A digital phoswich detector using time-over-threshold for depth of interaction in PET

Abstract

We present the performance of a digital phoswich positron emission tomography (PET) detector, composed by layers of pixilated scintillator arrays, read out by solid state light detectors and an application specific integrated circuit (ASIC). We investigated the use of integrated charge from the scintillation pulses along with time-over-threshold (ToT) to determine the layer of interaction (DOI) in the scintillator. Simulations were performed to assess the effectiveness of the ToT measurements for separating the scintillator events and identifying cross-layer-crystal-scatter (CLCS) events. These simulations indicate that ToT and charge integration from such a detector provide sufficient information to determine the layer of interaction. To demonstrate this in practice, we used a pair of prototype LYSO/BGO detectors. One detector consisted of a 19 × 19 array of 7 mm long LYSO crystals (1.36 mm pitch) coupled to a 16 × 16 array of 8 mm long BGO crystals (1.63 mm pitch). The other detector was similar except the LYSO crystal pitch was 1.63 mm. These detectors were coupled to an 8 × 8 multi-pixel photon counter mounted on a PETsys TOFPET2 ASIC. This high performance ASIC provided digital readout of the integrated charge and ToT from these detectors. We present a method to separate the events from the two scintillator layers using the ToT, and also investigate the performance of this detector. All the crystals within the proposed detector were clearly resolved, and the peak to valley ratio was 11.8 ± 4.0 and 10.1 ± 2.9 for the LYSO and BGO flood images. The measured energy resolution was 9.9% ± 1.3% and 28.5% ± 5.0% respectively for the LYSO and BGO crystals in the phoswich layers. The timing resolution between the LYSO-LYSO, LYSO-BGO and BGO-BGO coincidences was 468 ps, 1.33 ns and 2.14 ns respectively. Results show ToT can be used to identify the crystal layer where events occurred and also identify and reject the majority of CLCS events between layers.

Figure 1.

Left: example events modeled in the simulation for a phoswich 7 mm thick LYSO and 8 mm thick BGO crystal block (26 × 26 mm²). These are events that interact entirely in the top or bottom crystal layer, or those that interact in one layer and Compton scatter to deposit some energy in the other layer. Center: example LYSO and BGO pulses of the same energy from the pulse database. LYSO has much higher light yield and shorter decay time compared to BGO. Right: pulses from the database used for one forward scattered CLCS event. The LYSO (red) and BGO (blue) pulses are added together (black), and then subjected to a threshold.

Figure 2.

Experimental setup for the coincidence experiments: both detector blocks consist of a TOFPET2 ASIC attached to 64-channel Hamamatsu MPPCs. A 7 mm thick LYSO crystal array was stacked on an 8 mm thick BGO crystal array and then coupled via a light guide and optical grease to the MPPCs. A positron source was placed between the detector blocks.

Figure 3.

(a) 2D joint scatter plot of the charge integration versus ToT from the simulation. This simulation indicates LYSO, BGO and CLCS are well separated by using charge integration and ToT. The red dots are LYSO and the blue dots are BGO. The cyan dots represent CLCS events that strike LYSO first and the green are CLCS events striking the BGO first. (b) 2D plot of measured QDC values versus ToT as acquired from detector #1. The ToT for higher energy BGO events forms a vertical line at 510 ns (see text and figure 4). However, LYSO and BGO events can still be separated. The CLCS events have larger integrated charge than the BGO events and have a ToT of 510 ns—greater than any LYSO event. The separation between the BGO and CLCS events was plotted in the inset, with a different color scale for better visualization of the distribution characteristics. (c) Integrated charge spectra from the simulation. Red are LYSO events, blue are BGO events while cyan and green are the two types of CLCS events. (d) Measured QDC (charge integration) spectra formed from detector #1. The black trace is all events. The red is attributed to LYSO based on the separation shown in figure 3(b), while the blue trace is events attributed to BGO. The cyan trace represents those events determined to be CLCS. The relatively wide LYSO photo peak is due to the non-uniformity of light collection from the LYSO as it travels through the BGO crystals onto the MPPC array and not to saturation of the LYSO signal.

Figure 4.

ToT from GATE simulation. Red represents LYSO events, blue represents BGO events, cyan and green represent the two different categories of CLCS events. The black dashed arrow line indicates the maximum integration time possible with the current PETsys instrumentation. In the measurements, all events that generate pulses longer than 510 ns will have a ToT of 510 ns and appear as a single line in the plot of Charge Integration vs. ToT.

Figure 5.

(a) Flood image of LYSO layer events in detector #1 with a pitch of 1.36 mm. (b) Flood image of BGO layer events in the same detector with a pitch of 1.63 mm. (c) Flood image of CLCS events. (d)–(f) Horizontal profiles across one row of the LYSO, BGO and CLCS flood images. Events are identified from the same listmode file.

Figure 6.

Representative energy spectra of LYSO (a) and BGO (b) events from selected individual crystals, illustrating the energy resolution in detector #1. The average energy resolution for LYSO was 9.9% ± 1.3% and 28.5% ± 5.0% for BGO. Measured energy resolution and 511 keV photopeak position of LYSO (c), (e) and BGO (d), (f) events for each individual crystal.

Figure 7.

(a) LYSO–LYSO and (b) BGO–BGO coincidence timing resolution using three time pick-off methods for determining the time stamp of a gamma detection, based on the average (blue circle), the QDC weighted average (red triangle) and the QDC squared weighted average (green diamond) of the first n earliest time stamps. The black arrows indicate the optimized methods; (c) Timing spectra for different event types between all crystals in detector #1 and detector #2. Black: LYSO–LYSO coincidences with FWHM time resolution of 468 ps; Red solid: LYSO (det#2)–BGO (det#1) coincidences (Δt = 1257 ps); Red dashed: LYSO (det#1)–BGO (det#2) coincidences (Δt = 1420 ps), Blue: BGO–BGO coincidences (Δt = 2141 ps).