Time-of-flight positron emission tomography (TOF-PET) detectors exhibiting multiple coincidence time resolution (CTR) components, such as those induced by the mixing of Cherenkov and scintillation photons, have attracted increasing attention. However, to fully exploit the latent potential of multi-kernel TOF-PET, new iterative image reconstruction methods are required. In this study, assuming that the events are labeled with the appropriate kernels, we propose an alternating direction method of multipliers (ADMM) for multi-kernel TOF-PET reconstruction, termed TOF-decomp ADMM. As the convergence speed of the TOF-PET log-likelihood depends on the CTR, the proposed method splits the fast- and slow-CTR log-likelihood terms and optimizes them separately under a constraint. This strategy explicitly balances the contributions of fast- and slow-CTR components and enables early stopping at iterations that yield improved contrast-noise trade-offs compared with conventional methods. We validated the proposed method using brain and image quality phantom simulations, demonstrating improved contrast-noise characteristics from a more stabilized convergence. By addressing the convergence imbalance inherent to multi-kernel TOF-PET, this work establishes a framework for exploiting the timing information available in emerging detector technologies.

Deep learning is increasingly transforming medical imaging by enabling more accurate data interpretation and reconstruction from complex detector signals. In positron emission tomography (PET), accurate localization of photon interactions within detectors is crucial for improving image resolution and diagnostic value. This work focuses on semi-monolithic scintillation detectors, which balance pixelated and monolithic designs, and applies deep learning methods that exploit scintillation light patterns across photosensors to improve interaction positioning. We evaluated how different neural network architectures and combinations of input signals affect positioning accuracy using experimental data from two types of detector arrays: 1) a 1 & times;8 module used in the IMAS total-body scanner and 2) a 1 & times;16 module from a brain-dedicated PET, both coupled to 64-channel photosensor matrices. We compared multilayer perceptron and convolutional neural network, using either reduced 16-channel inputs (obtained via row and column summation) or the full 64-channel configuration, along with energy, time, and engineered features. Results show that deeper architectures and richer inputs improve positioning performance-especially in the depth-of-interaction direction, with gains of around 20%-by better exploiting light-sharing effects between slabs. Finally, we introduced a deep-learning-based signal demultiplexing approach which accurately reconstructs full 64-channel signals from reduced 16-channel measurements with structural similarity index measure above 0.98. This enables the combination of simplified hardware design crucial for data throughput together with the benefits of higher resolution positioning when using 64 signals. This work shows how deep learning, when combined with multiple signal inputs and the learned recovery of full signals from multiplexed data, can enhance the performance of PET instrumentation.

Hanbury Brown and Twiss interferometry was a milestone experiment that transformed our understanding of the nature of light. Originally demonstrated in 1956 to measure the radii of stars through photon-correlation detection, it later became a cornerstone of modern quantum optics. Here we connect Hanbury Brown and Twiss interferometry to the physics of scintillation, the process of spontaneous light emission upon excitation by high-energy particles, such as X-rays. By revealing the underlying photon bunching in the scintillation process, we use the photon correlations $${g}^{\left(2\right)}\left(\tau \right)$$ to quantify the intrinsic light emission properties of the scintillator, specifically the emission time and the number of optical photons emitted per X-ray photon. This approach provides a characterization method that we benchmark on a wide gamut of scintillators, including several rare-earth-doped (and undoped) oxide single-crystal scintillators and perovskite nanocrystals, thereby showing the dependence of their properties on temperature and X-ray flux. Our method is particularly important for nano- and microscale scintillators, whose properties are challenging to quantify by conventional means. We extract the scintillation properties even in quantum-dot superlattices of only a few hundreds of nanometres and observed strong photon bunching ( $${g}^{\left(2\right)}\left(0\right) > 50$$ ). Our research paves the way for the broader use of methods from quantum optics for studying materials with complex optical properties in extreme regions of the electromagnetic spectrum. Photon bunching is observed in the scintillation process upon excitation by X-rays. The photon intensity correlation function is measured by using Hanbury Brown and Twiss interferometry to extract the light yield and emission lifetime in scintillator materials.

Abstract Background Breast cancer causes the largest number of cancer-related deaths among women worldwide. With the aim of improving Positron Emission Tomography (PET) technology for accurate breast cancer diagnosis and staging, we propose a system design based on monolithic crystals with inherent Depth of Interaction (DOI) capabilities and an innovative edgeless detector ring. This approach eliminates the physical gaps between PET detectors, improving the system detection efficiency while potentially enhancing the image quality since edge effects are reduced. We have developed a dedicated breast PET system prototype (DeepBreast) to show the feasibility of this design. The system is composed of 14 curved LYSO monolithic scintillators of 12.5 mm thickness glued side-by-side with a high-refractive index compound. The useful transaxial and axial Field of View (FOV) of the system are 160 mm and 50 mm, respectively. A Neural Network technique was used for the x- and y- photon impact position estimation. The impact DOI and energy values were determined using the Voronoi calibration methodology. An initial experimental evaluation of the DeepBreast system has been performed inspired by the NEMA protocols for whole-body and small-animals PET scanners. Results A nearly flat spatial resolution as a function of radial position was obtained, which indicates the DOI capability of the system to mitigate parallax errors. An average spatial resolution of 1.9 ± 0.1 mm, 1.9 ± 0.1 mm and 1.7 ± 0.1 mm FWHM was achieved at the center of the axial FOV for the radial, tangential, and axial directions, respectively. A maximum sensitivity value of 2% was measured at the center of the FOV. The noise equivalent count rate peak reached 15 kcps at 13.4 MBq. Moreover, percent contrast values of 27.9%, 28.8%, 56.8%, 72.5%, 87.2% and 84.2% were achieved for 4.5 mm, 6 mm, 9 mm, 12 mm, 15 mm and 20 mm cylinders of a larger dedicated IQ phantom, respectively. Conclusions The initial experimental results demonstrate the feasibility of the DeepBreast as an innovative PET scanner for breast cancer imaging.

Semi-monolithic detectors, a hybrid configuration combining the benefits of pixelated arrays and monolithic blocks, present a compelling and cost-effective solution for positron emission tomography (PET) scanners with both time-of-flight (TOF) and depth-of-interaction (DOI) capabilities. In this work, we evaluate four LYSO-based semi-monolithic arrays with various surface treatments, read out with the PETsys TOFPET2 ASIC, to identify the optimal configuration for a novel brain PET scanner. The chosen array, featuring ESR on all surfaces except for the black-painted lateral pixelated ones, achieved 15.9 ± 0.6 % energy resolution and 253 ± 15 ps detector time resolution (DTR). neural network with multilayer perceptron architectures were used to estimate the annihilation photon impact position, yielding average accuracies of 3.7 ± 1.1 mm and 2.6 ± 0.7 mm (FWHM) along the DOI and monolithic directions, respectively. The comparative analysis of the four arrays also prompted an investigation into light sharing in semi-monolithic detectors, supported by a GATE-based simulation framework which was designed to complement the experimental results and confirm the observed trends in time resolution. By refining the detector design based on semi-monolithic geometry and optimized surface crystal treatment to enhance positioning accuracy, this study contributes to the development of a next-generation brain PET scanner, with competitive performance but at a moderate cost.

Objective.A key challenge in PET systems is collecting large amounts of data with the most accurate information-time, energy, and position-to produce high-resolution images while limiting the number of channels to reduce costs and improve data collection efficiency. The new ultra-high-performance brain (UHB) scanner under development aims to tackle this issue, using a semi-monolithic detector that combines pixelated arrays and monolithic designs, along with signal multiplexing techniques.Approach.We assessed the time, energy, and positioning performance of the multiplexing circuit (summing signals along rows and columns) and compared it to the standard readout, both using TOFPET2 ASIC.Main Results.While time resolution worsens by about 15%, energy and positioning resolution-more crucial in small diameter scanners-are unaffected by signal summation. Overall, a pair of detector modules (2 × 2 arrays each) features an energy resolution of 16.9 ± 1.3% and 405 ± 29 ps coincidence time resolution. Positioning accuracy-estimated using multilayer perceptron neural network-is 1.9 ±0.4 mm and 3.0 ±0.7 mm along the monolithic and depth-of-interaction direction, respectively.Significance.This study demonstrates that this channel reduction readout effectively maintains high performance while allowing for reduced costs and enhanced scalability.

Scintillators are materials converting high-energy radiation into optical light, essential in a range of technologies such as medical imaging systems and security scanners. Scintillator development and optimization have remained limited by the complexity of their underlying physics, involving stochastic cascades of electron-electron, electron-phonon, and electron-photon interactions. Such processes are typically modeled by non-differentiable Monte Carlo simulations, limiting the applicability of machine learning for scintillator development. Here we present a physics-informed neural network that learns the scintillation cascade process from the incident high-energy particle to photon emission, substantially accelerating scintillator design and optimization. Combining this neural network with photonic simulations enables end-to-end differentiable optimization of the scintillator geometry. This allows us to optimize for arbitrary figures of merit, such as specific target emission patterns.. We demonstrate the concept and characterize it relative to previous approaches by inverse design of nanophotonic scintillators for X-ray imaging.

Most preclinical PET scanners are based on pixelated detectors without Depth of Interaction (DOI) capabilities, which is crucial to correct for parallax errors. Semi-monolithic crystals have the potential to combine the timing capabilities of pixelated crystals and the 3D positioning accuracy of monolithic scintillators. In this work, we present a preclinical PET prototype consisting of 2 rings defining an inner diameter of 106 mm and an axial length of 52 mm. Each ring contains 14 arrays of 1 × 22 LYSO slabs of 0.97 mm × 25.6 mm × 12 mm, coupled to 8 × 8 SiPMs arrays. The sensitivity, spatial resolution, count rate performance and image quality were studied using the NEMA NU 4-2008 protocol. All images were reconstructed using the MLEM algorithm with 0.5 mm voxel size, including DOI information and normalization correction. A mean spatial resolution for all measured positions and across the three directions (axial, transaxial and radial) of 1.61 ± 0.19 mm was obtained at the center of the FOV. A peak sensitivity of 3.5% was obtained at the center of the scanner, for an energy window (EW) between 358 keV -664 keV. The noise equivalent count rate peak reached 106.9 kcps for an activity of 11.7 MBq using the mouse-sized phantom, the same EW window and a time coincidence window of 10 ns. For the image quality phantom, contrast recovery coefficients of 0.20, 0.62, 0.75, 0.81 and 0.85 were found for the 1, 2, 3, 4 and 5 mm rods, respectively. Spill-over-ratio values of 0.10 for air-filled and 0.23 for water-filled cylinders were measured. Also, according to the Rayleigh criterion, 1.5, 1.2 and 1 mm hot spots of a micro-Derenzo phantom were well distinguished.

Abstract Background The renewed interest in BGO scintillators for TOF-PET is driven by the improved Cherenkov photon detection with new blue-sensitive SiPMs. However, the slower scintillation light from BGO causes significant time walk with leading edge discrimination (LED), which degrades the coincidence time resolution (CTR). To address this, a time walk correction (TWC) can be done by using the rise time measured with a second threshold. Deep learning, particularly convolutional neural networks (CNNs), can also enhance CTR by training with digitized waveforms. It remains to be explored how timing estimation methods utilizing one (LED), two (TWC), or multiple (CNN) waveform data points compare in CTR performance of BGO scintillators. Results In this work, we compare classical experimental timing estimation methods (LED, TWC) with a CNN-based method using the signals from BGO crystals read out by NUV-HD-MT SiPMs and high-frequency electronics. For $${2 \times 2 \times 3}\,\hbox {mm}^{3}$$ 2 × 2 × 3 mm 3 crystals, implementing TWC results in a CTR of 129 ± 2 ps FWHM, while employing the CNN yields 115 ± 2 ps FWHM, marking improvements of 18 % and 26 %, respectively, relative to the standard LED estimator. For $${2 \times 2 \times 20}\,\hbox {mm}^{3}$$ 2 × 2 × 20 mm 3 crystals, both methods yield similar CTR (around 240 ps FWHM), offering a $$\sim$$ ∼ 15 % gain over LED. The CNN, however, exhibits better tail suppression in the coincidence time distribution. Conclusions The higher complexity of waveform digitization needed for CNNs could potentially be mitigated by adopting a simpler two-threshold approach, which appears to currently capture most of the essential information for improving CTR in longer BGO crystals. Other innovative deep learning models and training strategies may nonetheless contribute further in a near future to harnessing increasingly discernible timing features in TOF-PET detector signals.

PET is the modality of choice for studying the biochemistry and physiology of the human brain in vivo, although its low spatial resolution, attributable to inherent physical and technical constraints, has limited its ability to resolve small cerebral structures. Here, we report PET images of the human brain obtained at a volumetric resolution of nearly 2 µL. Methods: A dedicated ultra-high-resolution (UHR) PET scanner featuring 1.2-mm true pixelated detectors was developed to achieve microvolumetric spatial resolution. A partially assembled UHR PET scanner with an axial field of view of 143 mm was used to obtain 18F-FDG PET images of the human brain. Patients who had a clinical PET/CT scan subsequently underwent UHR PET on completion of their medical examination. UHR PET images were reconstructed using a 3-dimensional ordered-subset expectation maximization iterative algorithm with analytic coincidence function modeling. Reconstructed images were normalized to the Montreal Neurological Institute 152 brain template and analyzed using atlases for region identification. Relative SUVs to the cerebellum were extracted for selected small brain structures. Results: All major brain regions were easily identifiable in UHR PET images, including details of the primary motor and somatosensory cortices, caudate nucleus, putamen, thalamus, inferior colliculi, and dentate nuclei. Notably, regions rarely seen so distinctly with 18F-FDG PET, such as the subthalamic areas and brainstem nuclei, were successfully resolved, suggesting that UHR PET has the potential to provide enhanced quantification of these tiny cerebral structures. This was further confirmed by higher SUV ratios in the UHR PET images compared with the PET/CT images. The UHR PET image of 1 patient revealed hypermetabolic foci in the cerebellum that were not discernible on the PET/CT and MR images. Conclusion: UHR PET images of the human brain at nearly 2-µL volumetric spatial resolution were obtained. Previously indistinguishable, small, highly relevant regions of the brain were resolved, paving the way for more accurate and detailed studies with the potential for greater insight in neuropsychiatry, neurooncology, and neurodegenerative diseases.

One of the limiting factors of spatial resolution in positron emission tomography (PET) imaging is annihilation photon acollinearity (APA). For whole-body PET scanners, APA induces a blur ranging from 1.7 to 2.2 mm FWHM. For long axial field-of-view (FOV) scanners, this range increases even more, depending on the maximum ring difference. It was previously shown that perfect time-of-flight (TOF) resolution sharpens the APA-induced blur by altering its expected Gaussian shape into a profile resembling a 1/r function, thereby reducing its contribution to spatial resolution loss. This suggests that the conventional theoretical limit of PET spatial resolution could be overcome if sufficient TOF resolution can be achieved. However, the requirements to achieve an observable gain in spatial resolution have yet to be explored. We propose an investigation of these requirements for whole-body and long axial FOV scanners, in terms of TOF resolution and count statistics. Using a fictive 81-cm diameter scanner with 2-mm wide detectors, we show that ultrafast TOF resolution-13 ps FWHM-enables an observable gain in spatial resolution for a range of count statistics. In addition, we show that lower TOF resolutions (i.e., higher TOF values of 27 or 67 ps) could mitigate APA for the oblique tubes of response of long axial FOV systems subjected to larger APA blurring. This last observation is of particular interest as it suggests that the nonstationary nature of spatial resolution in PET imaging can be further mitigated when such TOF precision is achieved.

This study focuses on advancing metascintillators to break the 100 ps barrier and approach the 10 ps target. We exploit nanophotonic features, specifically the Purcell effect, to shape and enhance the scintillation properties of the first-generation metascintillator. We demonstrate that a faster emission is achievable along with a more efficient conversion efficiency. This results in a coincidence time resolution improved by a factor of 1.6, crucial for TOF-PET applications.

Hanbury Brown and Twiss (HBT) interferometry is a milestone experiment that transformed our understanding of the nature of light. The concept was demonstrated in 1956 to measure the radii of stars through photon coincidence detection. This form of coincidence detection later became a cornerstone of modern quantum optics. Here we connect HBT interferometry to the physics of scintillation, the process of spontaneous light emission upon excitation by high-energy particles, such as x-rays. Our work reveals intrinsic photon bunching in the scintillation process, which we utilize to elucidate its underlying light emission mechanisms. Specifically, g^((2) ) (τ) enables the quantitative extraction of scintillation lifetime and light yield, showing their dependence on temperature and X-ray flux as well. This approach provides a characterization method that we benchmark on a wide gamut of scintillators, including rare-earth-doped garnets and perovskite nanocrystals. Our method is particularly important for nano- and micro-scale scintillators, whose properties are challenging to quantify by conventional means: We extract the scintillation properties in perovskite nanocrystals of only a few hundreds of nanometers, observing strong photon bunching (g^((2) ) (0)>50). Our research paves the way for broader use of photon-coincidence measurement and methods from quantum optics in studying materials with complex optical properties in extremes regions of the electromagnetic spectrum.

Acollinearity of annihilation photons is a source of spatial blur in PET imaging that increases with the scanner radius. It was demonstrated that acollinearity follows a Gaussian distribution. This statement can refer to two concepts: the amplitude of the acollinearity angle or the angular deviation of the annihilation photons relative to the collinear case. Since the former is the partial integral of the latter, an error of interpretation could have significant repercussions. Previous works that have studied the effect of acollinearity in PET imaging have made the assumption that the angular deviation is Gaussian, which is in agreement with experimental studies. However, we show that in GATE, a PET simulation software, acollinearity is simulated as the former. This not only changes the shape of the spatial blur induced in the image space but also significantly reduces its effect on spatial resolution, e.g., from 2.1 mm FWHM to 0.4 mm FWHM for a PET scanner ≈ 80 cm in diameter. This underestimation is shared with other PET simulation softwares. Results obtained with these simulators would underestimate, sometimes severely, the blur induced by acollinearity. We propose an approach to simulate acollinearity that follows the latter interpretation and show that it would be adequate for PET imaging.

The production of prompt photons providing high photon time densities is a promising avenue to reach ultrahigh coincidence time resolution (CTR) in time-of-flight positron emission tomography. Detectors producing prompt photons are receiving high interest experimentally, ignited by past exploratory theoretical studies that have anchored some guiding principles. Here, we aim to consolidate and extend the foundations for the analytical modeling of prompt generating detectors. We extend the current models to a larger range of prompt emission kinetics where more stringent requirements on the prompt photon yield rapidly emerge as a limiting factor. Lower bound and estimator evaluations are investigated with different underlying models, notably by merging or keeping separate the prompt and scintillation photon populations. We further show the potential benefits of knowing the proportion of prompt photons within a detection set to improve the CTR by mitigating the detrimental effect of population (prompt versus scintillation) mixing. Taking into account the fluctuations on the average number of detected prompt photons in the model reveals a limited influence when prompt photons are accompanied by fast scintillation (e.g., LSO:Ce:Ca) but a more significant effect when accompanied by slower scintillation (e.g., bismuth germanate). Establishing performance characteristics and limitations of prompt generating detectors is paramount to gauging and targeting the best possible timing capabilities they can offer.

Acolinearity of the annihilation photons observed in Positron Emission Tomography (PET) is described as following a Gaussian distribution. However, it is never explicitly said if it refers to the amplitude of the acolinearity angle or its 2D distribution relative to the case without acolinearity (herein defined as the acolinearity deviation). Since the former is obtained by integrating the latter, a wrong interpretation would lead to very different results. The paper of Shibuya et al. (2007), differs from the previous studies since it is based on the precise measurement of the energy of the annihilation photons. They also show that acolinearity follows a Gaussian distribution in the context of PET. However, their notation, which relies on being on the plane where the two annihilation photons travel, could mean that their observation refers to the amplitude of the acolinearity angle. If that understanding is correct, it would mean that acolinearity deviation follows a 2D Gaussian distribution divided by the norm of its argument. Thus, we revisited the proof presented in Shibuya et al. (2007) by using an explicit description of the acolinearity in the 3D unit sphere.

Acolinearity of the annihilation photons observed in Positron Emission Tomography (PET) is described as following a Gaussian distribution. However, it is never explicitly said if it refers to the amplitude of the acolinearity angle or its 2D distribution relative to the case without acolinearity (herein defined as the acolinearity deviation). Since the former is obtained by integrating the latter, a wrong interpretation would lead to very different results. The paper of Shibuya et al. (2007), differs from the previous studies since it is based on the precise measurement of the energy of the annihilation photons. They also show that acolinearity follows a Gaussian distribution in the context of PET. However, their notation, which relies on being on the plane where the two annihilation photons travel, could mean that their observation refers to the amplitude of the acolinearity angle. If that understanding is correct, it would mean that acolinearity deviation follows a 2D Gaussian distribution divided by the norm of its argument. Thus, we revisited the proof presented in Shibuya et al. (2007) by using an explicit description of the acolinearity in the 3D unit sphere.