Unsupervised Method-Based TOF-PET Signal Reshaping

Time-of-flight (TOF) information can improve the imaging quality of positron emission tomography (PET) camera by calculating the original positron-electron annihilation position. In this work, we measured the preliminary TOF resolution of non-TOF-optimized LYSO arrays for TOF PET using our PMT-Quadrant-Sharing (PQS) technology which has higher light collection efficiency (no external light guide). We investigated digital signal processing methods that may achieve better time resolution. Digital oscilloscope Tektronix DPO7054 was used to capture the waveforms. Three timing methods, Linear Fitting (LF) to the pulse rising time, Constant Fraction Discriminator (CFD) and Leading edge Discriminator (LD), were studied. A new Time-Energy (T-E) correction using Multivariate Least Mean Square Fit was applied to minimize time walk caused by the variation of signal shape. The time resolution as a function of the depth of interaction was also studied. The average block-to-block coincident timing resolution of 13×13 (4×4×20 mm3 detector pixel) array with fast PMT R9779 (51-mm diameter) was measured to be 432 ps (FWHM). Another 13×13 crystal array with ultrahigh spatial resolution 1.4×1.4×10 mm3 detector pixels coupling to small 19-mm diameter poor-timing XP1912 PMTs was also tested and achieved 551 ps. This study shows that our regular PQS detector blocks (without optimization for TOF) have good TOF time resolution suitable for TOF PET systems, in addition to higher spatial resolution and reducing PMT cost.

Objective. In positron emission tomography (PET) reconstruction, the integration of time-of-flight (TOF) information, known as TOF-PET, has been a major research focus. Compared to traditional reconstruction methods, the introduction of TOF enhances the signal-to-noise ratio of images. Precision in TOF is measured by full width at half maximum (FWHM) and the offset from ground truth, referred to as coincidence time resolution (CTR) and bias. Approach. This study proposes a network combining transformer and convolutional neural network (CNN) to utilize TOF information from detector waveforms, using event waveform pairs as inputs. This approach integrates the global self-attention mechanism of Transformer, which focuses on temporal relationships, with the local receptive field of CNN. The combination of global and local information allows the network to assign greater weight to the rising edges of waveforms, thereby extracting valuable temporal information for precise TOF predictions. Experiments were conducted using lutetium yttrium oxyorthosilicate (LYSO) scintillators and silicon photomultiplier (SiPM) detectors. The network was trained and tested using the waveform datasets after cropping. Main results. Compared to the constant fraction discriminator (CFD), CNN, CNN with attention, long short-term memory (LSTM) and Transformer, our network achieved an average CTR of 189 ps, reducing it by 82 ps (more than 30%), 13 ps (6.4%), 12 ps (6.0%), 16 ps (7.8%) and 9 ps (4.6%), respectively. Additionally, a reduction of 10.3, 8.7, 6.7 and 4 ps in average bias was achieved compared to CNN, CNN with attention, LSTM and Transformer. Significance. This work demonstrates the potential of applying the Transformer for PET TOF estimation using real experimental data. Through the integration of both CNN and Transformer with local and global attention, it achieves optimal performance, thereby presenting a novel direction for future research in this field.

The timing information carried by the PMT waveform is important when reconstructing the particle trajectory, especially in the application of time-of-flight detectors. The traditionally used timing methods, including the leading edge discrimination (LED) and the constant fraction discrimination (CFD), are easily realized in the circuit. Since the timing point of the signal should be determined by the whole signal waveform, the CFD and LED method cannot reach the best time resolution. With the development of fast analog-to-digital converter which makes the whole waveform available, more intelligent methods are developed to realize more precise timing. Former studies show that PMT-waveform-analysis model based on convolutional neural network can improve the coincidence time resolution (CTR) of a pair of PMTs by nearly 20%, but the limit of number of labels directly affects the CNN timing performance. In this paper, a new method based on CNN model is introduced for the timing of a pair of Cherenkov-detection MCP-PMTs. The results show a great improvement compared with the CFD method.

Silicon photomultipliers (SiPMs) are now widely used for positron emission tomography (PET) applications because of their high gain and low noise characteristics. The PET image quality has been improved with the advancement of time-of-flight (TOF) and depth-of-interaction (DOI) measurement techniques. For brain-dedicated PET systems, both TOF and DOI information are beneficial for enhancing the reconstructed PET image quality. In a previous study, we proposed SiPM-based dual-ended readout PET detectors that used a mean time method to achieve coincidence timing resolution (CTR) of 349 ps and DOI resolution of 2.9 mm. However, the coincidence timing resolution (CTR) was worse than 300 ps since the crystal surface and the reflector type were not optimized. This study aimed at investigating the optimal crystal surface treatment and the reflector material to achieve a sub-200 ps CTR and sub-3 mm DOI resolution with a dual-ended readout PET detector using an LYSO crystal (2.9 × 2.9 × 20 mm3). The scintillation light inside the LYSO crystal was read out by two SiPMs using the dual-ended readout method. The CTR and DOI resolution were measured with two different crystal surfaces (polished and saw-cut) and three different reflector material scenarios of ESR without grease (i.e., air coupling), ESR with optical grease and Teflon. We digitized the timing and energy signals by using a V775N TDC module (35 ps bit−1) and V965 QDC module, respectively. The combination of the saw-cut LYSO crystal and the ESR with air coupling resulted in the best CTR (188 ± 32 ps) and DOI resolution (2.9 ± 0.2 mm) with the dual-ended readout configuration. We concluded the dual-ended readout method in combination with the saw-cut crystal and the ESR reflector with air coupling can provide a sub-200 ps CTR and sub-3.0 mm DOI resolution simultaneously.

To increase contrast in positron emission tomography (PET) images, researchers are investigating detectors that reach below the nanosecond timing resolution. This allows a tight coincidence window which reduces random coincidence counts in the acquired data, as well as to include time-of-flight information into the reconstruction algorithms. With this goal in mind, single photon avalanche diode (SPAD) arrays have been under study for their excellent timing performances. However, their spurious dark counts can blur the start of PET signals where timing information is the most precise and create false starts in the acquisition system, introducing dead time. To minimize these problems in digital SPAD systems using a single time to digital converter (TDC) per PET channel, dark count discriminator circuits are required to reduce timing errors and increase the triggering efficiency in presence of dark counts. This paper compares the performance of a probabilistic and a novel delay line based dark count discriminator. Simulations of a SPAD array investigate the impact of dark counts on triggering efficiency and coincidence timing. Results show that the probabilistic discriminator provides excellent event recovery with regard to dark counts at the cost of some coincidence timing resolution. On the other hand, the delay line discriminator maintains the peak coincidence timing resolution but does not provide as much efficiency at high dark count rate levels.

Scintillating crystal coupled to a silicon photomultiplier SiPM is a typical detection solution in many modern radionuclide imaging systems such as positron emission tomography (PET). Coincidence time resolution (CTR) and energy resolution (ER) are two key performance parameters in the field of time-of-flight PET (ToF-PET). Both CTR and ER are system level parameters related to crystal choice dimension, speed and noise of the read-out electronics, calibration and analysis algorithms, and SiPM performance. In this work, the effect of SiPM photon detection efficiency PDE, single photon time resolution SPTR, and correlated and uncorrelated noise on CTR and ER were studied. For this purpose, analytical calculation and Monte Carlo simulation tools for CTR and ER were developed and tested for the onsemi MicroFJ-60035 SiPM device at different operating bias voltages coupled with LYSO and BGO crystals. The obtained analytical and simulation results were compared with experimental measurements performed with a 22Na source and 3 × 3 × 3 mm3 LSYO crystal wrapped in Teflon reflector and coupled to SiPMs using optical grease. Good agreement between measurements, simulation, and analytical calculation was found for ER at all device working range, and the best ER of ∼ 9% for 511 keV photopeak was found for the 3 × 3 × 3 mm3 LYSO crystal. The CTR was simulated for LYSO crystals with different lengths from 3 mm up to 30 mm and varies from 110 ps (3 mm crystal) up to 170 ps (30 mm crystal). It was compared to experimental measurements performed for 3, 5 and 20 mm crystal length, where CTR was measured from 95 ps (3 mm crystal) up to 140 ps (20 mm crystal). A difference from 5 ps up to 20 ps was found between Monte Carlo simulation and CTR measurements depending on the crystal length. The analytical calculation shows on average ∼ 20 ps worse CTR than experimental measurements. The systematic discrepancy between experimental and simulated CTR values might be related to a difference in light collection efficiency and level of electronics noise used in calculations as well as SiPM PDE angular dependency, while the discrepancy between experimental and analytical CTR might be related to read-out electronics noise which seems to have too high effect on analytical calculation.

A silicon photomultiplier (SiPM) with electronics integrated on cell level has been developed by Philips Digital Photon Counting. The device delivers a digital signal of the detected photon counts and their time stamp, making it a potential candidate for positron emission tomography (PET) applications. Several operational parameters of the specifically developed acquisition protocol can be adjusted to optimize the photon detection. In this work, the combination of five different parameters (trigger scheme, validation scheme, cell inhibition, temperature and excess bias voltage) is analyzed. Their impact on both the intrinsic as well as the PET-oriented sensor's performance is studied when coupled to two different PET candidate scintillators, GAGG and LYSO (2 × 2 × 6 mm3). The results show that SiPM intrinsic properties such as breakdown voltage temperature coefficient (20 mV/K) and optical crosstalk (20%) are similar to state-of-the-art analog devices. The main differences are induced by the logic of the acquisition sequence and its parameters. The sensor's dark-count-rate (DCR) is 770 kHz/mm2 at 24°C and 100% active cells. It can be reduced through cell inhibition and lower temperatures (ca. 2 orders of magnitude at 0°C and 20% cell inhibition). DCR reduction is necessary to avoid acquiring dark-count-triggered and validated events, causing loss of detection sensitivity. The typical time fraction spent with these events is 42.5% (GAGG) and 35.5% (LYSO). Increasing percentages of cell inhibition affect the photodetection efficiency and with it the energy resolution and the coincidence time resolution (CTR). At 5.6 °C and 10% cell inhibition, the measured energy resolution is 11.9% and 13.5% (FWHM, saturation corrected) and a FWHM CTR of 458 ps and 177 ps is achieved, for GAGG and LYSO respectively. With the implemented setup, the optimum configuration for PET, in terms of sensitivity, energy resolution and CTR, is trigger scheme 1, validation scheme 8, 10% cell inhibition and 3.0 V excess bias. Cooling the sensor to 0°C considerably improves the device's performance.

Achieving fast coincidence timing resolution (CTR) is an important issue in clinical time-of-flight positron emission tomography (TOF-PET) to improve the reconstructed image quality. One of the major factors affecting the CTR is the crystal surface treatment, which is often parameterized as surface roughness. However, previous studies on the crystal surface treatment optimization had two limitations of crystal-by-crystal variation and worse CTR over 200 ps. Here, we report the effects of the crystal surface treatment on the performance of a 20 mm long fast-LGSO crystal based TOF detector by tracking the same crystals in the sub-180 ps CTR regime. The light collection efficiency (LCE), energy resolution (ER) and CTR of the TOF detector were evaluated with six different crystal surface treatments of chemically polished (C.P), C.P half side roughened (1/2S) treatment, and then the C.P one side roughened (1S) treatment, mechanically polished (M.P) treatment, M.P 1/2S treatment, and M.P 1S treatment. The four lateral surfaces of each crystal were wrapped by using enhanced specular reflector film while the top surface was covered by using Teflon tape. The bottom surface of the crystal was optically coupled to a silicon photomultiplier. The timing and energy signals were extracted by using a custom-made high-frequency readout circuit, and then digitized by using a waveform digitizer. All the experimental conditions were same except the crystal surface treatment. Among the six different crystal surface treatments, the M.P 1S would be the optimal crystal surface treatment which balanced enhancements in the CTR (165 ± 3 ps) and ER (10.5 ± 0.5%). Unlike the M.P 1S, the C.P 1S did not enhance the CTR and ER. Hence, the C.P without roughening would be the second-best optimal crystal surface treatment which balanced the CTR (169 ± 3 ps) and ER (10.5 ± 0.5%).

Time resolution in time-of-flight positron emission tomography (TOF-PET) has improved significantly over the last decade due to advancements in scintillation materials, photodetectors, and readout electronics, which has increased the signal-to-noise ratio (SNR) compared to conventional PET. Silicon photomultipliers (SiPMs) in TOF-PET detectors are often operated at high bias voltage to improve the time performance at the expense of increasing signal noise. SiPM noise, both correlated and uncorrelated, can cause baseline fluctuations, leading to time-walk effects when a leading edge trigger strategy is used, and thus limiting timing performance. We examined the effect of SiPM baseline fluctuations using the FastIC ASIC, a scalable multichannel readout for fast timing applications. We flagged noisy events by using a comparator signal triggered by dark counts before the actual scintillation event. We tested different classification and correction methods with scintillating crystals and Cherenkov radiators, coupled to analog SiPMs from Broadcom (NUV-MT) and Hamamatsu Photonics. We reduced the coincidence time resolution (CTR) in bismuth germanate 2×2×3 mm3 (BGO) crystals from 410 ± 10 ps to 388 ± 10 ps FWHM (5 %) by correcting the time-walk on the noisy events. We measured an improvement from 107 ±2 ps to 93.5 ± 0.6 ps (11 %) for LYSO 2×2×3 mm3 crystals by filtering the noisy events. An improvement of 9 % on the CTR of the EJ232 plastic scintillator was also achieved by filtering noisy events, reducing it from 82.2 ± 0.5 ps to 75 ± 1 ps. This study presents a scalable method for flagging undesired events in a full TOF-PET system and discusses the impact of SiPM noise on the FastIC readout.

Positron emission tomography (PET) with high resolution and high sensitivity is desirable for detecting cancers and neurological diseases. In this work, a depth-of-interaction (DOI)-time-of-flight (TOF) PET has been attempted to achieve both high spatial and high timing resolution. Dual-ended readout is a simple technique that can provide excellent timing and DOI resolutions and consistent signal arrival times, regardless of the DOI position along the scintillation crystal, as a mean-time method is used. A dual-ended readout DOI-TOF PET module consisting of a 6 × 6 array of 2 × 2 × 20 mm3 saw-cut cerium-doped lutetium-yttrium oxyorthosilicate (LYSO) crystals is constructed. Both ends of the LYSO crystal array are optically coupled to a multi-pixel photon counter (MPPC) with 4 × 4 channels. The sixteen MPPC outputs are reduced to four position signals using a charge division circuit (CDC) board, and the timing signal is extracted from the common cathode of the MPPC. The four position signals from the MPPC are digitized by a DRS4-based high-speed waveform digitizer with a sampling rate of 5 GSa/s. A 22Na source is placed in front of a reference detector and at the side of the dual-ended readout DOI-TOF PET module in five steps of 2 mm, 6 mm, 10 mm, 14 mm, and 18 mm to measure the DOI, coincidence timing resolutions (CTRs), and mean-times. The full-width-half-maximums (FWHMs) of DOI resolutions and CTRs varied from 3.0 mm to 3.8 mm, with an average of 3.5 mm, and from 333 ps to 367 ps, with an average of 349 ps, respectively. The average of the slopes of the mean-time versus DOI position, for the 36 crystals, was −0.60 ± 1.68 ps/nm, which was consistent with the null value. The dual-ended readout DOI-TOF PET module based on the mean-time method produced both good DOI and CTRs, and consistent signal arrival times. The found solution would be the most advantageous in the small aperture PET systems, such as those for brain and breast imaging.

The design of a positron emission tomography (PET) scanner is specially challenging since it should not compromise high spatial resolution, high sensitivity, high count-rate capability, and good energy and time resolution. The geometrical design of the system alongside the characteristics of the individual PET detector modules contributes to the overall performance of the scanner. The detector performance is mainly influenced by the characteristics of the photo-detector and the scintillation crystal. Although silicon photomultipliers (SiPMs) have already proven to be promising photo-detectors for PET, their performance is highly influenced by micro-cell structure and production technology. Therefore, five types of SiPMs produced by KETEK with an active area size of 1.2 × 1.2 mm2 were characterized in this study. The SiPMs differed in the production technology and had micro-cell sizes of 25, 50, 75, and 100 μm. Performance of the SiPMs was evaluated in terms of their breakdown voltage, temperature sensitivity, dark count rate, and correlated noise probability. Subsequently, energy resolution and coincidence time resolution (CTR) of the SiPMs were measured with five types of crystals, including two Ce:LYSO, two Ce:GAGG, and one Pr:LuAG. Two crystals with a geometry of 1.5 × 1.5 × 6 mm3 were available from each type. The best CTR achieved was ∼ 240 ps, which was obtained with the Ce:LYSO crystals coupled to the 50 μm SiPM produced with the trench technology. The best energy resolution for the 511 keV photo-peak was ∼ 11% and was obtained with the same SiPM coupled to the Ce:GAGG crystals.

Radiation detection used in positron emission tomography (PET) exploit the timing information to remove background noise and refine the position measurement through time-of-flight (TOF) information. In PET, very fine time resolution (in the order of 10 ps FWHM) would not only improve contrast in the image, but would also enable real-time image reconstruction without iterative or back-projected algorithms. The current performance limitations will be pushed off through the optimization of faster light emission mechanisms (prompts photons), after which the burden of timing resolution will fall to the readout optoelectronics. Digital SPAD arrays offer compelling possibilities to minimize timing jitter in these future detector systems such per-cell timestamps granularity and per-cell configuration parameters, providing a highly flexible signal processing environment. However, processing hundreds of timestamps per detection event places a toll on the real-time processing, which increases rapidly with embedded channel count. Furthermore, if the processing is sent to an external device such as an FPGA, the bandwidth and related power requirements also increase. The simulation flow presented here offers perspectives on how many time to digital converters (TDC) would be required to reach the 10 ps FWHM CTR range for PET. Using this information, designers can estimate the compromises between timing performance, bandwidth requirements, data transmission, power consumption and real-time dataflow processing in the DAQ at the chip and system level. With a standard 1.1 × 1.1 × 3.0 mm3 LYSO scintillator, the coincidence timing resolution (CTR) changed by less than 3% within the range of 4 to 484 implemented TDCs for evaluated system conditions. On the other hand, an LYSO-based photonic crystal with 2.5% prompt emission rate needs a detector with at least 36 TDCs to reach within 3% CTR of an equivalent array with one TDC per SPAD. This gives significant insights on how this change of crystal material will affect system real time requirements for future detector design.

TOF(Time of flight) LiDAR obtains the timing difference of the emitted pulse to the target and returned signal. TDC(Timing discriminator) measures the timing difference with specific threshold voltage level called LED (Leading edge discriminator) method. The timing difference varies since the timing at the threshold voltage is different depending on the intensity of the returned signal. This timing error is called walk error. To compensate the walk error, CFD(Constant fraction discriminator) and ZCD(Zero crossing discriminator) methods are used. These methods utilize the peak point. However when the signal is saturated, these methods bring significant error since the peak point can't be found. It is important to prevent the signal to be saturated for high performance of distance accuracy and wide input dynamic range. Walk error compensation methods composed of current sensing circuit and MOSFET will be presented. As a result, high performance of distance accuracy is achieved with the wide input dynamic range.

Silicon photomultipliers (SiPMs) have high gain, low bias voltage, excellent timing properties and are insensitive to magnetic fields, which have become the best readout choice for advanced positron emission tomography (PET) systems, such as PET/MRI instruments and time-of-flight (TOF) PET imaging. Novel Device Laboratory (NDL) has been developing an unique SiPM technology that uses the bulk resistors in the epitaxial layer as quenching resistors (EQR SiPM); it features small micro cells with high fill factor, fast response to even a single photon, and simple fabrication technology. This paper reports the results to verify that EQR SiPM is promising in the applications of advanced PET imaging. For the EQR SiPM with active area of 3 × 3 mm2 and 90000 individual cells, the dark count rate of 6.3 MHz at room temperature, a peak PDE of 34% for 420 nm photons and an intrinsic single photon timing resolution (SPTR) of 81 ps at 9 V over-voltage were obtained. By employing a pair of EQR SiPMs 1-to-1 coupled with 2.84 × 2.84× 6 mm3 LYSO crystals, the energy resolution (ER) of ∼10.1% and the coincidence timing resolution (CTR) of ∼195 ps (FWHM) were demonstrated; by employing a pair of EQR SiPMs 1-to-1 coupled with 2.84 × 2.84× 10 mm3 LYSO crystals, the ER of ∼9.8% and the CTR of ∼226 ps (FWHM) were demonstrated.

Sens-A-Ray characteristics with variations in beam quality