Compton PET: A Simulation Study for a PET Module with Novel Geometry and Machine Learning for Position Decoding

UMC-PET: a fast and flexible Monte Carlo PET simulator

Objective.The GPU-based Ultra-fast Monte Carlo positron emission tomography simulator (UMC-PET) incorporates the physics of the emission, transport and detection of radiation in PET scanners. It includes positron range, non-colinearity, scatter and attenuation, as well as detector response. The objective of this work is to present and validate UMC-PET as a a multi-purpose, accurate, fast and flexible PET simulator.Approach.We compared UMC-PET against PeneloPET, a well-validated MC PET simulator, both in preclinical and clinical scenarios. Different phantoms for scatter fraction (SF) assessment following NEMA protocols were simulated in a 6R-SuperArgus and a Biograph mMR scanner, comparing energy histograms, NEMA SF, and sensitivity for different energy windows. A comparison with real data reported in the literature on the Biograph scanner is also shown.Main results.NEMA SF and sensitivity estimated by UMC-PET where within few percent of PeneloPET predictions. The discrepancies can be attributed to small differences in the physics modeling. Running in a 11 GB GeForce RTX 2080 Ti GPU, UMC-PET is ∼1500 to ∼2000 times faster than PeneloPET executing in a single core Intel(R) Xeon(R) CPU W-2155 @ 3.30 GHz.Significance.UMC-PET employs a voxelized scheme for the scanner, patient adjacent objects (such as shieldings or the patient bed), and the activity distribution. This makes UMC-PET extremely flexible. Its high simulation speed allows applications such as MC scatter correction, faster SRM estimation for complex scanners, or even MC iterative image reconstruction.

Applications of machine learning and deep learning in SPECT and PET imaging: General overview, challenges and future prospects

The integration of positron emission tomography (PET) and single-photon emission computed tomography (SPECT) imaging techniques with machine learning (ML) algorithms, including deep learning (DL) models, is a promising approach. This integration enhances the precision and efficiency of current diagnostic and treatment strategies while offering invaluable insights into disease mechanisms. In this comprehensive review, we delve into the transformative impact of ML and DL in this domain. Firstly, a brief analysis is provided of how these algorithms have evolved and which are the most widely applied in this domain. Their different potential applications in nuclear imaging are then discussed, such as optimization of image adquisition or reconstruction, biomarkers identification, multimodal fusion and the development of diagnostic, prognostic, and disease progression evaluation systems. This is because they are able to analyse complex patterns and relationships within imaging data, as well as extracting quantitative and objective measures. Furthermore, we discuss the challenges in implementation, such as data standardization and limited sample sizes, and explore the clinical opportunities and future horizons, including data augmentation and explainable AI. Together, these factors are propelling the continuous advancement of more robust, transparent, and reliable systems.

The Past, Present, and Future Role of Artificial Intelligence in Ventilation/Perfusion Scintigraphy: A Systematic Review

Ventilation-perfusion (V/Q) lung scans constitute one of the oldest nuclear medicine procedures, remain one of the few studies performed in the acute setting, and are amongst the few performed in the emergency setting. V/Q studies have witnessed a long fluctuation in adoption rates in parallel to continuous advances in image processing and computer vision techniques. This review provides an overview on the status of artificial intelligence (AI) in V/Q scintigraphy. To clearly assess the past, current, and future role of AI in V/Q scans, we conducted a systematic Ovid MEDLINE(R) literature search from 1946 to August 5, 2022 in addition to a manual search. The literature was reviewed and summarized in terms of methodologies and results for the various applications of AI to V/Q scans. The PRISMA guidelines were followed. Thirty-one publications fulfilled our search criteria and were grouped into two distinct categories: (1) disease diagnosis/detection (N = 22, 71.0%) and (2) cross-modality image translation into V/Q images (N = 9, 29.0%). Studies on disease diagnosis and detection relied heavily on shallow artificial neural networks for acute pulmonary embolism (PE) diagnosis and were primarily published between the mid-1990s and early 2000s. Recent applications almost exclusively regard image translation tasks from CT to ventilation or perfusion images with modern algorithms, such as convolutional neural networks, and were published between 2019 and 2022. AI research in V/Q scintigraphy for acute PE diagnosis in the mid-90s to early 2000s yielded promising results but has since been largely neglected and thus have yet to benefit from today's state-of-the art machine-learning techniques, such as deep neural networks. Recently, the main application of AI for V/Q has shifted towards generating synthetic ventilation and perfusion images from CT. There is therefore considerable potential to expand and modernize the use of real V/Q studies with state-of-the-art deep learning approaches, especially for workflow optimization and PE detection at both acute and chronic stages. We discuss future challenges and potential directions to compensate for the lag in this domain and enhance the value of this traditional nuclear medicine scan.

Joint estimation of interaction position and energy deposition in semiconductor SPECT imaging sensors using fully connected neural network

Objective.Pixelated semiconductor detectors such as CdTe and CZT sensors suffer spatial resolution and spectral performance degradation induced by charge-sharing effects. It is critical to enhance the detector property through recovering the energy-deposition and position estimation.Approach.In this work, we proposed a fully-connected-neural-network-based charge-sharing reconstruction algorithm to correct the charge-loss and estimate the sub-pixel position for every multi-pixel charge-sharing event.Main results.Evident energy resolution improvement can be observed by comparing the spectrum produced by a simple charge-sharing addition method and the proposed energy correction methods. We also demonstrate that sub-pixel resolution can be achieved in projections obtained with a small pinhole collimator and an innovative micro-ring collimator.Significance.These achievements are crucial for multiple-tracer SPECT imaging applications, and for other semiconductor detector-based imaging modalities.

Hybrid PET/Compton-camera imaging: an imager for the next generation

Compton cameras can offer advantages over gamma cameras for some applications, since they are well suited for multitracer imaging and for imaging high-energy radiotracers, such as those employed in radionuclide therapy. While in conventional clinical settings state-of-the-art Compton cameras cannot compete with well-established methods such as PET and SPECT, there are specific scenarios in which they can constitute an advantageous alternative. The combination of PET and Compton imaging can benefit from the improved resolution and sensitivity of current PET technology and, at the same time, overcome PET limitations in the use of multiple radiotracers. Such a system can provide simultaneous assessment of different radiotracers under identical conditions and reduce errors associated with physical factors that can change between acquisitions. Advances are being made both in instrumentation developments combining PET and Compton cameras for multimodal or three-gamma imaging systems, and in image reconstruction, addressing the challenges imposed by the combination of the two modalities or the new techniques. This review article summarizes the advances made in Compton cameras for medical imaging and their combination with PET.

Experimental Evaluation of a 3-D CZT Imaging Spectrometer for Potential Use in Compton-Enhanced PET Imaging

We constructed a prototype positron emission tomography (PET) system and experimentally evaluated large-volume 3-D cadmium zinc telluride (CZT) detectors for potential use in Compton-enhanced PET imaging. The CZT spectrometer offers sub-0.5-mm spatial resolution, an ultrahigh energy resolution (~1% @ 511 keV), and the capability of detecting multiple gamma-ray interactions that simultaneously occurred. The system consists of four CZT detector panels with a detection area of around 4.4 cm × 4.4 cm. The distance between the front surfaces of the two opposite CZT detector panels is ~80 mm. This system allows us to detect coincident annihilation photons and Compton interactions inside the detectors and then, exploit Compton kinematics to predict the first Compton interaction site and reject chance coincidences. We have developed a numerical integration technique to model the near-field Compton response that incorporates Doppler broadening, detector's finite resolutions, and the distance between the first and second interactions. This method was used to effectively reject random and scattered coincidence events. In the preliminary imaging studies, we have used point sources, line sources, a custom-designed resolution phantom, and a commercial image quality (IQ) phantom to demonstrate an imaging resolution of approximately 0.75 mm in PET images, and Compton-based enhancement.

Simulation study of a brain PET scanner using TOF-DOI detectors equipped with first interaction position detection

Objective. The aim of this study is to evaluate the performance characteristics of a brain positron emission tomography (PET) scanner composed of four-layer independent read-out time-of-flight depth-of-interaction (TOF-DOI) detectors capable of first interaction position (FIP) detection, using Geant4 application for tomographic emission(GATE). This includes the spatial resolution, sensitivity, count rate capability, and reconstructed image quality.Approach. The proposed TOF-DOI PET detector comprises four layers of a 50 × 50 cerium-doped lutetium-yttrium oxyorthosilicate (LYSO:Ce) scintillator array with 1 mm pitch size, coupled to a 16 × 16 multi-pixel photon counter array with 3.0 mm × 3.0 mm photosensitive segments. Along the direction distant from the center field-of-view (FOV), the scintillator thickness of the four layers is 2.5, 3, 4, and 6 mm. The four layers were simulated with a 150 ps coincidence time resolution and the independent readout make the FIP detection capable. The spatial resolution and imaging performance were compared among the true-FIP, winner-takes-all (WTA) and front-layer FIP (FL-FIP) methods (FL-FIP selects the interaction position located on the front-most interaction layer in all the interaction layers). The National Electrical Manufacturers Association NU 2-2018 procedure was referred and modified to evaluate the performance of proposed scanner.Main results. In detector evaluation, the intrinsic spatial resolutions were 0.52 and 0.76 mm full width at half-maximum (FWHM) at 0° and 30° incidentγ-rays in the first layer pair, respectively. The reconstructed spatial resolution by the filter backprojection (FBP) achieved sub-millimeter FWHM on average over the whole FOV. The maximum true count rate was 207.6 kcps at 15 kBq ml^-1and the noise equivalent count rate (NECR_2R) was 54.7 kcps at 6.0 kBq ml^-1. Total sensitivity was 45.2 cps kBq^-1and 48.4 cps kBq^-1at the center and 10 cm off-center FOV, respectively. The TOF and DOI reconstructions significantly improved the image quality in the phantom studies. Moreover, the FL-FIP outperformed the conventional WTA method in terms of the spatial resolution and image quality.Significance. The proposed brain PET scanner could achieve sub-millimeter spatial resolution and high image quality with TOF and DOI reconstruction, which is meaningful to the clinical oncology research. Meanwhile, the comparison among the three positioning methods indicated that the FL-FIP decreased the image degradation caused by Compton scatter more than WTA.

Automated lesion detection of breast cancer in [18F] FDG PET/CT using a novel AI-Based workflow

Applications based on artificial intelligence (AI) and deep learning (DL) are rapidly being developed to assist in the detection and characterization of lesions on medical images. In this study, we developed and examined an image-processing workflow that incorporates both traditional image processing with AI technology and utilizes a standards-based approach for disease identification and quantitation to segment and classify tissue within a whole-body [18F]FDG PET/CT study.

METHODS

One hundred thirty baseline PET/CT studies from two multi-institutional preoperative clinical trials in early-stage breast cancer were semi-automatically segmented using techniques based on PERCIST v1.0 thresholds and the individual segmentations classified as to tissue type by an experienced nuclear medicine physician. These classifications were then used to train a convolutional neural network (CNN) to automatically accomplish the same tasks.

RESULTS

Our CNN-based workflow demonstrated Sensitivity at detecting disease (either primary lesion or lymphadenopathy) of 0.96 (95% CI [0.9, 1.0], 99% CI [0.87,1.00]), Specificity of 1.00 (95% CI [1.0,1.0], 99% CI [1.0,1.0]), DICE score of 0.94 (95% CI [0.89, 0.99], 99% CI [0.86, 1.00]), and Jaccard score of 0.89 (95% CI [0.80, 0.98], 99% CI [0.74, 1.00]).

CONCLUSION

This pilot work has demonstrated the ability of AI-based workflow using DL-CNNs to specifically identify breast cancer tissue as determined by [18F]FDG avidity in a PET/CT study. The high sensitivity and specificity of the network supports the idea that AI can be trained to recognize specific tissue signatures, both normal and disease, in molecular imaging studies using radiopharmaceuticals. Future work will explore the applicability of these techniques to other disease types and alternative radiotracers, as well as explore the accuracy of fully automated and quantitative detection and response assessment.

A Comparison of Neural Networks and Center of Gravity in Muon Hit Position Estimation

The performance of cosmic-ray tomography systems is largely determined by their tracking accuracy. With conventional scintillation detector technology, good precision can be achieved with a small pitch between the elements of the detector array. Improving the resolution implies increasing the number of read-out channels, which in turn increases the complexity and cost of the tracking detectors. As an alternative to that, a scintillation plate detector coupled with multiple silicon photomultipliers could be used as a technically simple solution. In this paper, we present a comparison between two deep-learning-based methods and a conventional Center of Gravity (CoG) algorithm, used to calculate cosmic-ray muon hit positions on the plate detector using the signals from the photomultipliers. In this study, we generated a dataset of muon hits on a detector plate using the Monte Carlo simulation toolkit GEANT4. We demonstrate that two deep-learning-based methods outperform the conventional CoG algorithm by a significant margin. Our proposed algorithm, Fully Connected Network, produces a 0.72 mm average error measured in Euclidean distance between the actual and predicted hit coordinates, showing great improvement in comparison with CoG, which yields 1.41 mm on the same dataset. Additionally, we investigated the effects of different sensor configurations on performance.

Artificial Intelligence and Cardiac PET/Computed Tomography Imaging

Artificial intelligence is an important technology, with rapidly expanding applications for cardiac PET. We review the common terminology, including methods for training and testing, which are fundamental to understanding artificial intelligence. Next, we highlight applications to improve image acquisition, reconstruction, and segmentation. Computed tomographic imaging is commonly acquired in conjunction with PET and various artificial intelligence methods have been applied, including methods to automatically extract anatomic information or generate synthetic attenuation images. Last, we describe methods to automate disease diagnosis or risk stratification. This summary highlights the current and future clinical applications of artificial intelligence to cardiovascular PET imaging.

Application of Artificial Intelligence in PET Instrumentation

Artificial intelligence (AI) has been widely used throughout medical imaging, including PET, for data correction, image reconstruction, and image processing tasks. However, there are number of opportunities for the application of AI in photon detector performance or the data collection process, such as to improve detector spatial resolution, time-of-flight information, or other PET system performance characteristics. This review outlines current topics, research highlights, and future directions of AI in PET instrumentation.

Simulation study of potential time-of-flight capabilities for a multilayer DOI-PET detector with an independent readout structure

A multilayer depth-of-interaction positron emission tomography (DOI-PET) detector with an independent readout structure has a potential advantage as a time-of-flight (TOF)-PET detector. The thin scintillator block of each detector layer can afford an improved coincidence time resolution (CTR) of ∼100 ps because the photon transfer time spread within the scintillator inherently decreases. To evaluate the potential TOF capabilities of a multilayer DOI-PET detector, which consists of thin layers of a cerium-doped lutetium-yttrium oxyorthosilicate (LYSO:Ce) scintillator coupled to a multi-pixel photon counter (MPPC) array, we examined the detector's CTR performance via Monte Carlo simulations. We used several types of scintillator structures: a monolithic plate, laser-processing array with 3.2 mm pitch, fine laser-processing array with 1.6 mm pitch, and pixelated array with 3.2 mm pitch, with 2, 4, 6, and 8 mm thickness values of a 25.6 mm × 25.6 mm scintillator cross-section. The MPPC array was composed of 3.0 mm × 3.0 mm photosensitive segments arranged in an 8 × 8 array. Here, we note that the CTR performance also significantly depends on the timing detection method, which generates a timing trigger signal for coincidence detection. Thus, we evaluated the CTRs for each scintillator structure by adopting four timing detection methods: using the total sum signal of 64 MPPC chips (T_sum), the maximum signal in the 64 MPPC chips (Max), the sum signal of a partial number of MPPC chips located at and in the vicinity of theγ-ray interaction position (P_sum), and the average of the timestamps generated at several MPPC chips (Ave). When using the T_sum for timing detection, the CTR full width at half-maximum values were ∼100 ps regardless of the scintillator structure. However, when using the Max signal approach, the CTRs of the monolithic plates, laser-processing arrays, and fine-pitch laser-processing arrays were drastically degraded with increasing thickness. On the other hand, the CTRs of the pixelated arrays exhibited almost no degradation. To improve the CTRs of the monolithic plate and the (fine-pitch) laser-processing array that exhibit a large light spread in the scintillator block, we applied the P_sum and Ave methods. The resulting CTRs significantly improved upon using P_sum; however, in the Ave approach the improvement effect disappeared when the thickness was <6 mm in case of our simulation.

Tomographic imaging with Compton PET modules: ideal case and first implementation

In our previous studies, we demonstrated that the Compton PET module, a layer structure PET detector with side readout, can provide high performance in terms of spatial/energy/timing resolution, as well as high gamma ray detection efficiency. In this study, we investigate how to translate the high performance of the detector module into good quality reconstructed tomographic images. This study is performed using GATE simulation, as well as with physical experiments. Similar detector geometry is used in the simulation and experiment: two identical 4-layer detector modules are placed with face to face distance of 56 mm. In the simulation study, each layer consists of a 1-mm-pitch pixelated crystal array. In the experimental study, each layer is a monolithic crystal, which is virtually binned into 1 mm² cells to group single events according to the gamma ray interaction locations. A customized Derenzo phantom was placed between the two detector modules. By rotating the phantom using a motorized rotary stage, data along lines of response (LORs) at different angles were collected for reconstructing the tomographic image. The same reconstruction algorithm was used for both simulation and experimental studies. The results demonstrate that the simulation study could resolve 0.8 mm rods while the experimental study was able to resolve 1.0 mm rods.

Gold nanoparticle detection and quantification in therapeutic MV beams via pair production

PURPOSE

We propose a new detection method of gold nanoparticles (AuNP) in therapeutic megavoltage (MV) x-ray beams by means of coincidence counting of annihilation photons following pair production in gold.

METHODS

The proposed MV x-ray induced positron emission (MVIPE) imaging technique is studied by radiation transport computations using MCNP6 (3D) and CEPXS/ONEDANT (1D) codes for two water phantoms: a 35 cm slab and a similarly sized cylinder, both having a 5 cm AuNP filled region in the center. MVIPE is compared to the standard x-ray fluorescence computed tomography (XFCT). MVIPE adopts MV x-ray sources (Co-60, 2 MV, 6 MV, 6 MV with closed MLC and 15 MV) and relies on the detection of 511 keV photon-pairs. XFCT uses kilovoltage sources (100 kVp, 120 kVp and 150 kVp) and imaging is characterized by analysis of k _α1,2 Au characteristic lines. Three levels of AuNP concentration were studied: 0.1%, 1% and 10% by weight.

RESULTS

Annihilation photons in the MVIPE technique originate both in the AuNP and in water along the x-ray beam path with significantly larger production in the AuNP-loaded region. MVIPE signal from AuNP is linearly increasing with AuNP concentration up to 10%wt, while XFCT signal reaches saturation due to self-absorption within AuNP. The production of annihilation photons is proportional to the MV source energy. MVIPE technique using a 15 MV pencil beam and 10 wt% AuNP detects about 4.5 × 10³ 511 keV-photons cm^-2 at 90° w/r to the incident beam per 10⁹ source photons cm^-2; 500 of these come from AuNP. In contrast, the XFCT technique using 150 kVp detects only about 100 k _α1-photons cm^-2 per 10⁹ source photons cm^-2.

CONCLUSIONS

In MVIPE, the number of annihilation photons produced for different MV-beam energies and AuNP concentrations is significantly greater than the k _α1 photons generated in XFCT. Coincidence counting in MVIPE allows to avoid collimation, which is a major limiting factor in XFCT. MVIPE challenges include the filtering of Compton scatter and annihilation photons originating in water.

Applications of artificial intelligence and deep learning in molecular imaging and radiotherapy

This brief review summarizes the major applications of artificial intelligence (AI), in particular deep learning approaches, in molecular imaging and radiation therapy research. To this end, the applications of artificial intelligence in five generic fields of molecular imaging and radiation therapy, including PET instrumentation design, PET image reconstruction quantification and segmentation, image denoising (low-dose imaging), radiation dosimetry and computer-aided diagnosis, and outcome prediction are discussed. This review sets out to cover briefly the fundamental concepts of AI and deep learning followed by a presentation of seminal achievements and the challenges facing their adoption in clinical setting.

Deep learning based methods for gamma ray interaction location estimation in monolithic scintillation crystal detectors

In this work, we explore deep learning based techniques using the information from mean detector response functions (MDRFs) as a new method to estimate gamma ray interaction location in monolithic scintillation crystal detectors. Compared with searching based methods, deep learning techniques do not require recording all the MDRF information once the prediction networks are trained, which means the memory cost could be significantly reduced. In addition, the event positioning process using deep learning techniques only requires running through the network once, without the need to do searching in the reference dataset. This could greatly speed up the positioning process for each event. We have designed and trained four different neural networks to estimate the gamma ray interaction location given the MDRF data. We have studied network structures consisting only of fully connected (FC) layers, as well as Conv neural networks (CNNs). In addition, we tried to use both regression and classification to generate the final prediction of the gamma ray interaction position. We evaluated the estimation accuracy, testing speed and memory cost (numbers of parameters) of different network architectures, and also compared them with the exhaustive search method. Our results indicate that deep learning based estimation methods with a well designed network structure can achieve a relative positioning error with respect to the ground truth determined by the exhaustive search method of below 1 mm in both x and y directions (depth information is not considered in this work), which would imply a very high performance positioning algorithm for practical monolithic scintillation crystal detectors. The deep learning network also achieves a testing speed that is more than 400 times faster than the exhaustive search method. With proper design of the network structure, the deep learning based positioning methods have the potential to save memory cost by a factor of up to 100.

Machine Learning in PET: from Photon Detection to Quantitative Image Reconstruction

Machine learning has found unique applications in nuclear medicine from photon detection to quantitative image reconstruction. While there have been impressive strides in detector development for time-of-flight positron emission tomography, most detectors still make use of simple signal processing methods to extract the time and position information from the detector signals. Now with the availability of fast waveform digitizers, machine learning techniques have been applied to estimate the position and arrival time of high-energy photons. In quantitative image reconstruction, machine learning has been used to estimate various corrections factors, including scattered events and attenuation images, as well as to reduce statistical noise in reconstructed images. Here machine learning either provides a faster alternative to an existing time-consuming computation, such as in the case of scatter estimation, or creates a data-driven approach to map an implicitly defined function, such as in the case of estimating the attenuation map for PET/MR scans. In this article, we will review the abovementioned applications of machine learning in nuclear medicine.