COMPTON SCATTER AND X-RAY CROSSTALK AND THE USE OF VERY THIN INTER-CRYSTAL SEPTA IN HIGH RESOLUTION PET DETECTORS C.S. Levin, M.P. Torni, S.R. Cherry, L.R. McDonld, E.J. Hoffmn Div. of Nucler Medicine & Biophysics nd Imging Sciences Division, Crump Institute for Biologicl Imging, UCLA School of Medicine Abstrct To improve sptil resolution, PET systems re being developed with finer detector elements. Unfortuntely, using smller crystl sies increses inter-crystl Compton sctter nd bismuth x-ry nd (to lesser degree) electron escpe crosstlk, cusing positioning errors tht led to degrdtion of imge contrst. We investigted the use of extremely thin (I 3 ym) led strips for pssive shielding of this inter-crystl crosstlk. Using nnihiltion photons nd smll (2 nd 3 mm wide) BGO crystls in coincidence, crosstlk studies were performed with either two smll djcent crystls (1-D) or one crystl inside volume of BGO (2-D). The frction of Compton scttered events from one crystl into n djcent one ws reduced, on verge, by fctor of 3.2 (2.2) in the 1-D experiment nd by fctor of 3. (2.1) in 2-D, with 3 (15) pm thick led strip in between the crystls nd 3-7 kev energy window in both crystls. We could not mesure reduction in bismuth x-ry crosstlk with the use of led sept becuse of the production of led x-rys. The width of the coincident point spred function ws not significntly different for the 1- nd 2-D studies, with or without the different thickness sept in plce. These results indicte tht intercrystl crosstlk does not ffect the positioning resolution. A simple insertion of very thin led strips my significntly reduce the inter-crystl scttering crosstlk of high resolution PET system, thereby ultimtely improving imge contrst, without introducing ded re. I. INTRODUCTION Current dvnces in PET detector hve resulted in nd re moving towrd the development of systems with smller detector elements designed to further immove sptil resolution. Unfortuntely utiliing smller crysti sies leds to higher frction of inter-crystl crosstlk due to Compton scttering. It hs been suggested tht inter-crystl scttering is the primry cuse of event mispositioning in high resolution block detectors used in commercil PET systems [ 11. Similr crosstlk effects occur in high resolution discretecrystl designs. For 5 3 mm wide crystls, we hve observed tht bismuth K-shell x-ry nd electron escpe cross tlk lso ply significnt role. These low energy effects result from the incomplete energy deposition of gmm ry interction in crystl. The l/e length of n 8 kev bismuth x-ry in BGO is pproximtely.2 mm. The verge rnge of 5 11 kev photoelectron in BGO is roughly.1 mm. Becuse 511 kev gmm ry interctions result in both Compton- nd photoelectron production, most electrons creted will hve energies much less thn 511 kev. In ddition, becuse n electron deposits its energy continuously in the crystl, the remining -783-3 18-X/96$5.1996 136 energy fter it escpes the crystl will be low, on verge. Thus, the electron escpe contribution of the low energy crosstlk will be smll in comprison to tht due to the 8 kev bismuth escpe x-ry. The conventionl method used to reduce the effects of Compton scttering is energy thresholding on the photopek. However, this method is somewht limited when the energy resolution is poor (s is the cse for systems using BGO s the scintilltor) nd ultimtely reduces the system sensitivity. We investigted the use of extremely thin led strips (I.3 mm) for pssive shieldin of inter-crystl crosstlk using mesurements with g2n source of 511 kev nnihiltion photons nd smll, squre BGO crystls (2 nd 3 mm width) in coincidence. In previous work [2-41, intercrystl sept ws for the most prt introduced in the context of improving the sptil resolution uniformity cross the imging re (cmer field of view) in conjunction with wedge shped or tpered crystls. In those investigtions, lrger crystls (2 4 mm) nd sept (2 1 mm) widths were utilied. The implementtion of these reltively thick sept necessrily reduced the totl geometricl efficiency (ring sensitivity) by reducing the crystl pcking frction. It ws concluded tht the benefits did not outweigh the disdvntges. We re interested in studying the effect of introducing very fine intercrystl sept (1.3 mm) between smll (2-3 mm) crystls. This fine sept cn be, for exmple, inserted in between crystls in current PET detector designs [ 11 without reducing the pcking frction (introducing ded re): These designs lredy hve 25-4ym intercrystl spces for reflector nd the mount of reflector cn be reduced with little loss in light output. In our mesurements we will use different energy thresholds to tke into ccount the possible contribution of intercrystl crosstlk of different energies. II. MATERIALS AND METHODS A. One-Dimensionl Experiment Experiments were performed using two different thicknesses of led inter-crystl sept,.15 nd.3 mm, plced in between two crystls, for both 2x2~1 nd 3x3~3 mm3 crystl sies. Figure 1 shows top-down schemtic drwing of the one-dimensionl experimentl set-up. Two crystls of the sme sie were coupled sidewys (for better light collection efficiency) with opticl grese to two seprte 1.5" dimeter RCA PMTs. Three lyers of thin teflon tpe covered the non-detecting fces of ech crystl nd blck tpe surrounded the crystl borders on ech tube. The two PMTs + the crystls (D1 nd D2) were then pressed together fce on so tht the sides of the crystls were in line with ech other with or without the thin led sept in plce. There ws no mesurble light lek between crystls. A third BGO crystl Authoried licensed use limited to: Stnford University. Downloded on June 1,21 t 22:19:48 UTC from IEEE Xplore. Restrictions pply.
mode on Mcintosh computer through GPIlB interfce to CAMAC vi LhView (Ntionl Instruments). A 22N source (-2 pci), positioned pproximtely 8 cm from D3, ws moved perpendiculr to the line between D3 nd D1 in.5 mm steps nd the CPSF ws mesured with or without the thin led strips in plce. Source sie nd positron rnge in the plstic source holder re roughly 1 nd.8 mm, 2x2~3 mm BGO crystl of interest -5, 9 mm BGO A-/ - 7 PMT-D1 =1 - _ over D1 (the crystl of interest). Becuse we re interested in the CPSF we hve loosely extended this definition to include those source positions where it is, by geometry, more likely tht the gmm ry is intercting in D2 first before scttering into DI (D1 must be hit for trigger). The sctter frction reduction fctor is defined s the rtio of sctter frction without to tht with inter-crystl sept in plcle between the two crystls. These definitions represent n verge over the number of counts in the mesured energy spectrum of D1 nd D2 bove given threshold. For exmple, if 3 kev threshold is chosen in both detectors, then for given event t most one detector cn hve >3 kev deposited within (511 kev mximum) nd count in our definition of sctter frction. Fig. 2. D1,D2 pbrtion of the 2-D experiment, Two dditionl effects were studied with the 2-D experiment: the contribution to the inter-crystl cross-tlk due to the 8 kev Bismuth K-shell x-ry, nd, possibly, electron escpe. We note however, tht electron escpe will effect crosstlk, t best, only within.1 mm perimeter of crystl (19% of the cross-sectionl re of 2 mm wide crystl), wheres, on verge, n 8 kev bismuth x-ry will ply role within the outer.2 mm of the sme crystl (36% of the crosssectionl rel of 2 mm wide BGO crystl). Thus, the 8 bismuth x-ry escpe will ply dominnt role in low energy crosstlk (see Section I). With 1.5 kev threshold set on DI 137 Authoried licensed use limited to: Stnford University. Downloded on June 1,21 t 22:19:48 UTC from IEEE Xplore. Restrictions pply.
the totl number of counts in D2 for 57-97 nd 3-125 kev energy windows were mesured for 5 collimted source positions over D1. The electron escpe events deposit continuum of energies weighted towrd lower energies. The second energy window will thus encompss lrger proportion of electron escpe events. We will use the crosstlk frction mesured in this ltter window to roughly ccess the combined contribution of escpe x-ry nd electron. For the smller (57-97 kev) energy window, we will ssume the counts re due only to bismuth x-ry crosstlk. The rtio of the number of events in observed in D2 within these energy windows to tht in D1 is the frction of events tht will be mispositioned. 111. RESULTS A. 1 -D Experiment: Coincident Point Spred Function nd Energy Spectr: D1 NO SEPTA WITH SEPTA Figure 3 shows the CPSF mesured in D1 nd crosstlk into D2 for 3 kev lower nd 7 kev upper energy threshold in ech of D1 nd D2 (2x2~1"~) with nd without.3 mm led strip in plce for the 1-D experiment. The FWHM of the CPSF of D 1 (1.54 mm) ws mesured to be only 3% nrrower (only one side hs the djcent crystl) with the.3 mm sept in plce. Note the different pek positions of the CPSF of D1 nd D2. The mximum of the CPSF of D1 occurs where the source position is in line with the center of the crystl. Figure 4 shows the corresponding full energy spectr (no threshold) mesured in the crystls with nd without the.3 mm sept. 1 14- SIDE WITHOUT 8- CRYSTAL 2 cn!- 5 6-3 kev ENERGY THRESHOLD I 4v - SIDE 2 -.-.-.... -... -----_ D1 NOSEPTA - D2 NO SEPTA D1 WITH SEPTA D2 WITH SEPTA WITH CRYSTAL 2 _.._...--...-..... : I,,,, I,, Wy4-:,, I,,,, I,,,, w-,..5 1 1.5 2 2.5 3 3.5 Fig. 3. Mesured coincident point spred function (CPSF) of D1 for 1 mm dimeter 22N point source with nd without.3 mm thick led strip in plce between D1 nd D2 (see Fig.1) for 3 kev energy threshold. B. 1 -D Experiment: Sctter Frction: Figure 5 shows the sctter frction mesured in D2 for 3 kev lower nd 7 kev upper energy threshold in ech of the 2 x 2 x 1 mm3 crystls, with nd without the.3 mm thick sept in plce s function of source position. We note tht the definition of sctter frction given in Section I1 might be most pproprite when the source position is collimted over D1, but this quntity is lso shown for positions outside of D1 (position 1.5 is the center of Dl)..3 1.25-1.2 c U.15- LT W +.1- Q cn.5- -NO LEAD SEPTA -WITH.3 mm INTERCRYSTAL SEPTA 2 X 2 X 1 mm CRYSTALS 3 kev THRESHOLD -- ----- I --/ _c-- o -,, ~. l, ~,, l,,,, l,, ~, l,,, ~,,, ~,.5 1 1.5 2 2.5 3 Figure 6 shows the sctter frction reduction fctor chieved with the.3 mm led strip in plce s function of vrious source positions nd energy thresholds. For 3 kev energy threshold, we found tht the frction of scttered events into n djcent crystl ws reduced, on verge, by fctor of 3.2 (2.1) when there ws.3 (.15) mm thick led strip in plce. In generl, this reduction fctor depended on energy threshold (see energy spectr in Fig. 2) nd source position. With the led strips in plce between lrger 3 x 3 x 3 mm3 crystls, we found similr results for the sctter frction reduction fctor: the lrger crystls hd, within lo%, pproximtely the sme sctter frction bove 3 kev energy threshold for points directly over the D1 crystl (but hd higher count rte). In Figure 7 we show different wy of representing the sctter dt. Let dl nd d2 be the digitied pulse heights per event of the signls from D1 nd D2, respectively. Figure 7 I 138 Authoried licensed use limited to: Stnford University. Downloded on June 1,21 t 22:19:48 UTC from IEEE Xplore. Restrictions pply.
ENERGYTHRESHOLD ----- ZOO kev - 3 kev /-- d fl --.e c 2 x 2 X 1 mm CRYSTAL 1..........................5 1 1.5 2 2.5 3 Fig. 6. Sctter frction reduction fctor mesured using the.3 mm thick intercrystl led sept s function of source position for vrious energy thresholds (2 x 2 x 1 mm3 BGO crystl). The dip seen is pproximtely t the position of the sept. 1.2 io4.- I io4 v) 8 5 6 4 HISTOGRAM OF RATIO=(dl -d2)/(dl +d2) NO SEPTA NO THRESHOLD shows histogrm of rtio (dl-d2)/(dl+d2) per event with nd without the.3 mm sept present. This rtio is the D nlogy of th.t used to position events in multiplexed (light shring) PET block detector designs [l]. We see the devition from the correct position ( rtio of 1) lessens with the sept in plce. This representtion is especilly relevnt to multiplexed crystl rry dlesigns where ll of the light (or chrge) creted from ll the crystls in the rry per event is used for crystl identifiction (positioning) nd energy thresholding of tht event. From this figure one cn better ccess the contribution of crosstlk to mispositioning errors in n rry, without the presence of light shring effects [l]. C. 2- E,xperiment: Coincident Point Spred Function: Figure 8 shows the CPSF mesured in D1 nd crosstlk component seen in D2 for 3 kev lower nd 7 kev upper energy threshold in both detectors with nd without the two sies of led sept (.15 nd.3 mm) in plce. There is no significnt difference in the CPSF FWHM or FWTM between the 1- nd 2-D experiments, even though there is much greter BGO sctter volume present nd nerly 4 times the sctter frction. Furthermore, there is no chnge in these quntities withi the introduction of sept. These results will be discussed in Section IV. - D2 (NO SEPTA).. -. -. D1 (.3 mm SEPTA) D2 (.3 mm SEPTA) 2 v) 1.2 lod -1.5-1 -.5.5 I I.5 RATIO 3 kev Energy Threshold I io4 RATIO=(dl -d2)/(dl +d2) WITH.3 MM SEPTA - NO THRESHOLD 8- U) 3 6-4- 2- - (B) -1.5-1 -.5.5 1 1.5 RATIO Fig. 7. Plot of the difference between the signls rnesured in D1 nd D2 normlied by the totl signl for () no sept nd (b) with.3 mm sept for the D experiment (Fig. 1). The presence of the sept reduces the errors in positioning (the correct position is rtio of 1). All energies were used. Fig. 8. Coincident PSF mesured for D1 nd crosstlk mesured in D2 s function of source position for no sept nd with.15 mm nd.3 mm thick led sept completely surrounding D1 (3 kev lower nd 7 kev upper energy thresholds in D1 nd D2). Upper 3 curves: response of D1, Lower 3 curves: D2. D. 2- Experiment: Sctter Frction Figure 9 displys the frctionl crosstlk (top) nd crosstlk reduction (bottom) for 15 nd 3 kev thresholds mesured in W2 with no sept nd with.15 nd.3 mm sept in plce. For source positions directly over D1, since the energy thresholds re high in this figure, the x-ry escpe pek of bismuth is not contributing. The sctter frction is improved, on verge, by fctor of 3. (2.1) with the.3 (1.5) mm sept in plce. 139 Authoried licensed use limited to: Stnford University. Downloded on June 1,21 t 22:19:48 UTC from IEEE Xplore. Restrictions pply.
o ---c 3 kev THRESH. (NO SEPTA) - B 3 kev THRESH. (.3 mm SEPTA) ---+--..-*... 15 kev THRESH. (NO SEPTA) - v 15 kev THRESH. (.3 mm SEPTA) 3 kev THRESH. (.1 5 mm SEPTA) 'F- P - 15 kev THRESH. (.15 mm SEPTA) 9.8- q.6- LL W.4- C o.2-1.4 2-D EXPERIMENT 2.35; 2.31 U y.25.2 U)...154 5.5 x -NO SEPTA (3-1 25 kev),+ -.3 mm SEPTA (3-125 kev) --.--.15 mm SEPTA (3-125 kev) NO SEPTA (57-97 kev) -..--.3 mm SEPTA (57-97 kev) *....15 mm SEPTA (57-97 kev).._. --- y- =E +-:-.-*-=-= 4- ---- ~--.-..~... :~,.~-.--~----~ ----- *---* 2-D EXPERIMENT 5 1 15 2 25 11.5 12 12.5 13 13.5 14 Fig. 1. X-ry nd electron escpe crosstlk into D2 from D1 vs. source position for 2-D experiment for two energy windows. In both cses lower energy threshold of 15 kev ws set in D1. The presence of led sept slightly increses crosstlk t low energies due to led 8 kev led x-rys produced. IV. DISCUSSION 2.54 SEPTA ' M - D -.1 5 mm (3 kev) --.-,3 mm (15 kev) --+---.1 5 mm (1 5 kev) :... I 11.5 12 12.5 13 13.5 14 1 POSITION (mm) Fig. 9. Top: Sctter frction vs. source position for no sept,.15 nd.3 mm led sept. The dt shows mesurements with thresholds of both 15 nd 3 kev in D1 nd D2. Bottom: Corresponding sctter reduction fctor. E. Cross-Tlk Due to X-Ry nd Electron Escpe Figure 1 displys the crosstlk from D1 into D2 for lower energy window in D2 centered bout the 77 kev bismuth x-ry with nd without sept in plce. Bismuth x-ry crosstlk into D2 (with no sept) is roughly 1% of events registered in D1 for 5% energy window centered bout 77 kev (57-97 kev) nd lower energy threshold of 15 kev set in D1. For 95 kev window bout 77 kev the estimtion of the combined x-ry nd electron escpe crosstlk is 25%. We observed n increse in x-ry crosstlk when.15 nd.3 mm sept is used. This is due to the fct tht gmm rys intercting with led lso produce 8 kev led x-rys. There will be thus competing effects of gmm nd bismuth x-ry (nd electron) ttenution in the led (crosstlk reduction) nd led x-ry production. In this experiment neither crosstlk (Compton sctter, bismuth x-ry nd electron escpe) nor intercrystl sept ppered to ffect the intrinsic resolution s mesured by the width of the coincident point spred function (CPSF). The CPSF width is mesure of the intrinsic resolution of n rry. In this experiment, this quntity is determined by the crystl sie, the sie of the point source nd the positron rnge in the source holder. The crosstlk sources do cuse positioning errors which reduce imge contrst. Our results indicte tht with the simple insertion of very thin led strips between 53 mm wide crystls, the crosstlk due to Compton sctter cn be reduced significntly. Whether n improvement in imge contrst would be worth the slight cost increse is open for debte. V. REFERENCES [l] SR Cherry, MP Torni, CS Levin, S Siege1 nd EJ Hoffmn. A Comprison of PET Detector Modules Employing Rectngulr nd Round Photomultiplier Tubes. IEEE Trns. Nucl. Sci. 42-4, August 1995: 164-8. [2] NA Keller nd LR Lupton. PET Detector Ring Aperture Function Clcultions using Monte Crlo Techniques. IEEE Trns. Nucl. Sci. NS-3-1, Februry 1983, 676-8. [3] JM Roney, CJ Thompson. Detector Identifiction with Four BGO Crystls on Dul PMT. ZEEE Trns. Nucl. Sci. NS-31: 122-7, 1984. [4] R Lecomte, D Schmitt nd G Lmoureux. Geometry Study of High Resolution PET Detection System Using Smll Detectors. ZEEE Trns. Nucl. Sci. NS-31-1, Februry 1984: 556-61. 14 Authoried licensed use limited to: Stnford University. Downloded on June 1,21 t 22:19:48 UTC from IEEE Xplore. Restrictions pply.