Four-layer DOI PET detectors using a multi-pixelphoton counter array and the light sharing method

Fumihiko Nishikido n, Naoko Inadama, Eiji Yoshida, Hideo Murayama, Taiga YamayaMolecular Imaging Center, National Institute of Radiological Sciences, Chiba 263-8555, Japan

a r t i c l e i n f o

Article history:Received 1 April 2013Received in revised form6 August 2013Accepted 6 August 2013Available online 26 August 2013

Keywords:PET detectorDOI detectorSilicon photomultiplierMPPC

a b s t r a c t

Silicon photomultipliers (SiPMs) provide many advantages for PET detectors, such as their high internalgain, high photon detection efficiency and insensitivity to magnetic fields. The number of detectablescintillation photons of SiPMs, however, is limited by the number of microcells. Therefore, pulse height ofPET detectors using SiPMs is saturated when large numbers of scintillation photons enter the SiPM pixels.On the other hand, we previously presented a depth-of-interaction (DOI) encoding method that is basedon the light sharing method. Since our encoding method detects scintillation photons with multiplereadout pixels, the saturation effect can be suppressed. We constructed two prototype four-layer DOIdetectors using a SiPM array and evaluated their performances.

The two prototype detectors consisted of four layers of a 6�6 array of Lu2(1�x)Y2xSiO5 (LYSO) crystalsand a SiPM (multi-pixel photon detector, MPPC, Hamamatsu Photonics K.K.) array of 4�4 pixels. The sizeof each LYSO crystal element was 1.46 mm�1.46 mm�4.5 mm and all surfaces of the crystal elementswere chemically etched. We used two types of MPPCs. The first one had 3600 microcells and high photondetection efficiency (PDE). The other one had 14,400 microcells and lower PDE. In the evaluationexperiment, all the crystals of the detector using the MPPC which had the high PDE were clearlyidentified. The respective energy and timing resolutions of lower than 15% and 1.0 ns were achieved foreach crystal element. No saturation of output signals was observed in the 511 keV energy region due tosuppression of the saturation effect by detecting scintillation photons with several MPPC pixels by thelight sharing method.

& 2013 Elsevier B.V. All rights reserved.

1. Introduction

Silicon photomultipliers (SiPMs) offer many advantages asphoto-detectors used in positron emission tomography (PET)detectors. These include having a high internal gain (4105), highphoton detection efficiency and insensitivity to magnetic fields.However, they have some disadvantages, such as high temperaturedependence, sensitivity to bias voltage fluctuations and a limita-tion in the number of detectable photons. SiPMs can detect asmaller number of scintillation photons than the number ofmicrocells at one time. As a result, output signals of PET detectorsusing SiPM are saturated when the number of photons enteringthe SiPMs increase (this is called the saturation effect). Thesaturation effect has been reported in some studies of PETdetectors, especially one-to-one coupling detectors in which eachcrystal element of a scintillator array is directly coupled to a singlephoto-detector pixel [1–4].

A “light sharing detector” which detects scintillation photonsusing several photo-detectors [5–8] is one way to avoid thesaturation effect. The scintillation crystals are identified with floodhistograms obtained from the centroid of the scintillation photondistribution detected in the photo-detectors. Detection of thescintillation photons by several photosensors leads to reducedinfluence by the saturation in the case of the SiPM detectors.Additionally, the number of photo-detectors can be smaller thanthat of the scintillation crystals when a crystal block consists ofsmaller size crystals than the photo-detector pixels.

The depth-of-interaction (DOI) encoding method is an impor-tant technique used in PET detectors to achieve a high imagingperformance. Generally, spatial resolution of reconstructed imagesis degraded in the peripheral field of view without the DOItechnique due to the parallax error. This is especially an issue forsmall animal PET scanners which consist of a small detector ring toachieve not only high spatial resolution but also high scannersensitivity. Such a geometry increases the number of gamma rayswhich are obliquely incident to the scintillation crystals in the PETdetectors. Therefore, the influence of the parallax error increasesand the DOI-PET detectors are required to suppress the parallaxerror. Many research groups have reported on various types of

Contents lists available at ScienceDirect

journal homepage: www.elsevier.com/locate/nima

Nuclear Instruments and Methods inPhysics Research A

0168-9002/$ - see front matter & 2013 Elsevier B.V. All rights reserved.http://dx.doi.org/10.1016/j.nima.2013.08.031

n Corresponding author. Tel.: þ81 43 206 3260.E-mail address: funis@nirs.go.jp (F. Nishikido).

Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761

DOI-PET detectors. For example, DOI information can be obtainedby stacking single layer crystal arrays [8–10], detecting scintillationphotons at two opposite sides of the crystal arrays (dual-endreadout detector) [11,12], using pulse shape discrimination (phos-wich detectors) [13,14], and using staggered dual layer crystalarrays [15].

We have proposed the four-layer DOI encoding method basedon the light sharing method [16]. In our method, reflectors areinserted alternately between the crystals in a crystal block. Thescintillation crystals in all four layers can be identified by photo-detectors mounted on only one side of the crystal block. Sharing ofthe scintillation photons can suppress the saturation effect and theone side readout can reduce the number of photo-detectorsneeded, compared with dual-end readout and stack type DOIdetectors.

In this paper, we constructed two prototype DOI detectorsbased on our four-layer DOI encoding method and evaluated thedetector performances. The detectors each use one of two types ofSiPM arrays produced by Hamamatsu Photonics (multi-pixelphoton counter, MPPC).

2. Materials and methods

2.1. Four-layer DOI detectors using two types of MPPC arrays

The prototype detectors consisted of a 4�4 MPPC array(S11064 series, Hamamatsu Photonics K.K.) and a Lu2(1�x)Y2xSiO5

(LYSO, Proteus Inc.) (Lu: 98% and Y: 2%) crystal block.Two types of the MPPCs, S11064-050P (microcell size: 50 mm)

and S11064-025P (microcell size: 25 mm) designated 050P and025P below, were used in the experiment. Each of the MPPC arrayshad a 3.0 mm�3.0 mm sensitive area. The 050P had a higher fillfactor (61.5%) than the 025P had (30.8%). Therefore, the photondetection efficiency (PDE) of the 050P was twice as large as that ofthe 025P. However, the number of microcells of the 050P (3600cells) was smaller than that of the 025P (14,400 cells). The gainvariations of the 16 MPPC pixels used in our study were 5.95% forthe 025P and 5.28% for the 050P. The same crystal array was usedto compare effects from only the photodetector characteristics onthe performance of the detectors for each MPPC type.

The crystal block consisted of four layers of a 6�6 array withLYSO crystals. The size of each crystal element was 1.46 mm�1.46 mm�4.5 mm and all surfaces were chemically etched. Weadopted the reflector arrangement between the crystals as proposedin Ref. [16] for DOI encoding. The reflector was the multilayerpolymer mirror (Sumitomo 3M, Ltd.) of 0.065 mm thickness and98% reflectivity. RTV rubber (KE420, Shin-Etsu Chemical Co., Ltd.)was applied between crystals except for the space where thereflector was inserted. Details of the arrangement of the reflectorand RTV rubber in each layer are shown in Fig. 1. The positionalrelationship of the LYSO array and the MPPC pixels is shown in Fig. 2.

2.2. Trigger and data acquisition system

Fig. 3 shows a block diagram of a trigger and data acquisitionsystem. The MPPC output signals were fed into simple non-inverting amplifier circuits and divided into two signals. The firstsignal was individually recorded by the 16 ch charge-sensitiveanalog-to-digital converter (ADC; C009H, Hoshin Electronics Co.,Ltd.) as list-mode data. The other signal was fed into sum modulesand used as the trigger and timing signal. For coincident timingmeasurements a BaF2 scintillator was coupled to a PMT (H5321,Hamamatsu) and amplified with a photomultiplier amplifier(PM-AMP; N018, Hoshin). The summed signal of the MPPC arraysand the PMT signal were fed into individual constant fractiondiscriminators (CFDs; 583B and 584, ORTEC). The CFD signals werefed to a time-to-digital converter (TDC; C021, Hoshin) to be used asstop signals. A coincidence signal of the two CFD signals was usedas the start signal of the TDC.

2.3. Experiment

Crystal identification performance and energy spectra for theprototype DOI detectors were measured with both MPPC arrays byusing a 22Na point source without the coincidence detector.Linearity of output signals for the gamma ray energy was alsomeasured for the detectors. The four-layer DOI detectors wereuniformly irradiated by gamma rays from point sources of 241Am(59.5 keV), 139Ce (165.9 keV), 51Cr (320 keV), 22Na (511 keV and1275 keV), 137Cs (662 keV) and 54Mn (835 keV). Each point sourcewas positioned at approximately 10 cm from the top of the DOIdetector. In the measurement for the two lower energy gamma raysources (241Am and 139Ce), another point source was also posi-tioned at the backside of the DOI detectors to increase the numberof gamma rays reaching the lower layer crystals. Temperature

2ndlayer 3rdlayer 4thlayer1stlayer

Fig. 1. Reflector arrangement for the crystal block. The black and gray lines show the reflectors and RTV rubber, respectively. Annihilation radiations enter the block fromboth the 1st layer side. The 4th layer coupled to the MPPC array.

Fig. 2. Positional relationship of the 6�6�4 LYSO array and the MPPC pixels ofthe prototype detectors. Dotted areas show sensitive areas of the MPPCs.

F. Nishikido et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761756

variance of the detectors was kept within 0.1 1C by placing them ina constant temperature box.

For comparison of the influence from the saturation effectbetween the one-to-one coupling detector and the four-layer DOIdetector, energy spectra were measured with a single LGSO crystalcoupled to a single pixel of both MPPCs for various energy gammarays. The crystal size was 3.0 mm�3.0 mm�3.0 mm. The crystalwas wrapped in the multilayer polymer mirror and opticallycoupled to the MPPC by optical grease. The other experimentalconditions were the same as described above.

3. Results

3.1. Comparison of the DOI detectors using the two types of MPPCs

To determine interacting crystals, we adopted the Anger-typecalculation for nine anode outputs and obtained 2D positionhistograms. Fig. 4(a) and (b) shows 2D position histogramsobtained by the uniform irradiation of 511 keV gamma rays forthe DOI detectors using the 050P and 025P arrays, respectively. Allof the crystal elements on the position histogram of the 050P areclearly identified. On the other hand, the spots on the positionhistogram of the 025P are blurred and some spots cannot be

distinguished. Some large gaps between the spots in the positionhistograms are caused by mismatch between the crystal size andthe MPPC pixel pitch.

Energy spectra were obtained by summing the data from allthe MPPC pixels in each event. Fig. 5 shows energy spectra ofselected single crystals at the center of the crystal array in eachlayer of the detectors using the 050P and 025P for uniformirradiation of 511 keV gamma rays from the 22Na point source.Energy resolutions of 11.0–13.6% (FWHM) are obtained for theenergy spectra of the detector using the 050P. The energyresolutions of the same crystals for the detector using the025P are 12.7–15.2% (FWHM).

Figs. 6 and 7 show pulse heights of the detectors using the 050Pand 025P arrays in each layer for various energy gamma rays. Theselected crystals are positioned at the center of the crystal arrayand just above one of the MPPC pixels. Therefore, the number ofdetected photons at the MPPC pixels from the selected crystals islarger than the other crystals in the same layer. Each plotted pointis obtained from photo-peak positions for each gamma ray source.Linearity of the pulse height of the detector using the 025P array iskept until 1.2 MeV. On the other hand, the pulse height of thedetector using the 050P array is saturated for high energy gammarays. This is caused by the smaller number of microcells and thehigher PDE of the 050P compared to the 025P. However, this resultis not problematic for use as a PET detector because the linearity isacceptable for energies under 511 keV.

The results of the crystal identification performance, energyresolution and linearity show that the detector using the 050Parray achieves better performance than the 025P array as a PETdetector. Therefore, only the results of the prototype DOI detectorusing 050P are described in the following sections.

3.2. Timing resolution of the four-layer DOI detector usingthe 050P MPPC array

Fig. 8 shows the timing spectrum for single crystal elements ineach layer of the detector using the 050P MPPC array. Timingresolution of 0.75 ns (FWHM) is obtained as coincidence betweenthe single crystal element of the 1st layer at the center of thecrystal array and the BaF2 detector. The timing resolution of theBaF2 detector itself is 0.42 ns estimated with two of the same typeof BaF2 detectors. The timing resolution of 0.62 ns for the crystalFig. 3. Block diagram of the trigger and data acquisition system.

Fig. 4. Position histograms for the DOI detectors using (a) S11064-050P (3600 pixels) and (b) S11064-025P (14,400 pixels).

F. Nishikido et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761 757

elements of the 1st layer is obtained after deconvolution of thecoincidence detector resolution. The timing resolutions of 2nd, 3rdand 4th layers are 0.64 ns, 0.58 ns and 0.66 ns, respectively.

3.3. Variations of detector performance in different positions

Figs. 9–11 show pulse height of photopeaks, energy resolutionand timing resolution of single crystal elements in all layers of theDOI detector using the 050P. The pulse heights of the 1st, 2nd, 3rdand 4th layers are 234.1733.6 ch, 235.1726.8 ch, 260.5732.4 chand 282.4733.9 ch, respectively. The 4th layer is the bottom layerand it is directly coupled to the MPPC array. As a result, the pulseheight is highest in the 4th layer and decreases in upper layers.From the 1st to 4th layers, the energy resolutions are 11.271.0%,11.471.0%, 10.971.1% and 11.371.5% (FWHM) and the timingresolutions are 0.7770.11 ns, 0.7670.09 ns, 0.6970.09 ns and0.6670.07 ns (FWHM), respectively.

Figs. 12 and 13 show energy and timing resolutions for singlecrystal elements of the DOI detector using the 050P as a functionof pulse height of photopeaks in each layer. The energy resolutionhas a slight tendency to improve as pulse height is increased. Thetiming resolution is clearly improved from 1.0 ns to 0.6 ns as pulseheight is increased 1.7 times.

3.4. Results of the single crystal detector

Fig. 14 shows pulse heights of photopeaks as a function ofgamma ray energy for the single LGSO crystal using 050P and 025Parrays. The dotted line shows fitting results for the plotted valuesof the 025P array using a linear function. The fitting resultsindicate that the pulse heights for 025P are not saturated in thewhole energy range below 1275 keV. On the other hand, the datafor 050P are clearly saturated at all energies above 300 keV. Theenergy resolutions of 13.0% for 050P and 12.5% for 025P areobtained after a non-linearity correction for 511 keV gamma rays.

4. Discussion

From the results of the crystal identification performanceand energy resolution, we concluded the detector using the

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

1st Layer

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

2nd Layer

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

4th Layer

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

3rd Layer

Fig. 6. Linearity of crystal elements positioned just above the sensitive area for various energy gamma rays in the four layers of the detector using the 050P MPPC array.

Fig. 5. Energy spectra for crystal elements on each layer of the detectors using thetwo types of MPPCs. (a) S11064-050P (3600 pixels) and (b) S11064-025P (14400pixels).

F. Nishikido et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761758

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

1st Layer

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

2nd Layer

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

4th Layer

Pea

k P

ulse

hei

ght[c

h]

Energy [keV]

3rd Layer

Fig. 7. Linearity of crystal elements positioned just above the sensitive area for various energy gamma rays in the four layers of the detector using the 025P MPPC array.

1400

1200

1000

800

600

400

200

0

Cou

nts

7654321Time [nsec]

1st Lyaer 2nd Layer 3rd Layer 4th Layer

Fig. 8. Timing spectra for single crystal elements in four layers.

150 200 250 300 350 4000

5

10

15

20

Pulse height

Num

ber o

f cry

stal

s

1st layer 2nd layer 3rd layer 4th layer

Fig. 9. Pulse height of each single crystal of the prototype DOI detector with the050P array for 511 keV gamma rays.

8 9 10 11 12 13 14 15 160

5

10

15

20

25

30

Num

ber o

f cry

stal

s

Energy resolution (%)

1st layer 2nd layer 3rd layer 4th layer

Fig. 10. Energy resolution of each single crystal of the prototype DOI detector withthe 050P array for 511 keV gamma rays.

0.5 0.6 0.7 0.8 0.9 1.00

2

4

6

8

10

12

14

Num

ber o

f cry

stal

s

Timing resolution (ns)

1st layer 2nd layer 3rd layer 4th layer

Fig. 11. Timing resolution of each single crystal of the prototype DOI detector withthe 050P array for 511 keV gamma rays.

F. Nishikido et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761 759

050P array achieves better performance than the 025P as a PETdetector. This comes from the PDE of the 050P being two timeshigher than that of the 025P. In addition, no saturation ofoutput signals was observed in the 511 keV energy region dueto suppression of the saturation effect by detecting scintilla-tion photons with several MPPC pixels by the light sharingmethod. These results indicate that the 050P is more appropriate

than the 025P for use as the photo-detector of the four-layer DOI PETdetector.

From the results of the single crystal detectors using the 050Pand 025P MPPCs, the energy resolution of the 025P is slightlybetter than that of the 050P even though the PDE of the 025P istwo times lower than that of the 050P. Although an increase of thePDE ordinarily improves energy resolution in conventional photo-detectors, the saturation effect degrades energy resolution of thedetector using the 050P MPPC since the number of the microcellsof the 050P is smaller than the number of scintillation photonsentering the MPPC pixels. This means that the saturation effectlimits the potential of the 050P in the case of the one-to-onecoupling detectors. On the other hand, the number of detectablephotons is increased in the four-layer DOI detector based onseveral photo-detector readouts. Therefore, the saturation effectis suppressed and the energy resolution of the four-layer DOIdetector using 050P is superior to that of the detector using 025P.

The timing and energy resolutions depend on the pulse heightwhich is proportional to the number of detected photons as shownin Figs. 13 and 14. In the four-layer DOI detectors used in theexperiment, 44% of the MPPC surface under the crystal block is aninsensitive area (see Fig. 2) and such a large insensitive area causesdegradation of the detector performance. Especially, performanceof the crystal elements just above the dead spaces of the MPPCarray is degraded. In order to improve performance of the four-layer DOI detector, it is necessary to recover the scintillationphotons lost in the dead space. We expect that use of an MPPCarray with a small insensitive area [4,17] can improve performanceof the four-layer DOI detector, increasing the number of detectablescintillation photons.

5. Conclusion

We developed prototype four-layer DOI detectors, based on thelight sharing method, using two types of MPPCs in the 4�4 arrayand evaluated their detector performances. The experimental resultsshowed that the S11064-050P was superior to the S11064-025P as aphoto-detector in the four-layer DOI-PET detector due to its higherPDE. In addition, the MPPC with a smaller number of microcells canbe used without performance degradation in the four-layer DOIdetector based on the light sharing method in comparison with theone-to-one readout detectors since the light sharing method cansuppress the saturation effect.

The experimental data indicated that the energy and timingresolutions depend on the number of the photons detected withthe MPPC array. In the future, we will improve the detectorperformances of the four-layer DOI detector, using a MPPC arraywith a small insensitive area.

Acknowledgment

This study was supported by JSPS KAKENHI Grant No.20700416.

References

[1] D.R. Schaart, S. Seifert, R. Vinke, et al., Phys. Med. Biol. 55 (2010) N179.[2] A.N. Otte, J. Barral, B. Dolgoshein, et al., Nucl. Instrum. Methods A 545 (2005)

705.[3] L. Cosentino, F. Cusanno, R. DeLeo, et al., Nucl. Instrum. Methods A 702 (2013)

10.[4] T. Kato, J. Kataoka, T. Nakamori, et al., Nucl. Instrum. Methods A 638 (2011) 93.[5] A. Kolb, E. Lorenz, M.S. Judenhofer, et al., Phys. Med. Biol. 55 (2010) 1815.[6] G. Ko, H. Yoon, S. Kwon, Nucl. Instrum. Methods A 703 (2013) 38.[7] T. Song, H. Wu, S. Komarov, Phys. Med. Biol. 55 (2010) 2573.[8] D.J. Herbert, S. Moehrs, N. D’Ascenzo, Nucl. Instrum Methods A 573 (2007) 84.

16

14

12

10

8

6

4

2

0

Ene

rgy

reso

lutio

n (%

)

400350300250200150Pulse height (ch)

'1st layer' '2nd layer' '3rd layer' '4th layer'

Fig. 12. Dependence of the energy resolution on pulse height of 511 keV gammaray events for each single crystal of the prototype DOI detector with the 050P array.

1.0

0.8

0.6

0.4

0.2

0.0

Tim

ing

reso

lutio

n (n

s)

400350300250200150Pulse height (ch)

'1st layer' '2nd layer' '3rd layer' '4th layer'

Fig. 13. Dependence of the timing resolution on pulse height of 511 keV gamma rayevents for each single crystal of the prototype DOI detector with the 050P array.

0 200 400 600 800 1000 1200 14000

500

1000

1500

2000

2500

3000

Pul

se h

eigh

t (a.

u.)

Energy (keV)

Fig. 14. The pulse height of the single crystal detectors as a function of gamma rayenergy.

F. Nishikido et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761760

[9] F.W.Y. Lau, A. Vandenbroucke, P.D. Reynolds, Phys. Med. Biol. 55 (2010) 7149.[10] T. Omura, T. Moriya, R. Yamada, Nuclear Science Symposium Conference

Record, IEEE, 2012, pp. 3560.[11] Y. Shao, H. Li, K. Gao, Nucl. Instrum. Methods A 580 (2007) 944.[12] Y. Yang, P.A. Dokhale, R.W. Silverman, et al., Phys. Med. Biol. 51 (2006) 2131.[13] M. Dahlbom, L.R. MacDonald, M. Schmand, et al., IEEE Trans. Nucl. Sci. NS-45

(3) (1998) 1128.

[14] J. Seidel, J.J. Vaquero, S. Siegel, et al., IEEE Trans. Nucl. Sci. NS-46 (3) (1999) 485.[15] H. Liu, T. Omura, M. Watanabe, et al., Nucl. Instrum. Methods A 459 (2001)

182.[16] T. Tsuda, H. Murayama, K. Kitamura, et al., IEEE Trans. Nucl. Sci. NS-51 (2004)

2537.[17] G. Llosá, J. Barrio, C. Lacasta, et al., Phys. Med. Biol. 55 (2010) 7299.

F. Nishikido et al. / Nuclear Instruments and Methods in Physics Research A 729 (2013) 755–761 761