Monte Carlo simulation of sensitivity and NECR of an entire-body PET scanner

Monte Carlo simulation of sensitivity and NECR of an entire-bodyPET scanner

Ismet Isnaini • Takashi Obi • Eiji Yoshida •

Taiga Yamaya

Received: 4 June 2013 / Revised: 10 December 2013 / Accepted: 12 December 2013

� Japanese Society of Radiological Technology and Japan Society of Medical Physics 2013

Abstract The current positron emission tomography

(PET) design is aimed toward establishing an entire-body

PET scanner. An entire-body PET scanner is a scanner

whose axial field of view (FOV) covers the whole body of

a patient, whereas whole-body PET scanner can be of any

axial FOV length, but was designed for a whole-body scan.

Despite its high production cost, an entire-body depth-of-

interaction PET scanner offers many benefits, such as

shorter and dynamic PET time acquisition, as well as

higher sensitivity and count rate performance. This PET

scanner may be cost-effective for clinical PET scanners

with high scan throughput. In this work, we evaluated the

sensitivity and count rate performance of a 2-m-long PET

scanner with conventional data acquisition (DAQ) archi-

tecture, using Monte Carlo simulation, and we evaluated

two ring diameters (60 and 80 cm) to reduce the scanner

cost. From simulation of scanning with a 2-m axial FOV,

the sensitivity for a 2-m-long PET scanner of 60 and 80-cm

diameter is around 80 and 68 times higher, respectively,

than that of the conventional PET scanner. In addition, for

the 2-m-long PET scanner with 60-cm diameter, the peak

noise equivalent count rate (NECR) was 843 kcps at

125 MBq, whereas the peak for the 80-cm diameter was

989 kcps at 200 MBq. This shows gains of 15.3 and 17.95,

respectively, in comparison with that of the conventional

PET scanner. The 2-m-long PET scanner with 60-cm ring

diameter could not only reduce the number of detectors by

21 %, but also had a 17 % higher sensitivity compared to

that with an 80-cm ring diameter. On the other hand,

despite the higher sensitivity, the NECR of the 60-cm ring

diameter was smaller than that of the 80-cm ring diameter.

This results from the single data loss due to dead time,

whereas grouping of axially stacked detectors was used in

the conventional DAQ architecture. Parallelization of the

DAQ architecture is therefore important for the 2-m-long

PET scanner to achieve its optimal performance.

Keywords Entire-body PET � Sensitivity � NECR

1 Introduction

Currently, clinical positron emission tomography (PET)

scanners have about 20-cm axial field of view (FOV).

These PET scanners can image the whole body using six or

more bed positions. On the other hand, an entire-body PET

scanner with an extended axial FOV, which can trace

whole-body uptake images at the same time and improve

the sensitivity dynamically, has been desired [1, 2]. We use

a different term entire-body scanner to distinguish between

entire-body and whole-body PET scanner. An entire-body

PET is a PET scanner whose axial FOV covers the whole

body of a patient, while whole-body PET can be of any

FOV length, but was designed for a whole-body scan. In

other words, an entire-body PET scanner can reduce the

scan time dramatically. Eriksson et al. [3, 4] have reported

the feasibility of a PET scan time of sub-minutes using an

extended axial FOV and time-of-flight (TOF) information.

This PET scanner may be cost-effective for clinical use

with a high scan throughput. However, PET detectors are

I. Isnaini � T. Obi

Imaging Science and Engineering Laboratory, Tokyo Institute of

Technology, 4259 Nagatsuta-cho, Midori-ku,

Yokohama 226-8503, Japan

e-mail: [email protected]

E. Yoshida (&) � T. Yamaya

National Institute of Radiological Sciences, 4-9-1 Anagawa,

Inage-ku, Chiba 263-8555, Japan

e-mail: [email protected]

Radiol Phys Technol

DOI 10.1007/s12194-013-0253-y

the main cause of the high production cost of PET scanner

even now. Some researchers have reported an entire-body

PET scanner with cheaper and thinner scintillation detec-

tors to reduce the PET scanner production cost [5], but

unfortunately the efficiency of such detectors is rather low.

In addition, an entire-body PET scanner would have to

process a large amount of data compared to conventional

PET scanners; hence, it would hamper the image con-

struction process.

The entire-body PET scanner also has many oblique

lines of response (LORs). Several characteristics of oblique

LORs, such as poor spatial resolution and decreased sen-

sitivity due to a smaller solid angle fraction and increased

scatter fraction and attenuation [6], should be discussed

carefully. Oblique LORs are able to maintain high spatial

resolution if a depth-of-interaction (DOI) detector is used

[7, 8].

In this work, we focused on an evaluation of the sen-

sitivity and count rate performance of entire-body PET

scanner without an imaging test. We evaluated the sensi-

tivity and count rate performance of a 2-m-long PET

scanner with conventional data acquisition architecture

(DAQ) using Monte Carlo simulation, and we evaluated

two types of ring diameter for reducing the scanner cost. In

addition, we discuss some issues to be resolved for the

feasibility of an entire-body PET scanner. Our simulation

model implements the 4-layered depth-of-interaction (DOI)

PET scanner, but DOI information was not used in this

simulation.

2 Materials and methods

2.1 Simulation setup

The simulations were conducted by use of the GEANT4

application for tomographic emission (GATE) version 6.2,

installed on a 64-bit Linux personal computer (Intel Xeon

X3330 2.66 GHz). A back-to-back photon source was also

used in the simulation to improve the computational effi-

ciency. The time taken for each simulation depends on the

source activity being used and the total time set in the

macro file of GATE programming. For example, a simu-

lation for a source of 400 MBq with a simulation time of

20 ms would last for 2–3 h. The software has been widely

used for simulation on PET scanners [9–11]. Six types of

4-layered DOI–PET scanners were simulated with three

axial FOVs and two ring diameters, as shown in Fig. 1.

A PET scanner with an 80-cm ring diameter and 19.2-cm

axial FOV length will be referred as the conventional PET

scanner in the rest of this paper. PET scanners with 100.8

and 201.6-cm axial FOVs are the extended version with 21

and 42 detector rings, respectively. Any scanner with 2-m

axial FOV will be referred as an entire-body PET scanner

in this paper. Scanners with 60 and 80-cm ring diameter

consisted of 38 and 48 detectors per detector ring. The

detector used is made of lutetium oxyorthosilicate (LSO)

crystals, which exhibit a reasonably fast scintillation decay

time for all PET scintillators currently in use. Each crystal

block was made of a 16 9 16 scintillation crystal with an

Fig. 1 Illustrations of six types

of PET scanners and DOI

detector. Ring diameter D is 60

and 80 cm

I. Isnaini et al.

area dimension of 2.9 9 2.9 mm and a thickness of 5 mm,

arranged in 4 layers. Table 1 shows the specifications of

the six types of entire-body PET scanners. During the

simulation, a paralysable dead time (250 ns) was applied

on the single data for each detector block. Conventionally,

detector blocks are arranged axially into a grouping circuit

with a non-paralysable dead time before judging of coin-

cidence events. However, the single data collision into the

grouping circuit is increased for the entire-body PET

scanner. Single data are applied with non-paralysable dead

time (256 ns) for grouping of axially stacked detectors.

2.2 Sensitivity test

A 175-cm line source of 50 kBq was used in the simula-

tion, and the relative sensitivity for the six types of PET

scanners was measured. The source was set in the center of

the FOV. The number of detected true coincidences was

computed, and system sensitivity profiles with a single-bed

position were calculated, with each measurement time

being 1200 s. Another test is the sensitivity test with

multiple-bed positions, which is calculated with the use of

the same line source. The sensitivity is defined as the total

of true counts for each slice, as in the following formula:

Sensitivity profile with a single bed positioni

¼XSlice

j¼�Slice

Truei;j ð1Þ

Total sensitivity

¼XNo:ofbed

i¼1

Sensitivity profile with single bed position i;

ð2Þ

where i defines the index of number of bed required for

multi-bed acquisition, while slice is the distance of a par-

ticular slice from the center in cm.

Several bed positions were required for a shorter axial

FOV scanner, with different scanning times per bed, as

shown in Table 2. The total axial FOVs and measurement

times for different number of bed positions were 201.8 cm

and 1200 s, respectively. This total scanning time includes

the time taken to shift the bed from one position to another.

For example, in the case of 21 rings scanner, 396 s is the

individual scanning time and the rest of the time is used to

shift the bed twice, about 50 cm away to the next position,

with the assumption that the bed speed is 10 cm/s, the

shifting time is about 5 s. Likewise is for 4 rings scanner.

However, the real calculation time during the simulation

was about 12 h for the 42 rings scanner.

2.3 NECR test

For the noise equivalent count rate (NECR) test, we uti-

lized a solid polyethylene cylinder phantom (175 cm long

and 20 cm in diameter) with a 175-cm line source. This

phantom was a modified version of the NEMA NU-2 2001

standard phantom [12]. The phantom was placed in the

center of the FOV. The NECR was calculated as follows:

NECR ¼ T2

T þ Sþ 2R; ð3Þ

where, T, S, and R are the true, scatter, and random count

rates, respectively. By simply retrieving the true, scatter,

and random counts for different source activities, an NECR

graph was plotted and the peak NECRs was noted. The real

calculation time for a complete NECR graph was about

12 h for the 42 rings. The method of acquiring the scatter,

random, and true count is by counting the number of

occurrences for each event.

3 Results

3.1 Sensitivity

Figure 2 shows sensitivity profiles for six types of PET

scanners for single-bed-position scanning with full maximum

ring difference (MRD), namely, 1343 MRD for 42 rings, 671

for 21 rings, and 127 for 4 rings. As the axial FOV is extended,

the sensitivity profile, instead of being triangular, becomes

dome-shaped. Table 3 shows the total sensitivities of the six

Table 1 Specifications of the DOI–PET scanners

Crystal size 2.9 9 2.9 9 5 mm3

Crystal material LSO

Number of crystals 16 9 16 9 4

(per detector)

Axial FOV 19.2 cm (4 detectors), 100.8 cm (21 detectors),

201.6 cm (42 detectors)

Ring diameter 60 cm (38 detectors), 80 cm (48 detectors)

Timing resolution 1.4 ns

Coincidence time

window

6 ns

Energy resolution 15 %

Energy window 450–600 keV

Table 2 Setup for multiple-bed-position studies

Number

of rings

Scanner

axial

FOV

(cm)

Number

of bed

positions

Scan

time per

bed (s)

Shifting

time (s)

Position

overlapping

(%)

42 201.8 1 1200 0 0

21 100.6 3 396.67 5.0 58

4 19.2 19 62.210 1.0 52

Monte Carlo simulation of sensitivity and NECR

types of PET scanners for single-bed-position scanning. At the

center of the axial FOV, the sensitivity for a 2-m-long PET

scanner of 60-cm diameter was about 5.33 times higher than

for the conventional PET scanner. This means an increase of

approximately 17 % for the same length of scanner (2 m) with

a larger diameter (80 cm).

Figure 3 shows the sensitivity profile of the scanners for

multiple-bed-position scanning. The black line indicates

the sensitivity for an individual bed position, whereas the

gray line shows the total sensitivity for whole-body

Fig. 2 Sensitivity profiles for single-bed-position scanning

Table 3 Total sensitivities of six types of PET scanners for single-

bed-position scanning and their gain (in parentheses) compared with

the conventional PET scanner

Number of rings Sensitivity (kcps/MBq)

60-cm diameter 80-cm diameter

4 95.04 (1.38) 68.79 (1)

21 314.91 (4.58) 261.49 (3.8)

42 366.94 (5.33) 328.13 (4.77)

I. Isnaini et al.

scanning. The total sensitivity for a smaller scanner (60 cm

diameter) is almost 80 times higher than that for the con-

ventional PET scanner, as shown in Table 4.

3.2 NECR

Figure 4 shows the complete NECR graph as well as ran-

dom, true, and scatter count rates for an entire-body PET of

80-cm diameter. As the random count increases especially

at higher activities, the value of the NECR decreases. This

is due to the fact that other count such as true and scatter

count does not have a significant increase on higher

activities.

Figure 5 and Table 5 show NECR curves and peak

NECRs for the six types of PET scanners within 400 MBq.

In the case of 4 rings, the NECR maximum value was not

Fig. 3 Sensitivity profiles for multiple-bed-position scanning


achieved below 400 MBq. For the 2-m-long PET scanner

with 60-cm diameter, the peak NECR was 843 kcps at

125 MBq, whereas the peak for the 80-cm diameter was

989 kcps at 200 MBq. This shows gains of 15.3 and 17.95,

respectively, in comparison to conventional PET.

Table 6 shows NECR gain at 125 MBq for a clinical

situation (the injection activity was 200 MBq, and the

resting time was 60 min.). This gain is even higher than the

gain achieved at the peak NECR.

4 Discussion and conclusion

The 2-m-long PET scanner with the conventional DAQ

architecture promises a significant increase in both NECR

and sensitivity compared with the conventional PET

scanner. In addition, the 2-m-long PET scanner with 60-cm

ring diameter not only could reduce the number of detec-

tors by 21 %, but also had 17 % higher sensitivity com-

pared to that with an 80-cm ring diameter. On the other

hand, despite the higher sensitivity, the NECR of the 60-cm

ring diameter was smaller than that of the 80-cm ring

diameter. This results from the single data loss due to dead

time applied in the grouping of axially stacked detectors

with the conventional DAQ architecture.

In the 2-m-long PET scanner with the conventional

DAQ architecture, the single data throughput was saturated

in limited activity. As a result, single data loss at the

Table 4 Total sensitivities of six types of PET scanners for multiple-

bed-position scanning and their gain (in parentheses) compared with

the conventional PET scanner

Number of rings Sensitivity (kcps/MBq)


4 5.67 (1.48) 3.84 (1)

21 122.67 (31.98) 89.0 (23.21)

42 306.73 (79.97) 260.8 (68.00)

Fig. 4 NECR graph for entire-body PET with 80-cm diameter for 42

rings scannerFig. 5 NECR curves for six types of PET scanners. 80-cm (top) and

60-cm ring diameter (bottom)

Table 5 Peak NECR for six types of PET scanners and its gain (in

parentheses) compared with the conventional PET scanner

Number of

rings

Activity at peak NECR

(MBq)

Peak NECR (kcps)

60-cm

diameter

80-cm

diameter

60-cm

diameter

80-cm

diameter

4 400 400 69.4 (1.25) 55.1 (1)

21 175 275 511.9 (9.29) 546.1 (9.9)

42 125 225 843.9 (15.31) 989.1 (17.95)

Table 6 NECR gain for different PET scanners at 125 MBq in

comparison with conventional PET scanner

Number of rings NECR gain


42 26.41 29.34

21 14.90 14.73

4 1.32 1.00

I. Isnaini et al.

grouping circuits was limited to the peak NECR for the

entire-body PET scanner. To resolve this problem, one

needs to separate the grouping processes in the axial

direction [13]. However, coincidence detection circuits

become complex.

By saving 20–30 min of scanning time per patient for a

PET/CT; for example, the number of patients has increased

up to 40 % [14]. Hence, clinical centers can have more

patients examined daily, which means a faster return on the

value of such equipment. If we assume that the scan time of

a conventional PET scanner is 20 min, the scan time of the

2-m-long PET scanner with 60-cm ring diameter, achieving

the same sensitivity level, can be reduced to 15 s by simple

calculation. This PET scanner may be cost-effective for

clinical PET scanners with high scan throughput, but the

feasibility of entire-body PET scanners remains with many

agendas.

As we can see, the time savings based on the sensitivity

gain, we can also predict a great time saving by looking at

the gain in the NECR performance. As seen in Fig. 5,

entire-body PET has an NECR peak much higher than the

21 and 4 rings scanner for the same activity per volume.

Hence, we can either reduce the scanning time to achieve

the same value of the NECR or lower the dose for each

patient.

The axial sensitivity profile of the conventional PET

scanner had a triangular distribution. Conventionally, the

sensitivity of a PET scanner depends on the detection

efficiency and the acceptance angle. As the axial FOV was

extended and the ring diameter was narrowed, the increase

of the acceptance angle was saturated; thus the sensitivity

profile became dome-shaped distribution. Therefore, at the

FOV center, the saturation of the acceptance angle has its

maximum effect. In addition, oblique LORs sacrifice the

maximum sensitivity by the choice of maximum ring dif-

ference (MRD), to obtain a uniform sensitivity profile for

the entire-body PET scanner. On the other hand, oblique

LORs cause a parallax error. The use of a DOI detector can

reduce the parallax error; however, the introduction of this

type of detector has increased the oblique LORs by a great

amount. Nevertheless, these extremely many oblique LORs

may not contribute to the imaging performance. We will

try to evaluate the image quality for the entire-body PET

scanner.

The entire-body PET scanner would have to process a

large amount of data compared to conventional PET

scanners. At the activity of the peak NECR, the single

count rate was 53.02 Mcps; this value is 6.3 times that of

the conventional PET scanner. Also, the true and random

count rate was 7.69 Mcps; this value is 26.94 times that of

the conventional PET scanner. Therefore, coincidence

detection circuits need more parallel and high-throughput

performance. In addition, the total number of LORs with

the 2-m-long PET scanner with 80-cm ring diameter was

above 130.4E9. This was 110.25 times that of the con-

ventional PET scanner. On the other hand, the entire-body

PET scanner with 60-cm ring diameter can reduce this

value by 0.62 times.

Cost is the most important problem. The production cost

of PET scanners mainly depends on the amount of the

scintillator and photomultiplier tubes (PMT). The number

of detectors used for the entire-body PET scanner with

60-cm ring diameter was 8.3 times higher, in comparison to

that of the conventional PET scanner. The gain of scan

time of the entire whole-body PET scanner was 80 times in

comparison to that of the conventional PET scanner, by

simple calculation. We think that the entire-body PET

scanner has the possibility of cost savings in comparison to

conventional PET scanners. For reducing the detector cost,

we will try to develop a low cost DOI detector using

conventional PMTs.

Another alternative for overcoming the cost issue is the

Open PET scanner [15–17]. Its design introduces a gap

between the detector ring to reduce the number of crystals

involved, but at the same time it compromises the sensi-

tivity value, NECR, cost, and the dose of radionuclide

injected into the patient. The number of crystals used for

the same scanner length varies according to the design.

Conflict of interest The authors declare that they have no conflict

of interest.

References

1. Karakatsanis NA, et al. Dynamic multi-bed FDG PET imaging:

feasibility and optimization. In: Proceedings of the IEEE NSS-

MIC, 2011. p. 3863–70.

2. Couceiro M, Ferreirac NC, Fonte P. Sensitivity assessment of

wide axial field of view PET systems via Monte Carlo simula-

tions of NEMA-like measurements. Nucl Instrum Methods Phys.

2007;A580:485–8.

3. Eriksson L, et al. An investigation of sensitivity limits in PET

scanners. Nucl Instrum Methods Phys. 2007;A 580:836–42.

4. Eriksson L, et al. Towards sub-minute PET examination. IEEE

Trans Nucl Sci. 2011;58(1):76–81.

5. Crespo P, et al. Whole-body single-bed time-of-flight RPC-PET:

simulation of axial and planar sensitivities with NEMA and

anthropomorphic phantoms. In: Nuclear Science Symposium

Conference Record M09-353, 2009. p. 3420–5.

6. Saha GB. Performance characteristic of PET scanner. In: Basic of

PET imaging: physics, chemistry and regulations. New York:

Springer; 2010. p. 97–116.

7. Murayama H, Ishibashi H, Uchida H, Omura T, Yamashita T.

Depth encoding multicrystal detectors for PET. IEEE Trans Nucl

Sci. 1998;45(3):1152–7.

8. Tsuda T, et al. Performance evaluation of a subset of a four-layer

LSO detector for a small animal DOI PET scanner: jPET-RD.

IEEE Trans Nucl Sci. 2006;53(1):35–9.

9. Buvat I, Lazaro D. Monte Carlo simulations in emission

tomography and GATE: an overview. Nucl Instrum Methods

Phys Res A. 2006;A569:323–9.


10. Jan, et al. GATE: a simulation toolkit for PET and SPECT. Phys

Med Biol. 2004;49:4543–61.

11. Assie K, et al. Monte Carlo simulation in PET and SPECT

instrumentation using GATE. Nucl Instrum Methods Phys Res A.

2004;A527:180–9.

12. Performance measurements of positron emission tomographs,

NEMA standards Pub. NU 2–2007, National Electrical Manu-

facturers Association, Rosslyn, 2007.

13. Yoshida E, Yamaya T, Nishikido F, Inadama N, Murayama H.

Basic study of entire whole-body PET scanners based on the

OpenPET geometry. Nucl Instrum Methods Phys Res. 2010;A

621:576–80.

14. Muhammad WS, Ifigenia T, Nektaria M, Kostas S. Role and cost

effectiveness of PET/CT in management of patients with cancer.

Yale J Bio Med. 2010;83:53–65.

15. Yamaya T, Inaniwa T, Minohara S, Yoshida E, Inadama N,

Nishikido F, Shibuya K, Lam CF, Murayama H. A proposal of an

open PET geometry. Phys Med Biol. 2008;53(3):757–77.

16. Yoshida E, Yamaya T, Shibuya K, Nishikido F, Inadama N,

Murayama H. Simulation study on sensitivity and count rate

characteristics of ‘‘OpenPET’’ geometries. IEEE Trans Nucl Sci.

2010;57(1):111–6.

17. Yoshida E, Yamaya T, Nishikido F, Inadama N and Murayama

H. Feasibility study of entire whole-body PET scanners based on

the OpenPET geometry. In: Proceedings of the IEEE Nuclear

Science Symposium Conference Record, 2009. p. 3628–9.

I. Isnaini et al.

Documents

Monte Carlo simulation of sensitivity and NECR of an entire-body PET scanner