A Continuous Crystal Detector for TOF PET

Transcription

1 1 A Continuous Crystal Detector for TOF PET T. Szczęśniak, Member, IEEE, M. Moszyński, Fellow, IEEE, Ł. Świderski, Member, IEEE, A. Nassalski, Member, IEEE, A. Syntfeld-KaŜuch, Member, IEEE, P. Ojala, and C. Bohm Abstract Presently, a majority of Positron Emission Tomography (PET) systems is based on block-detectors consisting of many scintillating pixels read by photomultipliers (PMTs). An improvement of time resolution, due to a common light readout by a cluster of PMTs, proposed by Kuhn et al, and tested by us for an LSO crystal triggered an idea of the new PET detector optimized for the Time of Flight (TOF) systems, based on a continuous crystal. In the present work, optimization of timing for a 20x20x20 mm 3 LYSO crystal coupled to the 16-channel photomultiplier H MOD from Hamamatsu is presented. First, the measurements of transit time jitter, number of photoelectrons and time resolution with a small 10x10x5 mm 3 LSO crystal coupled to the H MOD PMT were performed. Results were compared with data collected for fast timing photomultipliers like Photonis XP1020, XP3060, XP20D0 or Hamamatsu R9800. In the second part of the study, the time resolution measurements and the optimization of the system with the continuous LYSO crystal were made. Also simple tests with positioning of gamma interaction inside the scintillator are reported. The final results are discussed in terms of the measured photoelectron number and requirements for the TOF-PET scanners. recorded with the 10x10x5 mm 3 and 4x4x20 mm 3 LSO crystals coupled to the Photonis XP20D0 photomultiplier [2] suggesting that an LSO or LYSO scintillators (7.4 g/cm3) seems to be a very good candidates for the future TOF-PET [3]. A further improvement of the time resolution, due to the common light readout by the cluster of PMTs in the block detector, was proposed by Kuhn et al [4]. The study of obtainable time resolution with the LSO crystal with light readout by several photomultipliers was reported by us in [5]. A large improvement in timing for the sum of signals from two photomultipliers has been achieved for the case when the 4x4x20 mm 3 LSO scintillator was placed in the middle of the light guide, between PMTs. This was a consequence of a better light collection by the two PMTs, which gave 1000 photoelectrons for 511 kev in comparison to only about 500 photoelectrons measured with PMT 1 and PMT 2 separately. It led to an improvement of the time resolution from about 490 ps to 350 ps (see Fig. 1). Index Terms Time-of-flight PET, Fast timing, Fast PMTs, LSO scintillator. A I. INTRODUCTION DDITIONAL information about time-of-flight (TOF) of annihilation quanta collected in Positron Emission Tomography (PET) scanners allowed improving their performance and image quality. The achieved level of advancement depends very strongly on a design and a precise optimization of the detectors used, particularly their timing capabilities. Presently, a majority of detectors for PET systems is based on scintillator crystals read by several photomultipliers. Commercially available TOF PET systems with pixelated block detectors are characterized by coincidence timing resolution around 600 ps [1]. In our previous work, a very good time resolution was Manuscript received November 14, This work was supported in part by the FP6 Commission "Biocare" (Molecular Imaging for Biologically Optimized Cancer Therapy) under the contract number and by SPUB 621/E-78/SPB/6 PR UE/DIE 458/ of Polish Committee for Scientific Research. T. Szczęśniak ( t.szczesniak@ipj.gov.pl), M. Moszyński, Ł. Świderski, A. Nassalski, A. Syntfeld-KaŜuch are with The Sołtan Institute for Nuclear Studies, PL Otwock-Świerk, Poland. P. Ojala and C. Bohm, are with Department of Physics, Stockholm University, AlbaNova University Centre, SE Stockholm, Sweden. Fig. 1. The time spectra measured with the 4x4x20 mm 3 LSO scintillator coupled to a system consisting of a light guide and two Photonis XP20D0 PMTs. The crystal was placed in the middle, between the PMTs and measurements were made for the 511 kev full energy peak in relation to a reference detector based on BaF 2 crystal and fast PMT. The values corrected for the reference detector are equal to 350 ± 12 and 486 ± 16 ps, respectively [5]. A further progress in production of new photomultipliers [6] and scintillators [7] is one of the possibilities of an improvement of the time resolution in PET systems. However the development process is long and not always successful. The characterization and comparison of different sensor configurations for monolithic scintillation crystal presented by

2 2 Ojala et al [8] triggered an idea to replace common pixelated block detectors used in the present TOF PET scanners by the continuous crystals. A comparative study of a number of different photomultipliers in fast timing, done in [9], showed an importance of the number of photoelectrons collected in a photomultiplier for the optimal time resolution. Application of one, continuous scintillator, instead of many small pixels should considerably increase the number of photons collected at photocathode of the PMT. Since the time resolution is inversely proportional to the square root of the photoelectron number, the timing performance should be highly improved. Of course, such a change involves introducing multichannel photodetector to assure good determination of position of gamma interaction inside the crystal. The aim of this paper is to study and optimize timing properties of the detector consisting of 16-channel photomultiplier H MOD from Hamamatsu and 20x20x20 mm 3 LYSO crystal from Saint-Gobain. The PMT with a high quantum efficiency of 43% was selected for the study. II. EXPERIMENTAL DETAILS A. Scintillators and Photomultiplier The studies were carried out on three scintillating crystals. The 10x10x5 mm 3 LSO was used for general timing characterization of the used photomultiplier. The 20x20x20 mm 3 LYSO crystal was tested as a candidate for the future continuous TOF PET detector. Finally, measurements with the 4x4x20 mm 3 LSO scintillator were performed for comparison with the results obtained for monolithic LYSO. In each case the mechanically polished LSO/LYSO crystals were wrapped with several layers of white Teflon tape. The Hamamatsu 16-channel, multianode photomultiplier H MOD was chosen for the presented experiments. The main properties of the PMT are presented in Table I and the dimensional outline of its input window is presented in Fig. 2. Note, a very high quantum efficiency of 43%, and single channel size comparable to a typical finger-like crystal commonly used in PET. TABLE I THE MAIN PROPERTIES OF THE HAMAMATSU H MOD PHOTOMULTIPLIER White Sens. [ua/lm] Type H MOD Blue Sens. [ua/lm] Serial No. ZB0730 QE [%] at 350nm Uniformity Max/Min Photocathode Window Structure Stages Bialkali Borosolicate Glass Metal Channel Dynodes 12 Fig. 2. The dimensional outline of the Hamamatsu H MOD input window. The linear voltage divider with the ratios of: was made by Hamamatsu and was integrated into the PMT assembly. B. Experimental Methods The number of photoelectrons per energy unit (phe/mev) for all the crystals was measured by the Bertolaccini et al method [10], it means by a comparison of the peak position due to the single photoelectrons (which determines the gain of the photomultiplier) with the position of the 662 kev full energy peak from a 137 Cs gamma source [11]. The time jitter of the tested photomultipliers was measured using a light pulser based on an XP22 light emitting diode (LED) driven by an avalanche transistor. The FWHM of the light pulse was determined in the past by a microchannel plate photomultiplier to be equal to 130 ps [12]. The effective wavelengths of the light seen by the PMT with the bi-alkali photocathode were centered at 560 nm. The time resolution study was performed in coincidence experiments with 511 kev annihilation quanta from a 22 Na gamma source. In a reference detector, a truncated cone, 20 mm and 25 mm in diameter and 15 mm high BaF 2 crystal coupled to the XP20Y0Q/DA PMT was used. Its time resolution of 128 ± 4 ps for the 511 kev full energy peak, was reported in [13]. In all the timing measurements, a slow-fast arrangement was used for a precise selection of the required energy windows set at the 511 kev full energy peaks from a 22 Na γ-source. C. Experimental Set-up The detailed diagram of the experimental set-up is presented in Fig. 3. In the fast signal electronic chain, that used last dynode signal, the time spectrum of the response difference of the detectors was taken. In the slow signal electronic chain, formed using sum of anode signals, the gate was generated, to choose the energy range of interest. In the case of the time resolution measurements energy windows were set at 511 kev

3 3 full energy peaks. For the time jitter the energy window was set at the single photoelectron peak and instead of a reference detector the pulse from the XP22 LED was used to trigger the starting discriminator. Due to fast decay time of the anode pulse the spectroscopy amplifier was working properly even without preamplifier. Examples of the signal after amplifier and the sum of 16 anode signals are presented in Fig. 4. Fig. 4. The screen shots from a digital scope showing the output signal after spectroscopy amplifier (left) and the sum of 16 anode signals (right). Time scale is equal to 20ns/div. III. RESULTS AND DISCUSSION Fig. 3. The slow-fast arrangement for timing measurements. In the fast part (red), the time spectrum of the response difference of the detectors is taken. In the slow part (blue), the gate is generated, to choose the energy range of interest. The timing measurements were done using a fast leading edge (LE) discriminator Polon 1520 and a Constant Fraction Discriminator (CFD) Ortec 935. The time spectra were measured with an Ortec 566 Time-to-Amplitude Converter and recorded by a PC-based multichannel analyzer (Tukan8k) [14]. Time calibration of the Time-to-Amplitude Converter was done using a precise Time Calibrator Ortec 462, based on a quartz clock. In our previous experiment, described in [5] a sum of anode signals was made by simple connection of two short cables of anode outputs into one cable. Such a simple method is not useful in the case of a PET detector since the anode signals are needed for determination of a gamma interaction position. Moreover, summing of the 16 anode signals into one cable destroys the rise time of the output pulse, leading to significant deterioration of a time resolution. The metal channel photomultipliers like H8711 possess a common dynode structure, so the signal from the last dynode can be used instead of the sum of the anodes. In the experimental set-up the sum of the anode signals was used for gate generation. Each signal was plugged into a spectroscopy amplifier (2x Stelzer MA8000) and then the output signals were send to the summing module. The module had sixteen 500 ohm inputs and one output with 10k ohms connected parallel. The output signal was send to the Single Channel Analyzer. A. Time jitter In the first part of the study a general characterization of timing capabilities of the Hamamatsu H MOD photomultiplier was made. The tests covered a time jitter and timer resolution measurements with the 10x10x5 mm 3 LSO crystal. The time jitter spectra recorded with the tested PMT are presented in Fig. 5. During the experiments XP22 LED illuminated the whole photocathode but the signal was read in three different modes corresponding to the area of interest of the photocathode. First, the gate was generated using a sum of all 16 anode signals. In the next step only 4 central pixels were used and finally the time jitter was tested for one pixel. Fig. 5. Transit time jitter spectra measured with XP22 LED illuminating whole area of the photocathode. First the gating was set to choose all channels of the PMT (top-left), and then the gate was changed to allow events only for central channels (top-right) and one channel (bottom).

4 4 The result obtained for the tested Hamamatsu metal channel PMT are similar to the properties of commonly used fast photomultipliers with linear focusing like Hamamatsu R9800, XP1020, XP3060 or XP20D0 [9], [15]. However, considerable deterioration of the time jitter is seen when the output signal involves more anode signals. In such a case the final spectra is broadened due to transit time differences between individual channels. B. Time resolution with 10x10x5 mm 3 LSO This section is the second part of the general characterization of timing capabilities of the Hamamatsu H MOD photomultiplier. The presented results are compared with the timing data collected in our previous works [2], [9], [15]. The size of the selected scintillator allowed the determination of the limit of achievable time resolution not affected by light transport and attenuation in the big crystal. Because for PET application the anode signals in the PMT are used for identification, where gamma interaction took place, in the case of time resolution measurements the signal from last dynode was used. Such way of signal readout has another big advantage. Since dynode structure of H8711 photomultiplier (metal channel dynodes) is common for all channels and only anode signals are separated, the last dynode signal can be used for optimal light collection instead of summing of anode signals, as it was presented in [5]. Fig. 6 presents the time spectrum measured with the 10x10x5 mm 3 LSO crystal in relation to the BaF 2 reference detector. In this case the gate was created using 4 central pixels. detector, after subtracting a contribution of a reference BaF 2 detector is presented in the third column. The number of photoelectrons per MeV and for 511keV is collected in column fourth and fifth, respectively. In the last column the time resolution normalized to the number of photoelectrons and excess noise factor is showed. Timing capabilities of the Hamamatsu H8711 phototube are as good as other candidates for a future TOF PET. However the Hamamatsu tube suffers because of a poor electron collection efficiency which is reflected in a low number of photoelectrons. Such good value as 198 ps was possible due to a very high QE of 43%. TABLE II THE TIME RESOLUTION MEASURED WITH THE LSO10X10X5 COUPLED TO VARIOUS FAST PHOTOMULTIPLIERS The final result of 198 ps is an outcome of an optimization of the experimental set-up settings, especially the threshold set at a LE discriminator. The dependency of timing with fraction of a signal triggering LE discriminator for the H8711 photomultiplier in comparison to the linear focused PMTs is presented in Fig. 7. The plot was discussed in more details in [9]. In the case of H8711 phototube an optimum is seen around 5% fraction, similar to the classic fast photomultipliers. Fig. 6. The time spectrum measured with the 10x10x5 mm 3 LSO crystal placed in the center of the Hamamatsu H MOD 16-channel photomultiplier. The more detailed results of the time resolution measurements together with a number of photoelectrons are collected in Table II. Also the data collected for various fast timing photomultipliers with linear focusing dynode structure are presented. The time resolution at FWHM for a single Fig. 7. The time resolution dependence on a LE threshold for the H MOD PMT and two fast photomultipliers with linear focusing dynode structure. Measurements were done for the 10x10x5 mm 3 LSO crystal. Presented values were corrected for the contribution of the reference detector. In our previous work [9] the plot of time jitter versus normalized time resolution with a 10x10x5 mm 3 LSO crystal

5 5 was presented for various types of fast photomultipliers. Such plot is presented in Fig. 8 together with the data for Hamamatsu H MOD tested in this study (diamond point). The square points represent measurements done with classic fast PMTs. The relation is linear and proportional to the time jitter. The circle points at the bottom represent fast PMTs with a different way of anode construction, in particular with the screening grid at the anode (XP20D0) [16] or with the design characteristic for metal channel photomultipliers. In this case also a linear dependence could be assumed but shifted alongside of the classic PMTs line. phe/mev comparing to 5900 phe/mev obtained for LSO. The second effect which deteriorates time resolution is a contribution of a time spread of the light collection inside the larger LYSO crystal. However, the value of about 275 ps measured for a single continuous LYSO detector (after correction for the reference BaF 2 detector) suggests that proposition of TOF PET system with monolithic crystal could establish a new approach for this kind of tomography. Fig. 9. The time spectrum measured with the 20x20x20 mm 3 LYSO crystal coupled to the Hamamatsu H MOD 16-channel photomultiplier. Fig. 8. The dependence of normalized time resolution versus time jitter for the tested H MOD phototube and other fast PMTs. Measurements were made for the 10x10x5 mm 3 LSO crystal. The results obtained for the tested Hamamatsu phototube are in very good agreement with the data collected for various kinds of classic fast photomultipliers. However, a last dynode and anode construction of a metal channel PMT should allow the further improvement of time resolution with lower fraction level just like it is observed in the case of XP20D0 equipped with the screening grid at the anode [16] and 4-channel PMT XP1485. The anode pulse of a metal channel PMT should not be affected by the prepulse resulting from electrons travelling between penultimate dynode and last dynode as it takes place in the linear focused PMTs. C. Time resolution with 20x20x20 mm 3 LYSO In the next part of the study the detector with 20x20x20 mm 3 LYSO continuous crystal was tested and optimization for the best time resolution was made. Fig. 9 presents the time spectrum measured with the 20x20x20 mm 3 LYSO crystal in relation to the BaF 2 reference detector. In this case the gate was created using all pixels. Again, the final timing result of 272 ps is an effect of the optimization of a threshold set at the LE discriminator. The plot of the time resolution versus fraction is presented in Fig. 10. The time resolution measured with the large LYSO crystal is by about 20% poorer than that of the small 10x10x5 mm 3 LSO because of a lower photoelectron number of only 4000 Fig. 10. The time resolution dependence on a LE threshold for the detector consisting of the H MOD photomultiplier and 20x20x20 mm 3 LYSO crystal. Presented values were corrected for the contribution of the reference detector. The time resolution of the tested detector together with a number of photoelectrons is collected in Table III. Also the data recorded with the same 20x20x20 mm 3 LYSO crystal and XP20D0 (QE=35%) photomultiplier were added for comparison. Note, a big difference in photoelectron number

6 6 despite the higher quantum efficiency of the Hamamatsu tube (QE=43%). TABLE III THE TIME RESOLUTION MEASURED WITH THE LYSO 20X20X20 COUPLED TO THE H8711 AND XP20D0 PHOTOMULTIPLIERS In Table IV a comparison between the continuous crystal detector and a single 4x4x20 mm 3 LSO pixel placed in various positions on a standard fast photomultipliers is presented. In the case of 16-channel Hamamatsu H8711 the energy gate was set using single anode signal corresponding to the chosen pixel. In the case of the linear focused PMTs the time resolution was measured in five positions over photocathode. The drawing of the crystal positions and pixels used in tests with H8711 PMT are showed in Fig. 11. timing measured for the three channels in the continuous crystal detector shows that homogeneity of a time resolution over a whole area of the detector is its big advantage. D. Positioning Application of a continuous crystal instead of a pixelated block detector needs modifications in methods of spatial identification of gammas interacting inside the scintillator. Presented work is focused mainly on time resolution optimization, however also simple tests on positioning were made. The main aim of the described measurements was finding the limitations of a continuous crystal detector introduced into the spatial resolution, not the exact algorithm. Tests of determination of interaction position of gamma inside the crystal were made with the LYSO crystal covered by black tape instead of Teflon. The simplest, center of mass algorithm was chosen, so this kind of wrapping simplified the calculations by removing the problem of the light reflected by walls of a scintillator. Two examples of counts distributions in pixels in case when the collimated 22 Na gamma source was pointed on one pixel are presented in Fig. 12. Fig. 11. The drawing of positions of the 4x4x20 mm 3 LSO on the photocathode in linear focused PMTs and pixels used in the H MOD photomultiplier. TABLE IV THE TIME RESOLUTION MEASURED WITH THE LYSO 20X20X20 COUPLED TO THE H8711 AND XP20D0 PHOTOMULTIPLIERS In the case of the H8711 photomultiplier the time resolution is identical in all the three measured positions. Also a number of photoelectrons in each pixel are the same. Situation is different in the case of the linear focused PMTs. Values in the middle of the standard fast photomultipliers are better in comparison to the one pixel in the tested Hamamatsu, but the results obtained on the rest four positions are much worse. The Fig. 12. Two examples of counts distributions in the detector consisting of H8711 PMT and 20x20x20 mm 3 LYSO crystal. In the first measurement (left) a collimated gamma source of 22 Na was pointed on pixel 6, in the second measurement (right) - on pixel 3. High level of background can be seen in both figures. In both cases the determination of the illuminated area is possible since the proper pixel has the highest number of counts. However a very high level of background is seen on both figures. This is a consequence of a composition of the LYSO crystal. The LSO and LYSO materials are lutetium-based scintillators which contains a radioactive isotope 176 Lu, a naturally occurring beta emitter. In the tested 20x20x20 mm 3 LYSO a natural background gives around 2300 c/s and is highly dominant over events resulting from collimated 511keV annihilation quanta. In Fig. 13 spectra of the 176 Lu background and 22 Na gamma source recorded with the tested LYSO are presented. In this case a non-collimated gamma source was used giving around 6000c/s. In the case of collimated source the counting rate was equal to about 3000 c/s and 511keV peak was not separated from the background. Second problem is the reflective layer of the scintillator walls. For the time of flight capability the amount of light collected from the crystal has to be maximized and directional

7 7 information cannot be lost so the black wrapping is not useful. Mirror like walls seems to be the best choice, since only this kind of wrapping gave a chance to keep the directional information. The Teflon walls will keep the light but the reflected photons will only create the statistical background. Fig. 13. The 511 kev spectrum recorded with a strong 22 Na source and LYSO scintillator. The final spectrum was obtained by subtraction of a background from 176 Lu. Another important aspect of the discussed detector is a fact that attenuation length for 511 kev in the LYSO crystal is equal to 1.2 cm. It means that the interaction of gammas occurs in whole volume of the crystal. Such a situation leads to inaccuracy in spatial resolution due to poor deep of interaction (DOI) determination. One of the possibilities of improving the DOI information is adding 4 APDs on the top of the crystal as proposed by Ojala [8]. During the presented study also a time resolution measurement with a single 4x4x20 mm 3 LSO placed on the one of the pixels in H8711 photomultiplier were performed. The result of 255 ps for a single detector and photoelectron number of 4000 phe/mev indicate that a pixelated block detector on a position sensitive photodetector could be an alternative solution for the future TOF PET. IV. CONCLUSIONS The presented results suggest that metal channel position sensitive PMT can assure very good timing capabilities of a TOF PET detector. The timing performance obtained with the small 10x10x5 mm 3 LSO crystal is as good as fast linear focused photomultipliers. The time resolution of 272ps measured for the single detector with monolithic LYSO scintillator corresponds to the 385 ps for the two detectors in coincidence. This value is around 200ps better then the present properties of commercially available TOF PET scanners. The continuous crystal detector with improved uniformity of the photocathode is characterized by a much better homogeneity of the timing response in comparison to detectors based on linear focused PMTs and pixelated block detector. Determination of a position of gamma interaction inside the continuous crystal wrapped with reflective layer is a highly difficult task. Analysis of the experimental data and requirements of spatial resolution in PET suggests that the pixelated block detector placed on a position sensitive PMT could be a more practical solution. REFERENCES [1] S. Surti, A. Kuhn, M.E. Werner, A.E. Perkins, J. Kolthammer and J.S. Karp, Performance of Philips Gemini TF PET/CT scanner with special consideration for its Time-of-Flight imaging capabilities, Journal of Nuclear Medicine, vol. 48, no. 3, pp , [2] M. Moszyński, M. Kapusta, A. Nassalski, T. Szczęśniak, D. Wolski, L. Eriksson, C. L. Melcher, New prospects for Time-of-Flight PET with LSO scintillators, IEEE Trans. Nucl. Sci., vol. 53, no. 5, pp , Oct [3] W. W. Moses, S. E. Derenzo, Prospects for Time-of-Flight PET using LSO scintillator, IEEE Trans. Nucl. Sci., vol. 46, pp , Jun [4] A. Kuhn, S. Surti, J. S. Karp, G. Muehllehner, F. M. Newcomer, R. Van Berg, Performance assessment of pixilated LaBr 3 detector modules for TOF PET, IEEE Trans. Nucl. Sci., vol. 53, no. 3, pp , Jun [5] T. Szczęśniak, M. Moszyński, A. Nassalski, P. Lavoute, A. G. Dehaine, A further study of timing with LSO on XP20D0 for TOF PET, Nucl. Instrum. Meth., vol. 54, no 5, pp , Oct [6] M. Kapusta, P. Lavoute, F. Lherbet, E. Rossignol, C. Moussant, F. Fouché, Breakthrough in quantum efficiency of bi-alkali photocathodes PMTs, 2007 IEEE Nuclear Science Symposium Conference Record, vol. 1, pp , Oct Nov. 3, [7] M.A. Spurrier, P. Szupryczyński, A.A. Carey, C.L. Melcher, Brightest and faster LSO, presented at IEEE 9 th International Conference on Inorganic Scintillators and Their Applications (SCINT), June 4-8, 2007, Winston-Salem, NC USA. [8] P. Ojala, A. Bousselham, L. Eriksson, A. Brahme, C. Bohm, Influence of sensor arrangements and scintillator crystal properties on the 3D precision of monolithic scintillation detectors in PET, 2005 IEEE Nuclear Science Symposium Conference Record, Vol. 5, pp , Oct [9] T. Szczęśniak, M. Moszyński, Ł. Świderski, A. Nassalski, P. Lavoute, and M. Kapusta, Fast Photomultipliers for TOF PET, IEEE Trans. Nucl. Sci., in press. [10] M. Bertolaccini, S. Cova, C. Bussolati, A technique for absolute measurements of the effective photoelectron yield in scintillation counters, Proc. Nucl. Electr. Symp. Versailles, France, [11] M. Moszyński, M. Kapusta, M. Mayhugh, D. Wolski, S.O. Flyckt, Absolute light output of scintillators, IEEE Trans. Nucl. Sci., vol. 44, no. 3, pp , Jun [12] M. Moszyński, J. Vacher, R. Odru, Application of HR400 microchannel plate photomultiplier to study the light pulse shape from fast and slow scintillators by means of the single photon method, Nucl. Instrum. Meth, vol. 204, no. 1, pp , Dec [13] M. Moszyński, M. Kapusta, D. Wolski, M. Balcerzyk, S.O. Flyckt, P. Lavoute, C. Marmonier, H. Mach, New fast photomultipliers with a screening grid at the anode, IEEE Trans. Nucl. Sci., vol. 51, pp , Aug [14] Z. Guzik, S. Borsuk, K. Traczyk, M. Płomiński, TUKAN-an 8K pulse height analyzer and multi-channel scaler with a PCI or a USB interface, IEEE Trans. Nucl. Sci., vol. 53, no. 1, pp , Feb [15] T. Szczęśniak, M. Moszyński, Ł. Świderski, A. Nassalski, A. Syntfeld- KaŜuch, A.G. Dehaine, and M. Kapusta, A comparative study of fast photomultipliers for timing experiments and TOF PET, presented at 2008 IEEE SORMA-WEST Conference, June 2-5, 2008, Berkeley, California, USA. [16] M. Moszyński, M. Kapusta, D. Wolski, M. Balcerzyk, S.O. Flyckt, P. Lavoute, C. Marmonier, H. Mach, New fast photomultipliers with a screening grid at the anode, IEEE Trans. Nucl. Sci., vol. 51, pp , Aug