSEM ordered subsets expectation maximumPA pathologyPET positron emission tomographyPS positron emission tomography-single photon emission computed tomographyPN pulmonary noduleRAMLA row action maximum likelihood algorithmRLL right lower lobeRML right middle lobeROC receiver operating characteristicROI region of interestS/B sphere to backgroundSD standard deviationSPECT single photon emission computed tomographySPET single photon emission tomographySPN solitary pulmonary noduleTN true negativeTP true positiveUPT unknown primary tumour

List of abbreviations





Documents

Assessment of PET imaging devices: the case of a LSO/NaI PET-SPECT