Upload
taiga
View
212
Download
0
Embed Size (px)
Citation preview
Reduction method for intrinsic random coincidence eventsfrom 176Lu in low activity PET imaging
Eiji Yoshida • Hideaki Tashima • Fumihiko Nishikido •
Hideo Murayama • Taiga Yamaya
Received: 11 October 2013 / Revised: 17 January 2014 / Accepted: 18 January 2014
� Japanese Society of Radiological Technology and Japan Society of Medical Physics 2014
Abstract For clinical studies, the effects of the intrinsic
radioactivity of lutetium-based scintillators such as LSO
used in PET imaging can be ignored within a narrow
energy window. However, the intrinsic radioactivity
becomes problematic when used in low-count-rate situa-
tions such as gene expression imaging or in-beam PET
imaging. Time-of-flight (TOF) measurement capability
promises not only to improve PET image quality, but also
to reduce intrinsic random coincidences. On the other hand,
we have developed a new reduction method for intrinsic
random coincidence events based on multiple-coincidence
information. Without the energy window, an intrinsic
random coincidence is detected simultaneously with an
intrinsic true coincidence as a multiple coincidence. The
multiple-coincidence events can serve as a guide to iden-
tification of the intrinsic coincidences. After rejection of
multiple-coincidence events detected with a wide energy
window, data obtained included a few intrinsic random and
many intrinsic true coincidence events. We analyzed the
effect of intrinsic radioactivity and used Monte Carlo
simulation to test both the TOF-based method and the
developed multiple-coincidence-based (MC-based) method
for a whole-body LSO-PET scanner. Using the TOF- and
MC-based reduction methods separately, we could reduce
the intrinsic random coincidence rates by 77 and 30 %,
respectively. Also, the intrinsic random coincidence rate
could be reduced by 84 % when the TOF?MC reduction
methods were applied. The developed MC-based method
showed reduced number of the intrinsic random
coincidence events, but the reduction performance was
limited compared to that of the TOF-based reduction
method.
Keywords PET � Simulation
1 Introduction
Lutetium-based scintillators, such as lutetium oxyorthosi-
licate (LSO) [1], are widely used for positron emission
tomography (PET) scanners. These scintillators contain a
small amount of 176Lu intrinsic radioactivity. As shown in
Fig. 1a, 176Lu decays into 176Hf with simultaneous b-par-
ticle emission (average energy 420 keV) and three prompt
c-ray emissions (88, 202, and 307 keV). In a PET detector,
the b particle emitted from the LSO array tends to be
detected in the same LSO array. In contrast, c-rays are
detected not only in the same LSO array, but also in
another LSO array by emission outside the first LSO array.
In this scenario, the b particle and the c-rays can be rec-
ognized as an intrinsic true coincidence, as shown in
Fig. 1b. Intrinsic true coincidence detects two radiations
from one 176Lu nuclide, simultaneously. When two bparticles are detected within the coincidence time window,
these b particles can be recognized as an intrinsic random
coincidence, as shown in Fig. 1b. Intrinsic random coin-
cidence detects two radiations from different 176Lu nuc-
lides, accidentally. In this case, the intrinsic coincidences
are detected as a multiple coincidence, because the intrinsic
random coincidence is detected simultaneously with the
intrinsic true coincidences. A multiple coincidence occurs
when more than two photons are detected in different
detectors within the coincidence time window. Naturally,
both intrinsic true and random coincidences are false
E. Yoshida (&) � H. Tashima � F. Nishikido � H. Murayama �T. Yamaya
Molecular Imaging Center, National Institute of Radiological
Sciences, 4-9-1 Inage-ku, Chiba 263-8555, Japan
e-mail: [email protected]
Radiol Phys Technol
DOI 10.1007/s12194-014-0258-1
coincidences. With a narrow energy window, most intrinsic
coincidences can be prevented. Thus, for clinical studies,
the count rate of the intrinsic radioactivity can be ignored
with a narrow energy window and a narrow coincidence
time window [2–4].
On the other hand, when imaging is done with a low
activity level, a PET scanner detects only a limited number
of true coincidences. In a small-animal PET scanner, for
example, for cell-trafficking studies [5] or gene expression
imaging [6], the injection activity tends to be limited to a
very low level (below 1 kBq for some situations). In these
cases, the intrinsic radioactivity affects the imaging per-
formance with the conventional energy window [7].
Another example is in-beam PET, which is a method for
the in situ monitoring of charged particles therapy [8, 9].
For the usual carbon (12C) beam irradiation, it has been
reported that the activity of positron emitters (e.g., 9C, 10C,11C, 15O, etc.) produced through fragmentation reactions
between the projectiles and the atomic nuclei of the tissue
is generally low. Instead of the 12C beam, therefore, we
aimed at using a 11C radioactive beam as an incident beam
directly. Because the projectiles themselves are positron
emitters, we expected that PET images could be obtained
directly, corresponding to the distribution of the primary
particles. Also, 11C has shown tenfold magnification of the
activity concentration compared to 12C. For in-beam PET
studies, we have developed a small prototype PET scanner
[10, 11] that uses a lutetium gadolinium oxyorthosilicate
scintillator. This small prototype PET scanner had 6.5 %
system sensitivity, 14 % energy resolution (energy window
of 400–600 keV), and 3.0 ns timing resolution (coinci-
dence time window of 20 ns). For 11C irradiation with
70 Gy, the average true coincidence rate was about 2500
cps. This means that the true coincidence rate is expected
to be limited to 35.7 cps at a typical clinical irradiation
dose (1 Gy). Calculated simply from our experiment, the
assumed minimum average activity was about 500 Bq. On
the other hand, the intrinsic coincidence rate of the small
prototype PET scanner has been counted as about 50 cps,
and we expected that the intrinsic coincidence rate of a
human-size PET scanner would be higher, because human-
size PET scanners have about 10 times greater scintillator
volumes than the small prototype PET scanner. It is rec-
ommended that all possible intrinsic coincidences be pre-
vented when a scintillator like LSO is used. The simplest
solution to this is to use scintillators without intrinsic
radioactivity, such as bismuth germinate or gadolinium
oxyorthosilicate; however, LSO offers a better perfor-
mance than do these scintillators, because it has a suffi-
ciently good time-of-flight (TOF) measurement capability
to give improved imaging performance.
Recently, some new TOF-PET scanners have been
developed [12–14]. The time difference between the
detection of two annihilation photons is used so that the
uncertainty of the annihilation positions is reduced. TOF
information decreases the statistical noise values of
reconstructed images. Conti [15] reported that the TOF
measurement capability promised not only to improve the
PET image quality, but also to reduce the number of ran-
dom coincidences. We thought that the LSO-PET scanner
might be able to image low activity levels using TOF
information, although LSO has intrinsic radioactivity.
Thus, we used TOF information to reduce the number of
intrinsic random coincidences.
On the other hand, when coincidence events are not
limited by the energy window, many intrinsic coincidences
are detected as multiple coincidences. At a low activity
level, the multiple-coincidence events are a unique phe-
nomenon for identifying intrinsic coincidences. In this
work, we developed a new reduction method for the
intrinsic random coincidences from 176Lu, based on mul-
tiple-coincidence information. Also, we evaluated the
effects of intrinsic random coincidence reduction by the
narrow energy window and TOF- and multiple coinci-
dence-based (MC-based) reduction methods using Monte
Carlo simulation.
Fig. 1 Illustrations of a the
decay level scheme of 176Lu and
b intrinsic coincidence from176Lu. Intrinsic true coincidence
is caused by b–c coincidence,
and intrinsic random
coincidence is caused by b–bcoincidence
E. Yoshida et al.
2 Materials and methods
2.1 Energy window
In imaging at a low activity level, the energy window must
be set as narrow as possible so that contamination from
intrinsic coincidences is prevented. Especially, in intrinsic
random coincidences, b particles have a broad energy
distribution. Typically, in the newest whole-body LSO-
PET scanners, the lower-level discriminator (LLD) and the
upper-level discriminator (ULD) are set at 425 and
650 keV, respectively. For reducing the number of intrinsic
random coincidences, the LLD should be set at a higher
energy level and the ULD should be set at a lower energy
level.
2.2 TOF-based reduction method
A fast scintillator like LSO has the advantage of being able
to reduce the number of random coincidences using a
narrow coincidence time window. But, for a clinical PET
scanner, the minimum coincidence time window is limited
by the ring diameter to about 4 ns. On the other hand, by
use of TOF information, the coincidence events detected
outside an object can be identified as random coincidence
events. Therefore, TOF information has random coinci-
dence reduction capability beyond the hardware coinci-
dence timing window [15]. The intrinsic random
coincidences can be reduced in the same way, as shown in
Fig. 2. The reduction performance of intrinsic random
coincidences within the object is defined as follows:
TOF reduction rate ¼ Dring � Dobject
Dring
; ð1Þ
where Dring is the ring diameter and Dobject is the object
diameter. Therefore, the reduction performance of the
TOF-based reduction method depends on the size of the
object. Because the coincidence events detected outside an
object can be identified as random coincidence events. In
particular, we think that the TOF-based reduction method
is suitable for in-beam PET, because the activity distribu-
tion is confined to a localized area in the body.
2.3 MC-based reduction method
An intrinsic true coincidence is a rare event for a narrow
energy window [7], because the maximum energy of the c-
rays is 307 keV. On the other hand, intrinsic random
coincidences cannot be prevented completely by the nar-
row energy window. If the energy window is set wide, both
intrinsic random coincidences and intrinsic true coinci-
dences are detected as multiple coincidences. The multiple-
coincidence events can serve as a guide to the identification
of the intrinsic coincidences at the low activity level.
Figure 3 shows the procedure for the MC-based reduction
method. In the first step, coincidence events within the
wide energy window are measured. After rejection of the
multiple-coincidence events from all coincidence events,
the remaining coincidence events are discriminated within
the narrow energy window again to reduce the number of
intrinsic true coincidences. With the wide energy window,
the number of coincidence events is increased drastically.
Fig. 2 Illustration of TOF
measurement of the random
coincidence event. Events
detected for TOF bins outside
the object can be identified as
random coincidence events. It
should be noted that the random
coincidence rates of each TOF
bin are equal
Reduction method for intrinsic random coincidence events
However, at a low activity level, the PET scanner has
sufficient detectability with the wide energy window.
The MC-based reduction method can be implemented if
the list-mode data include energy information. Acquired
list-mode data reject or allow all multiple coincidences,
depending on the data acquisition system of the PET
scanners. In the data acquisition, the energy window is set
wide. After the data acquisition, acquired list-mode data
reject all multiple coincidences, and the conventional
energy window is applied.
2.4 Simulation setup
We analyzed the coincidence count rate of the 176Lu
intrinsic radioactivity and tested both methods for the
whole-body LSO-PET scanner using the Geant4 Appli-
cation for Tomographic Emission (GATE) [16, 17]
simulation. GATE does not support treatment of the
intrinsic radioactivity. However, the intrinsic radioactiv-
ity in the LSO has been simulated successfully as a176Lu ion source with an activity of 276.75 Bq/cm3 [7].
McIntosh et al. [18] reported that simulation data from
GATE agreed with experimental data for the Siemens
Inveon PET detector modules. We evaluated validation
of the GATE simulation that compared a model of 176Lu
intrinsic radioactivity to measured data for our small
prototype PET scanner [10, 11].
As shown in Fig. 4, a simulated whole-body LSO-PET
scanner was designed with four detector rings of 48
detector blocks. Each detector consisted of a 16 9 16 array
of 2.9 9 2.9 9 20.0 mm3 LSO crystals. Table 1 lists the
basic specifications of the scanner. The intrinsic count rate
was calculated by recording of the number of prompt and
delayed coincidences. However, a subtraction-based ran-
dom correction may not give statistically meaningful
events for a random count rate estimated from the delayed
coincidences, which have a sparse distribution for a limited
measurement time. On the other hand, if emission data do
not depend on intrinsic random coincidences at a very low
activity level, we think that intrinsic random coincidences
can be measured over a long time before the emission data
Fig. 3 Procedure for the MC-
based reduction method. For
conventional PET scanners,
coincidence events are limited
by the narrow energy window.
On the other hand, for the MC-
based reduction method, in the
first step, coincidence events
within the wide energy window
are measured. After rejection of
the multiple-coincidence events
from all coincidence events, the
remaining coincidence events
are discriminated within the
narrow energy window again to
reduce the number of intrinsic
true coincidences
Fig. 4 Schematic geometry of the whole-body LSO-PET scanner
E. Yoshida et al.
measurement. However, the calculation time is just not
realistic for obtaining statistically meaningful events.
Therefore, in this study, we used only prompt coincidences.
The intrinsic true and random coincidences were cal-
culated with several energy windows. Also, for evaluation
of the TOF- and MC-based reduction methods, the imaging
target was a set of cylindrical polyethylene phantoms,
without activity, with diameters of 10 and 20 cm (both
were 20 cm long). In this case, the lines of response
(LORs) outside the imaging target were rejected before the
reduction process because these LORs cannot contribute to
the imaging quality. For the MC-based reduction method,
the energy window was set at 100–1000 keV. Also, we
evaluated the reduction performance of the TOF- and MC-
based reduction methods for several scanner configurations
(varying timing resolution, axial field of view (FOV), and
crystal thickness). The reduction performance was calcu-
lated as follows:
Reduction rate ¼ Ro � Rp
Ro
; ð2Þ
where Ro and Rp are the random count rate without the
reduction method and the random count rate of the devel-
oped method, respectively. Each measurement time was
10 s. (The calculation time was about 10 h with a 2.8 GHz
Xeon Processor.)
To evaluate the effective count rate for practical use, we
calculated the noise equivalent count rate (NECR) of this
PET scanner as follows:
NECR ¼ T2
T þ Sþ k � R ; ð3Þ
where T, S, and R are the true, scatter, and random count
rates, respectively. The factor k is the fraction of the pro-
jection occupied by the cylindrical phantom. Because the
random count rate was determined directly from the sim-
ulation data, there was no factor of 2 in front of R, as would
be the case from error propagation considerations when a
delayed coincidence window is used. The NECR test uti-
lized a solid polyethylene cylinder phantom (70 cm long
and 20 cm in diameter) with a 70-cm line source at 4 cm
off-axis. The phantom was based on the NEMA NU-2 2001
standard [19]. The phantom was placed in the center of the
FOV, and the acquisition time was 10 s.
2.5 Imaging simulation
Images of intrinsic random coincidence events were gen-
erated for several FOVs by use of random coincidence
reduction methods. The measurement time was 600 s. (The
calculation time was about 20 days with the 2.8 GHz Xeon
Processor.) The 2D ML-EM algorithm with a matrix size of
200 9 200 9 48 was applied to the data rebinned by the
single-slice rebinning method [20]. The voxel size was
4 9 4 9 4 mm3. An imaging quality test was calculated
by uses of four spherical sources with 3-cm diameter.
Source activities were 100, 250, 500, and 1000 Bq. The
FOV was limited to 20 cm in diameter. Normalization was
applied, but a scatter correction and attenuation correction
were not applied to the data because we used the ideal
phantom without the scatter and attenuation defined only
for activity. Also, no random correction was applied to the
data because the sinogram of delayed coincidences had too
few counts. The contrast-to-noise ratio (CNR) was calcu-
lated for each source as follows:
CNR ¼ SROI � Sbackgroundffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
r2ROI þ r2
background
� �
=2
r ð4Þ
where SROI and rROI are the mean signal and standard
deviation of each weak source, and Sbackground and
rbackground are the mean signal and standard deviation of
the background data, respectively.
3 Results
3.1 Evaluation of intrinsic coincidence events
For validation of the GATE simulation, Fig. 5 shows
simulated and measured energy spectra from 176Lu intrin-
sic radioactivity for the small prototype PET scanner. The
simulated energy spectrum was obtained with the
0–1000 keV energy window. There was a good agreement
between the results with the simulated model and the
measurements. This energy spectrum showed photopeaks
of 88, 202, and 307 keV. Most of the 88-keV c-rays
seemed to be detected by the detector that detected the bparticles. The peak at 650 keV was the sum of the peaks
for the three c-rays and the b particles.
Figure 6 shows the simulated energy spectrum with the
intrinsic coincidence events from the 176Lu intrinsic radio-
activity and 511 keV annihilation photons (back-to-back
Table 1 Basic specifications of the whole-body LSO-PET scanner
Crystal material LSO
Crystal size 2.9 9 2.9 9 20 mm3
Number of crystals 16 9 16 (per detector)
Number of rings 64
Ring diameter 80 cm
Axial FOV 19.2 cm
Time resolution 500 ps
Energy resolution 15 %
Coincidence time window 4.5 ns
Energy window 425–650 keV
Reduction method for intrinsic random coincidence events
annihilation photons) for the whole-body LSO-PET scanner.
The sum peak at 650 keV as shown in Fig. 5 was rejected by
coincidence measurement. The b particles had a very broad
distribution. Also, this energy spectrum showed three peaks
that corresponded to the photopeaks of 88, 202, and 307 keV.
For coincidence measurement, most events of the summed
peak at 650 keV were rejected. Also, the 425 keV LLD was
suitable, but the 650 keV ULD was too large for the 15 %
energy resolution.
Figure 7a and b show the intrinsic true and random
count rates for the LLD and the ULD, respectively. The
ULD of Fig. 7a and the LLD of Fig. 7b were 600 and
425 keV, respectively. The intrinsic true coincidence rate
tended to depend on only the LLD, but the intrinsic random
coincidence rate depended on both LLD and ULD. The
intrinsic true coincidence could be ignored with the
375 keV LLD. Our prototype system’s energy window of
400–600 keV completely included the photopeak.
However, it is recommended that all possible intrinsic
coincidences be prevented when LSO is used. Therefore,
the energy window was selected as 425–575 keV by cut-
ting of 25 keV from both the LLD and ULD. The count
rate of the photopeak with the energy window of
425–575 keV decreased by only 3 % against that with the
energy window of 400–600 keV. In this energy window,
the 511 keV annihilation photons could be detected ade-
quately, and the intrinsic random coincidence rate was 226
cps.
3.2 Evaluation of TOF- and MC-based reduction
methods
Figure 8 shows the intrinsic random count rates for the
TOF- and MC-based reduction methods for the 10- and
20-cm-diameter cylinder phantoms without activity. The
coincidence timing window was 4.5 ns. The intrinsic ran-
dom coincidences were counted only as LORs through the
phantom. After selection of the narrow energy window for
the 15 % energy resolution, the intrinsic random count rate
with the energy window of 425–575 keV was 56 cps for
the 20-cm-diameter cylinder phantom. By using the TOF-
and MC-based reduction methods separately, we could
reduce the intrinsic random count rate by 77 and 30 %,
respectively. Also, the intrinsic random count rate could be
reduced by 84 % when the TOF?MC reduction method
was applied. Figure 9 shows reduction rates of the TOF-
and MC-based reduction methods, separately and com-
bined, as a function of the diameter of the cylinder phan-
tom for several timing resolutions. Figure 10 shows
reduction rates of TOF- and MC-based reduction methods,
separately and combined, for the 20-cm-diameter cylinder
phantom as a function of the number of detector rings for
several crystal thicknesses. The random coincidence
reduction performance of the TOF-based reduction method
Fig. 5 Simulated and measured energy spectra from 176Lu intrinsic
radioactivity for the small prototype PET scanner. Measured energy
spectrum was obtained by our small prototype PET scanner with 16
detectors consisting of a 14 9 14 9 4 array of 2.9 9 2.9 9 5.0 mm3
LGSO crystals. Measurement time was 1 h
Fig. 6 Simulated energy spectra with intrinsic coincidence events from a 176Lu intrinsic radioactivity and b 511-keV annihilation photons for
the whole-body LSO-PET scanner
E. Yoshida et al.
depends only on the diameter of the cylinder phantom. On
the other hand, the random coincidence reduction perfor-
mance of the MC-based reduction method depends on the
number of the LSO scintillator crystals.
Figure 11 shows the random fraction with use of the
TOF- and MC-based reduction methods, separately and
combined, for the NEMA cylinder phantom. At 50 kBq,
the true coincidence rate was about 60 cps, and the random
fraction was 10 % with the TOF?MC reduction methods.
Figure 12 shows the NECR and the NECR gain for the
TOF- and MC-based reduction methods, separately and
combined, for the NEMA cylinder phantom. At lower
activity, the NECR gain was increased with the TOF-based
and TOF?MC reduction methods.
3.3 Imaging study
Figure 13a shows reconstructed images of intrinsic random
coincidences for the 20-cm FOV with use of the random
coincidence reduction methods. These images were sum-
med for two centered slices. In background images, the
intrinsic random count had the same tendency as in Fig. 8.
The TOF?MC methods could provide the cleanest image.
The TOF- and MC-based reduction methods were not
observed to have any characteristic distribution.
Figure 13b shows reconstructed images of the phantom
with four weak sources for the random coincidence
reduction methods. The image obtained identified each
sphere for all situations. However, for the results of
425–650 keV, background from the intrinsic random
coincidences reduced the visual contrast. Table 2 lists
CNRs for weak-source phantom images. There was an
improvement in CNR for all sources for all three methods.
The TOF?MC method provided the highest CNR.
4 Discussion and conclusion
The intrinsic random coincidences depended on the energy
window. The intrinsic random coincidences had a broad
energy distribution and could not be prevented completely.
To prevent intrinsic random coincidences, using the narrow
Fig. 7 Intrinsic true and random count rates as a function of a LLD (ULD was 600 keV) and b ULD (LLD was 425 keV)
Fig. 8 Intrinsic random count rates with use of TOF- and MC-based reduction methods for cylinder phantoms with diameters of a 10 and
b 20 cm. The coincidence timing window was 4.5 ns for all cases
Reduction method for intrinsic random coincidence events
energy window that does not reduce the sensitivity is the
simplest method. After selection of the narrow energy
window (425–575 keV), the intrinsic random coincidence
rate could be reduced by about 70 % compared to the
energy window of 425–650 keV.
Using the MC-based reduction method, we could
reduce the intrinsic random coincidence rate by 30 %
compared to the energy window of 425–575 keV. The
intrinsic random coincidence rate with the energy window
of 100–1000 keV was 165 times larger than that with the
energy window of 425–575 keV. This is not a problem,
because the newest PET scanners have sufficient detect-
ability for the energy window of 100–1000 keV. How-
ever, NECR improvement of the MC-based reduction
method was limited. The MC-based reduction method can
be used for a conventional non-TOF-PET scanner. Also,
the performance of the MC-based reduction method
depended on the number of LSO scintillator crystals, as
shown in Fig. 10. As the axial FOV was extended and the
crystal thickness was large, the performance of the MC-
based reduction method was high. When the axial FOV is
extended for an entire whole-body PET scanner [21, 22],
for clinical studies it may not be possible to ignore the
intrinsic random coincidence for these LSO-PET scanners.
By using the TOF-based reduction method, we could
reduce the intrinsic random coincidence rate by 77 %
compared to the energy window of 425–575 keV. Also, the
NECR of the TOF-based reduction method was clearly
improved. The TOF-based reduction method had a higher
reduction capability than did the MC-based reduction
method. As shown in Fig. 9, the random coincidence
reduction performance did not depend on the timing reso-
lution. This was because the intrinsic random coincidences
had no true annihilation position on the LOR, and they
were spread out uniformly across the FOV. When the PET
scanner had a sufficient number of TOF bins, this reduction
performance depended on the occupancy of the imaging
target in the ring diameter. For the 10-cm-diameter cylin-
der phantom, the intrinsic random coincidence rate could
be ignored. However, the reduction performance of the
TOF-based reduction method depended on the size of the
object because the coincidence events detected outside an
object can be identified as random coincidence events. As
the diameter of the phantom was large, the reduction per-
formance was degraded. On the other hand, TOF infor-
mation can improve the imaging quality for well-known
enhancement of true coincidence. This enhancement per-
formance depends on the timing resolution. Crespo et al.
Fig. 9 Reduction rates of TOF- and MC-based reduction methods as a function of the diameter of the cylinder phantom for several timing
resolutions. The ideal line was defined by Eq. (1)
E. Yoshida et al.
[9] reported the feasibility of in-beam TOF-PET for
obtaining highly precise images.
As shown in Table 2, the images obtained had better
CNRs for all three reduction methods. As the activity of the
source was low, the CNR improvement was high for all
three reduction methods. Especially, the TOF?MC
reduction method had the highest CNR for all activities. On
the other hand, the difference between the TOF-based
reduction method and the MC reduction method was small.
This was because the background of the MC reduction
method that remained was about three times that of the
TOF-based reduction method for the whole FOV, but the
difference for the background in the region of the source
was limited.
Random and scatter corrections were not applied to the
data in this work. Simple subtraction-based random and
scatter corrections may not give statistically meaningful
events in a sparse distribution for a limited measurement
time. As a result, subtraction-based random and scatter
corrections may increase the statistical error.
For our 11C beam irradiation study, the average true
coincidence rate was assumed to be about 36 cps [10]. On
the other hand, the intrinsic random coincidence rate has
been calculated as 9 cps for the 20-cm-diameter cylinder
phantom with use of the TOF?MC reduction method. We
think that the intrinsic random coincidence can be ignored
for in-beam PET when the TOF?MC reduction method is
used because the activity distribution of in-beam PET is
confined to a localized area in the body.
In summary, we developed a reduction method for the
intrinsic random coincidence rate based on multiple-coin-
cidence information. We evaluated the reduction perfor-
mance using the developed MC-based reduction method
and the TOF-based reduction method, separately and in
Fig. 10 Reduction rates of TOF- and MC-based reduction methods as a function of the number of detector rings. There were three crystal
thicknesses, and the 20-cm-diameter cylinder phantom was used
Fig. 11 Random fraction with use of TOF- and MC-based reduction
methods for the NEMA cylinder phantom
Reduction method for intrinsic random coincidence events
combination. The developed method provided a 30 %
reduction in the intrinsic random coincidence rate, but the
contribution of the NECR was limited. On the other hand,
the TOF-based reduction method showed good promise for
reducing the intrinsic random coincidence rate, but the
level of the reduction for the TOF-based reduction method
depended on the size of the object. When the developed
method and the TOF-based reduction method were com-
bined, the reduction performance of the intrinsic random
coincidence rate was higher than that with use of only the
TOF-based reduction method.
Acknowledgments This work was supported by Grants-in-Aid for
Scientists Research of Kakenhi (22240065, 25242052, 24601019).
Conflict of interest The authors declare that they have no conflict
of interest.
Fig. 12 Results obtained with the TOF- and MC-based reduction methods for the NEMA cylinder phantom: a NECR and b NECR gain
Fig. 13 Reconstructed images of a the intrinsic random coincidence
events and b the phantom with four weak sources for the 20-cm FOV.
Background images are scaled for maximum intensity of the
425–575 keV energy window. Phantom images are scaled for
maximum intensity of the 250-Bq source
Table 2 CNRs for weak-source phantom images
ROI
activity
(Bq)
CNR
425–650 keV 425–575 keV MC TOF TOF?MC
100 1.45 1.64 1.76 1.89 1.95
250 2.01 2.62 3.07 3.19 3.22
500 3.78 4.36 4.40 4.52 4.55
1000 4.71 4.94 5.34 5.38 5.50
E. Yoshida et al.
References
1. Melcher CL, Schweitzer JS. Cerium-doped lutetium oxyorthosi-
licate: a fast, efficient new scintillator. IEEE Trans Nucl Sci.
1992;39:502–5.
2. Watson CC, Casey ME, Eriksson L, Mulnix T, Adams D,
Bendriem B. NEMA NU 2 performance tests for scanners with
intrinsic radioactivity. J Nucl Med. 2004;45:822–6.
3. Eriksson L, Watson CC, Wienhard K, Eriksson M, Casey ME,
Knoess C, et al. The ECAT HRRT: an example of NEMA scatter
estimation issues for LSO-based PET systems. IEEE Trans Nucl
Sci. 2005;52:90–4.
4. Yamamoto S, Horii H, Hurutani M, Matsumoto K, Senda M.
Investigation of single, random, and true counts from natural
radioactivity in LSO-based clinical PET. Ann Nucl Med.
2005;19:109–14.
5. Adonai N, Nguyen KN, Walsh J, Iyer M, Toyokuni T, et al. Ex
vivo cell labeling with 64Cu-pyruvaldehyde-bis(N4-methylthi-
osemicarbazone) for imaging cell trafficking in mice with posi-
tron-emission tomography. Proc Natl Acad Sci USA.
2002;99:3030–5.
6. Sundaresan G, Yazaki PJ, Shively JE, Finn RD, Larson SM,
Raubitschek AA, et al. 124I-labeled engineered anti-CEA mini-
bodies and diabodies allow high-contrast, antigen-specific small-
animal PET imaging of xenografts in athymic mice. J Nucl Med.
2003;44:1962–9.
7. Goertzen AL, Suk JY, Thompson CJ. Imaging of weak-source
distributions in LSO-based small-animal PET scanners. J Nucl
Med. 2007;48:1692–8.
8. Enghardt W. Charged hadron tumor therapy monitoring by means
of PET. Nuclear Inst Methods Phys Res A. 2004;525:284–8.
9. Crespo P, Shakirin G, Fiedler F, Enghardt W, Wagner A. Direct
time-of-flight for quantitative, real-time in-beam PET: a concept
and feasibility study. Phys Med Biol. 2007;52:6795–811.
10. Yamaya T, Yoshida E, Inaniwa T, Sato S, Nakajima Y, Wakizaka
H, et al. Development of a small prototype for a proof-of-concept
of OpenPET imaging. Phys Med Biol. 2011;56:1123–37.
11. Yoshida E, Kinouchi S, Tashima H, Nishikido F, Inadama N,
Murayama H, et al. System design of a small OpenPET prototype
with 4-layer DOI detectors. Radiol Phys Technol. 2012;5:92–7.
12. Surti S, Kuhn A, Werner ME, Perkins AE, Kolthammer J, Karp
JS. Performance of Philips Gemini TF PET/CT scanner with
special consideration for its time-of-flight imaging capabilities.
J Nucl Med. 2007;48:471–80.
13. Daube-Witherspoon ME, Surti S, Perkins A, Kyba CCM, Wiener
R, Werner ME, et al. The imaging performance of a LaBr 3-based
PET scanner. Phys Med Biol. 2009;55:45–64.
14. Jakoby BW, Bercier Y, Conti M, Casey ME, Bendriem B,
Townsend DW. Physical and clinical performance of the mCT
time-of-flight PET/CT scanner. Phys Med Biol.
2011;56:2375–89.
15. Conti M. Effect of randoms on signal-to-noise ratio in TOF PET.
IEEE Trans Nucl Sci. 2006;53:1188–93.
16. Jan S, Santin G, Strul D, Staelens S, Assie K, Autret D, et al.
GATE: a simulation toolkit for PET and SPECT. Phys Med Biol.
2004;49:4543–61.
17. Jan S, Benoit D, Becheva E, Carlier T, Cassol F, Descourt P, et al.
GATE V6: a major enhancement of the GATE simulation plat-
form enabling modeling of CT and radiotherapy. Phys Med Biol.
2011;56:881–901.
18. McIntosh B, Stout DB, Goertzen AL. Validation of a GATE
Model of 176Lu intrinsic radioactivity in LSO PET systems. IEEE
Trans Nucl Sci. 2011;58:682–6.
19. Performance measurements of positron emission tomographs.
Nat. Elect. Manufact. Assoc., Rosslyn, VA, NEMA Standards
Pub. NU 2-2001, 2001.
20. Daube-Witherspoon ME, Muehllehner G. Treatment of axial data
in three dimensional PET. J Nucl Med. 1987;28:1717–24.
21. Couceiro M, Blanco A, Ferreira F, Ferreira Marques R, Fonte P,
Lopes L. RPC-PET: status and perspectives. Nuclear Inst Meth-
ods Phys Res A. 2007;580:915–8.
22. Eriksson L, Conti M, Melcher CL, Townsend DW, Eriksson M,
Rothfuss H, et al. Towards sub-minute PET examination times.
IEEE Trans Nucl Sci. 2011;58:76–81.
Reduction method for intrinsic random coincidence events