Artificial intelligence guided enhancement of digital PET: scans as fast as CT?

PURPOSE: Both digital positron emission tomography (PET) detector technologies and artificial intelligence based image post-reconstruction methods allow to reduce the PET acquisition time while maintaining diagnostic quality. The aim of this study was to acquire ultra-low-count fluorodeoxyglucose (FDG) ExtremePET images on a digital PET/computed tomography (CT) scanner at an acquisition time comparable to a CT scan and to generate synthetic full-dose PET images using an artificial neural network. METHODS: This is a prospective, single-arm, single-center phase I/II imaging study. A total of 587 patients were included. For each patient, a standard and an ultra-low-count FDG PET/CT scan (whole-body acquisition time about 30 s) were acquired. A modified pix2pixHD deep-learning network was trained employing 387 data sets as training and 200 as test cohort. Three models (PET-only and PET/CT with or without group convolution) were compared. Detectability and quantification were evaluated. RESULTS: The PET/CT input model with group convolution performed best regarding lesion signal recovery and was selected for detailed evaluation. Synthetic PET images were of high visual image quality; mean absolute lesion SUVmax (maximum standardized uptake value) difference was 1.5. Patient-based sensitivity and specificity for lesion detection were 79% and 100%, respectively. Not-detected lesions were of lower tracer uptake and lesion volume. In a matched-pair comparison, patient-based (lesion-based) detection rate was 89% (78%) for PERCIST (PET response criteria in solid tumors)-measurable and 36% (22%) for non PERCIST-measurable lesions. CONCLUSION: Lesion detectability and lesion quantification were promising in the context of extremely fast acquisition times. Possible application scenarios might include re-staging of late-stage cancer patients, in whom assessment of total tumor burden can be of higher relevance than detailed evaluation of small and low-uptake lesions.

Generalization of deep learning models for ultra-low-count amyloid PET/MRI using transfer learning.

PURPOSE: We aimed to evaluate the performance of deep learning-based generalization of ultra-low-count amyloid PET/MRI enhancement when applied to studies acquired with different scanning hardware and protocols. METHODS: Eighty simultaneous [18F]florbetaben PET/MRI studies were acquired, split equally between two sites (site 1: Signa PET/MRI, GE Healthcare, 39 participants, 67 ± 8 years, 23 females; site 2: mMR, Siemens Healthineers, 64 ± 11 years, 23 females) with different MRI protocols. Twenty minutes of list-mode PET data (90-110 min post-injection) were reconstructed as ground-truth. Ultra-low-count data obtained from undersampling by a factor of 100 (site 1) or the first minute of PET acquisition (site 2) were reconstructed for ultra-low-dose/ultra-short-time (1% dose and 5% time, respectively) PET images. A deep convolution neural network was pre-trained with site 1 data and either (A) directly applied or (B) trained further on site 2 data using transfer learning. Networks were also trained from scratch based on (C) site 2 data or (D) all data. Certified physicians determined amyloid uptake (+/-) status for accuracy and scored the image quality. The peak signal-to-noise ratio, structural similarity, and root-mean-squared error were calculated between images and their ground-truth counterparts. Mean regional standardized uptake value ratios (SUVR, reference region: cerebellar cortex) from 37 successful site 2 FreeSurfer segmentations were analyzed. RESULTS: All network-synthesized images had reduced noise than their ultra-low-count reconstructions. Quantitatively, image metrics improved the most using method B, where SUVRs had the least variability from the ground-truth and the highest effect size to differentiate between positive and negative images. Method A images had lower accuracy and image quality than other methods; images synthesized from methods B-D scored similarly or better than the ground-truth images. CONCLUSIONS: Deep learning can successfully produce diagnostic amyloid PET images from short frame reconstructions. Data bias should be considered when applying pre-trained deep ultra-low-count amyloid PET/MRI networks for generalization.

Feasibility of one-eighth time gated myocardial perfusion SPECT functional imaging using IQ-SPECT.

PURPOSE: IQ-SPECT, an add-on to general purpose cameras based on multifocal collimation, can reduce myocardial perfusion imaging (MPI) acquisition times to one-fourth that of standard procedures (to 12 s/view). In a phantom study, a reduction of the acquisition time to one-eighth of the standard time (to 6 s/view) was demonstrated as feasible. It remains unclear whether such a reduction could be extended to clinical practice. METHODS: Fifty patients with suspected or diagnosed CAD underwent a 2-day stress-rest (99m)Tc-sestamibi MPI protocol. Two consecutive SPECT acquisitions (6 and 12 s/view) were performed. Electrocardiogram-gated images were reconstructed with and without attenuation correction (AC). Polar maps were generated and visually scored by two blinded observers for image quality and perfusion in 17 segments. Global and regional summed stress score (SSS), summed rest score (SRS) and summed difference score (SDS) were determined. Left ventricular volumes and ejection fraction were calculated based on automated contour detection. RESULTS: Image quality was scored higher with the 12 s/view acquisition, both with and without AC. Summed scores were statistically comparable between the 6 s/view and the 12 s/view acquisition, both globally and in individual coronary territories (e.g. in images with AC, SSS were 6.6 ± 8.3 and 6.2 ± 8.2 with 6 s and 12 s/view, respectively, p = 0.10; SRS were 3.9 ± 5.6 and 3.5 ± 5.3, respectively, p = 0.19; and SDS were 2.8 ± 5.7 and 2.6 ± 5.7, respectively, p = 0.59). Both acquisitions allowed MPI-based diagnosis of CAD in 25 of the 50 patients (with AC). Calculated end-diastolic volume (EDV) and end-systolic volume (ESV) were modestly higher with the 6 s/view acquisition than with the 12 s/view acquisition (EDV +4.8 ml at rest and +3.7 ml after stress, p = 0.003; ESV +4.1 ml at rest and +2.6 ml after stress, p = 0.01), whereas the ejection fraction did not differ (-1.2 % at rest, p = 0.20, and -0.9 % after stress, p = 0.27). CONCLUSION: Image quality and LV functional parameters obtained with a one-eighth acquisition time were statistically comparable to the previously validated one-fourth time protocol using IQ-SPECT. Shorter acquisition times without loss of diagnostic accuracy provide improved patient comfort and streamlined departmental efficiency.

ISIT-QA: In Silico Imaging Trial to Evaluate a Low-Count Quantitative SPECT Method Across Multiple Scanner-Collimator Configurations for 223Ra-Based Radiopharmaceutical Therapies.

Personalized dose-based treatment planning requires accurate and reproducible noninvasive measurements to ensure safety and effectiveness. Dose estimation using SPECT is possible but challenging for alpha (α)-particle-emitting radiopharmaceutical therapy (α-RPT) because of complex γ-emission spectra, extremely low counts, and various image-degrading artifacts across a plethora of scanner-collimator configurations. Through the incorporation of physics-based considerations and skipping of the potentially lossy voxel-based reconstruction step, a recently developed projection-domain low-count quantitative SPECT (LC-QSPECT) method has the potential to provide reproducible, accurate, and precise activity concentration and dose measures across multiple scanners, as is typically the case in multicenter settings. To assess this potential, we conducted an in silico imaging trial to evaluate the LC-QSPECT method for a 223Ra-based α-RPT, with the trial recapitulating patient and imaging system variabilities. Methods: A virtual imaging trial titled In Silico Imaging Trial for Quantitation Accuracy (ISIT-QA) was designed with the objectives of evaluating the performance of the LC-QSPECT method across multiple scanner-collimator configurations and comparing performance with a conventional reconstruction-based quantification method. In this trial, we simulated 280 realistic virtual patients with bone-metastatic castration-resistant prostate cancer treated with 223Ra-based α-RPT. The trial was conducted with 9 simulated SPECT scanner-collimator configurations. The primary objective of this trial was to evaluate the reproducibility of dose estimates across multiple scanner-collimator configurations using LC-QSPECT by calculating the intraclass correlation coefficient. Additionally, we compared the reproducibility and evaluated the accuracy of both considered quantification methods across multiple scanner-collimator configurations. Finally, the repeatability of the methods was evaluated in a test-retest study. Results: In this trial, data from 268 223RaCl2 treated virtual prostate cancer patients, with a total of 2,903 lesions, were used to evaluate LC-QSPECT. LC-QSPECT provided dose estimates with good reproducibility across the 9 scanner-collimator configurations (intraclass correlation coefficient > 0.75) and high accuracy (ensemble average values of recovery coefficients ranged from 1.00 to 1.02). Compared with conventional reconstruction-based quantification, LC-QSPECT yielded significantly improved reproducibility across scanner-collimator configurations, accuracy, and test-retest repeatability ([Formula: see text] Conclusion: LC-QSPECT provides reproducible, accurate, and repeatable dose estimations in 223Ra-based α-RPT as evaluated in ISIT-QA. These findings provide a strong impetus for multicenter clinical evaluations of LC-QSPECT in dose quantification for α-RPTs.

Deep denoiser prior driven relaxed iterated Tikhonov method for low-count PET image restoration.

Objective. Low-count positron emission tomography (PET) imaging is an efficient way to promote more widespread use of PET because of its short scan time and low injected activity. However, this often leads to low-quality PET images with clinical image reconstruction, due to high noise and blurring effects. Existing PET image restoration (IR) methods hinder their own restoration performance due to the semi-convergence property and the lack of suitable denoiser prior.Approach. To overcome these limitations, we propose a novel deep plug-and-play IR method called Deep denoiser Prior driven Relaxed Iterated Tikhonov method (DP-RI-Tikhonov). Specifically, we train a deep convolutional neural network denoiser to generate a flexible deep denoiser prior to handle high noise. Then, we plug the deep denoiser prior as a modular part into a novel iterative optimization algorithm to handle blurring effects and propose an adaptive parameter selection strategy for the iterative optimization algorithm.Main results. Simulation results show that the deep denoiser prior plays the role of reducing noise intensity, while the novel iterative optimization algorithm and adaptive parameter selection strategy can effectively eliminate the semi-convergence property. They enable DP-RI-Tikhonov to achieve an average quantitative result (normalized root mean square error, structural similarity) of (0.1364, 0.9574) at the stopping iteration, outperforming a conventional PET IR method with an average quantitative result of (0.1533, 0.9523) and a state-of-the-art deep plug-and-play IR method with an average quantitative result of (0.1404, 0.9554). Moreover, the advantage of DP-RI-Tikhonov becomes more obvious at the last iteration. Experiments on six clinical whole-body PET images further indicate that DP-RI-Tikhonov successfully reduces noise intensity and recovers fine details, recovering sharper and more uniform images than the comparison methods.Significance. DP-RI-Tikhonov's ability to reduce noise intensity and effectively eliminate the semi-convergence property overcomes the limitations of existing methods. This advancement may have substantial implications for other medical IR.

Deep learning generation of preclinical positron emission tomography (PET) images from low-count PET with task-based performance assessment.

BACKGROUND: Preclinical low-count positron emission tomography (LC-PET) imaging offers numerous advantages such as facilitating imaging logistics, enabling longitudinal studies of long- and short-lived isotopes as well as increasing scanner throughput. However, LC-PET is characterized by reduced photon-count levels resulting in low signal-to-noise ratio (SNR), segmentation difficulties, and quantification uncertainties. PURPOSE: We developed and evaluated a novel deep-learning (DL) architecture-Attention based Residual-Dilated Net (ARD-Net)-to generate standard-count PET (SC-PET) images from LC-PET images. The performance of the ARD-Net framework was evaluated for numerous low count realizations using fidelity-based qualitative metrics, task-based segmentation, and quantitative metrics. METHOD: Patient Derived tumor Xenograft (PDX) with tumors implanted in the mammary fat-pad were subjected to preclinical [18F]-Fluorodeoxyglucose (FDG)-PET/CT imaging. SC-PET images were derived from a 10 min static FDG-PET acquisition, 50 min post administration of FDG, and were resampled to generate four distinct LC-PET realizations corresponding to 10%, 5%, 1.6%, and 0.8% of SC-PET count-level. ARD-Net was trained and optimized using 48 preclinical FDG-PET datasets, while 16 datasets were utilized to assess performance. Further, the performance of ARD-Net was benchmarked against two leading DL-based methods (Residual UNet, RU-Net; and Dilated Network, D-Net) and non-DL methods (Non-Local Means, NLM; and Block Matching 3D Filtering, BM3D). The performance of the framework was evaluated using traditional fidelity-based image quality metrics such as Structural Similarity Index Metric (SSIM) and Normalized Root Mean Square Error (NRMSE), as well as human observer-based tumor segmentation performance (Dice Score and volume bias) and quantitative analysis of Standardized Uptake Value (SUV) measurements. Additionally, radiomics-derived features were utilized as a measure of quality assurance (QA) in comparison to true SC-PET. Finally, a performance ensemble score (EPS) was developed by integrating fidelity-based and task-based metrics. Concordance Correlation Coefficient (CCC) was utilized to determine concordance between measures. The non-parametric Friedman Test with Bonferroni correction was used to compare the performance of ARD-Net against benchmarked methods with significance at adjusted p-value ≤0.01. RESULTS: ARD-Net-generated SC-PET images exhibited significantly better (p ≤ 0.01 post Bonferroni correction) overall image fidelity scores in terms of SSIM and NRMSE at majority of photon-count levels compared to benchmarked DL and non-DL methods. In terms of task-based quantitative accuracy evaluated by SUVMean and SUVPeak, ARD-Net exhibited less than 5% median absolute bias for SUVMean compared to true SC-PET and lower degree of variability compared to benchmarked DL and non-DL based methods in generating SC-PET. Additionally, ARD-Net-generated SC-PET images displayed higher degree of concordance to SC-PET images in terms of radiomics features compared to non-DL and other DL approaches. Finally, the ensemble score suggested that ARD-Net exhibited significantly superior performance compared to benchmarked algorithms (p ≤ 0.01 post Bonferroni correction). CONCLUSION: ARD-Net provides a robust framework to generate SC-PET from LC-PET images. ARD-Net generated SC-PET images exhibited superior performance compared other DL and non-DL approaches in terms of image-fidelity based metrics, task-based segmentation metrics, and minimal bias in terms of task-based quantification performance for preclinical PET imaging.

Progressive multifocal leukoencephalopathy as the presenting feature in a patient with occult low-count monoclonal B-cell lymphocytosis.

Introduction: Low Count Monoclonal B-Cell Lymphocytosis (LC-MBL) is a relatively poorly understood entity which has been suggested to be very common in asymptomatic adults and possibly related to infectious complications despite not progressing to CLL. Methods: We describe the first case of Progressive Multifocal Leukoencephalopathy (PML) presenting in a 72-year-old man with LC-MBL but no other immunocompromising conditions. Results: A diagnosis of PML was confirmed with classic MRI findings in association with a high CSF John Cunningham polyomavirus (JCV) viral load (4.09' 105 copies/mL). An extensive search for underlying immunocompromising conditions only demonstrated LC-MBL representing approximately 4% of total leukocytes (0.2' 109/L). Discussion: This is the first report of PML in association with LC-MBL. Careful review of peripheral blood flow cytometry results is necessary to identify this disorder. Further study of the epidemiology and infectious complications of LC-MBL are warranted.

Introduction: La lymphocytose monoclonale à cellules B (LMB) est une maladie relativement mal comprise qui serait très courante chez des adultes asymptomatiques et qui pourrait être liée à des complications infectieuses, même si elle n'évolue pas en leucémie lymphocytique chronique. Méthodologies: Nous décrivons le premier cas de leucoencéphalopathie multifocale progressive (LEMP) observé chez un patient (72 ans) atteint de LMB, mais ne présentant pas d'autres pathologies induisant une immunodéficience. Résultats: Des résultats d'IRM classiques et une forte charge du virus JC (John Cunningham) dans le liquide céphalorachidien (4,09 × 105 copies/mL) ont confirmé un diagnostic de LEMP. De nombreux tests visant à révéler une immunodéficience sous-jacente ont seulement montré que les cellules B monoclonales représentaient environ 4% des leucocytes totaux (0,2 × 109/L). Discussion: Il s'agit du premier cas observé de LEMP en association avec une LMB. Il faut analyser soigneusement les résultats d'une cytométrie en flux du sang périphérique pour diagnostiquer ce trouble. Il convient de continuer d'étudier l'épidémiologie et les complications infectieuses de la LMB.

FedFTN: Personalized federated learning with deep feature transformation network for multi-institutional low-count PET denoising.

Low-count PET is an efficient way to reduce radiation exposure and acquisition time, but the reconstructed images often suffer from low signal-to-noise ratio (SNR), thus affecting diagnosis and other downstream tasks. Recent advances in deep learning have shown great potential in improving low-count PET image quality, but acquiring a large, centralized, and diverse dataset from multiple institutions for training a robust model is difficult due to privacy and security concerns of patient data. Moreover, low-count PET data at different institutions may have different data distribution, thus requiring personalized models. While previous federated learning (FL) algorithms enable multi-institution collaborative training without the need of aggregating local data, addressing the large domain shift in the application of multi-institutional low-count PET denoising remains a challenge and is still highly under-explored. In this work, we propose FedFTN, a personalized federated learning strategy that addresses these challenges. FedFTN uses a local deep feature transformation network (FTN) to modulate the feature outputs of a globally shared denoising network, enabling personalized low-count PET denoising for each institution. During the federated learning process, only the denoising network's weights are communicated and aggregated, while the FTN remains at the local institutions for feature transformation. We evaluated our method using a large-scale dataset of multi-institutional low-count PET imaging data from three medical centers located across three continents, and showed that FedFTN provides high-quality low-count PET images, outperforming previous baseline FL reconstruction methods across all low-count levels at all three institutions.

SPECT reconstruction with a trained regularizer using CT-side information: Application to 177Lu SPECT imaging.

Improving low-count SPECT can shorten scans and support pre-therapy theranostic imaging for dosimetry-based treatment planning, especially with radionuclides like 177Lu known for low photon yields. Conventional methods often underperform in low-count settings, highlighting the need for trained regularization in model-based image reconstruction. This paper introduces a trained regularizer for SPECT reconstruction that leverages segmentation based on CT imaging. The regularizer incorporates CT-side information via a segmentation mask from a pre-trained network (nnUNet). In this proof-of-concept study, we used patient studies with 177Lu DOTATATE to train and tested with phantom and patient datasets, simulating pre-therapy imaging conditions. Our results show that the proposed method outperforms both standard unregularized EM algorithms and conventional regularization with CT-side information. Specifically, our method achieved marked improvements in activity quantification, noise reduction, and root mean square error. The enhanced low-count SPECT approach has promising implications for theranostic imaging, post-therapy imaging, whole body SPECT, and reducing SPECT acquisition times.

Low-count PET image reconstruction based on truncated inverse radon layer and U-shaped network.

Objective.Positron emission tomography (PET) is a functional imaging widely used in various applications such as tumour detection. PET image reconstruction is an ill-posed inverse problem, and the model-based iterative reconstruction methods commonly used in clinical practice have disadvantages such as long time consumption and low signal-to-noise ratio, especially at low doses.Approach.In this study, we propose a deep learning-based reconstruction method that is capable of reconstructing images directly from low-count sinograms. Our network consists of two parts, a truncated inverse radon layer for implementing domain transform and a U-shaped network for image enhancement.Main result.We validated our method on both simulation data and real data. Compared to ordered subset expectation maximization with a post-Guassian filter, the structural similarity can be improved from 0.9357 to 0.9613 and the peak signal-to-noise ratio can be improved by 5 dB.Significance.The proposed method can directly convert low-count sinograms into PET images, while obtaining improved image quality and having less time consumption than iterative reconstruction algorithms and the state-of-the-art convolutional neural network.

Dynamic low-count PET image reconstruction using spatio-temporal primal dual network.

Objective.Dynamic positron emission tomography (PET) imaging, which can provide information on dynamic changes in physiological metabolism, is now widely used in clinical diagnosis and cancer treatment. However, the reconstruction from dynamic data is extremely challenging due to the limited counts received in individual frame, especially in ultra short frames. Recently, the unrolled model-based deep learning methods have shown inspiring results for low-count PET image reconstruction with good interpretability. Nevertheless, the existing model-based deep learning methods mainly focus on the spatial correlations while ignore the temporal domain.Approach.In this paper, inspired by the learned primal dual (LPD) algorithm, we propose the spatio-temporal primal dual network (STPDnet) for dynamic low-count PET image reconstruction. Both spatial and temporal correlations are encoded by 3D convolution operators. The physical projection of PET is embedded in the iterative learning process of the network, which provides the physical constraints and enhances interpretability.Main results.The experiments of both simulation data and real rat scan data have shown that the proposed method can achieve substantial noise reduction in both temporal and spatial domains and outperform the maximum likelihood expectation maximization, spatio-temporal kernel method, LPD and FBPnet.Significance.Experimental results show STPDnet better reconstruction performance in the low count situation, which makes the proposed method particularly suitable in whole-body dynamic imaging and parametric PET imaging that require extreme short frames and usually suffer from high level of noise.

Poisson Phase Retrieval in Very Low-count Regimes.

This paper discusses phase retrieval algorithms for maximum likelihood (ML) estimation from measurements following independent Poisson distributions in very low-count regimes, e.g., 0.25 photon per pixel. To maximize the log-likelihood of the Poisson ML model, we propose a modified Wirtinger flow (WF) algorithm using a step size based on the observed Fisher information. This approach eliminates all parameter tuning except the number of iterations. We also propose a novel curvature for majorize-minimize (MM) algorithms with a quadratic majorizer. We show theoretically that our proposed curvature is sharper than the curvature derived from the supremum of the second derivative of the Poisson ML cost function. We compare the proposed algorithms (WF, MM) with existing optimization methods, including WF using other step-size schemes, quasi-Newton methods such as LBFGS and alternating direction method of multipliers (ADMM) algorithms, under a variety of experimental settings. Simulation experiments with a random Gaussian matrix, a canonical DFT matrix, a masked DFT matrix and an empirical transmission matrix demonstrate the following. 1) As expected, algorithms based on the Poisson ML model consistently produce higher quality reconstructions than algorithms derived from Gaussian noise ML models when applied to low-count data. Furthermore, incorporating regularizers, such as corner-rounded anisotropic total variation (TV) that exploit the assumed properties of the latent image, can further improve the reconstruction quality. 2) For unregularized cases, our proposed WF algorithm with Fisher information for step size converges faster (in terms of cost function and PSNR vs. time) than other WF methods, e.g., WF with empirical step size, backtracking line search, and optimal step size for the Gaussian noise model; it also converges faster than the LBFGS quasi-Newton method. 3) In regularized cases, our proposed WF algorithm converges faster than WF with backtracking line search, LBFGS, MM and ADMM.

3D Convolutional Neural Network-Based Denoising of Low-Count Whole-Body 18F-Fluorodeoxyglucose and 89Zr-Rituximab PET Scans.

Acquisition time and injected activity of 18F-fluorodeoxyglucose (18F-FDG) PET should ideally be reduced. However, this decreases the signal-to-noise ratio (SNR), which impairs the diagnostic value of these PET scans. In addition, 89Zr-antibody PET is known to have a low SNR. To improve the diagnostic value of these scans, a Convolutional Neural Network (CNN) denoising method is proposed. The aim of this study was therefore to develop CNNs to increase SNR for low-count 18F-FDG and 89Zr-antibody PET. Super-low-count, low-count and full-count 18F-FDG PET scans from 60 primary lung cancer patients and full-count 89Zr-rituximab PET scans from five patients with non-Hodgkin lymphoma were acquired. CNNs were built to capture the features and to denoise the PET scans. Additionally, Gaussian smoothing (GS) and Bilateral filtering (BF) were evaluated. The performance of the denoising approaches was assessed based on the tumour recovery coefficient (TRC), coefficient of variance (COV; level of noise), and a qualitative assessment by two nuclear medicine physicians. The CNNs had a higher TRC and comparable or lower COV to GS and BF and was also the preferred method of the two observers for both 18F-FDG and 89Zr-rituximab PET. The CNNs improved the SNR of low-count 18F-FDG and 89Zr-rituximab PET, with almost similar or better clinical performance than the full-count PET, respectively. Additionally, the CNNs showed better performance than GS and BF.

A new radionuclide identification method for low-count energy spectra with multiple radionuclides.

Radionuclide identification is to recognize the radionuclides in the environment by analyzing the energy spectrum. Rapid and accurate identification is important for nuclear security. Current radionuclide identification methods based on traditional peak search require background subtraction. As a result, they have difficulties to deal with complex situations in practical applications such as low-count energy spectrum and mixed nuclides. In this paper, we propose a new radionuclide identification method with a feature enhancer and a one-dimensional neural network. The training dataset in this method is from simulated data generated by Geant4. By preprocessing the input energy spectrum data through the feature enhancer and extracting the nonlinear information through the neural network, this approach performs well on experimental energy spectra even at low count. The method also shows a high recognition accuracy and little misjudgments when dealing with mixed radionuclides spectra. Due to its good performance in identifying mixed nuclides and low-count spectra, the method has been deployed in portable instrument for radionuclide identification in real-time measurement.

Spatial adaptive and transformer fusion network (STFNet) for low-count PET blind denoising with MRI.

PURPOSE: Positron emission tomography (PET) has been widely used in various clinical applications. PET is a type of emission computed tomography and operates by positron annihilation radiation. With magnetic resonance imaging (MRI) providing anatomical information, joint PET/MRI reduces the radiation exposure risk of patients. Improved hardware and imaging algorithms have been proposed to further decrease the dose from radioactive tracers or the bed duration, but few methods focus on denoising low-count PET with MRI input. The existing methods are based on fixed conventional convolution and local attention, which do not sufficiently extract and fuse contextual and complementary information from multimodal input. There is still much room for improvement. Therefore, we propose a novel deep learning method for low-count PET/MRI denoising called the spatial-adaptive and transformer fusion network (STFNet), which consists of a Siamese encoder with a spatial-adaptive block (SA-block) and the transformer fusion encoder (TFE). METHODS: Our proposed STFNet consists of a Siamese encoder with an SA-block, TFE, and two branches of the decoder. First, in the encoder, we adapt the SA-block in the Siamese encoder. The SA-block comprises deformable convolution with fusion modulation (DCFM) and two convolutional operations, which can promote network extraction of more relative and long-range contextual features. Second, the pixel-to-pixel TFE helps the network establish a local and global relationship between high-level feature maps of PET and MRI. In the decoder part, we design two branches for PET denoising and MRI translation, and predictions are obtained by trainable weighted summation. This proposed algorithm is implemented to predict synthetic standard-dose neck PET images from low-count neck PET images and MRI. Additionally, this method is compared with the existing U-Net and residual U-Net methods with and without MRI input. RESULTS: To demonstrate the advantages of our method, we introduce configuration studies about TFE, ablation studies, and empirical comparative studies. Quantitative analyses are based on root mean square error (RSME), peak signal-to-noise ratio (PSNR), structural similarity (SSIM), and Pearson correlation coefficient (PCC). Additionally, qualitative results show the comparisons between our proposed method and other existing methods. All experimental results and visualizations show that our method achieves state-of-the-art performance in quantification and qualification. CONCLUSIONS: Based on our experiments, STFNet performs better than existing methods in measurement and visualization. However, our proposed method may still be suboptimal because we apply only the L1 loss to train our data set, and the data set includes corrupted PET with different low counts. In the future, we may exploit a generative adversarial network (GAN)-based paradigm in our STFNet to further improve the visual quality.

A personalized deep learning denoising strategy for low-count PET images.

Objective. Deep learning denoising networks are typically trained with images that are representative of the testing data. Due to the large variability of the noise levels in positron emission tomography (PET) images, it is challenging to develop a proper training set for general clinical use. Our work aims to develop a personalized denoising strategy for the low-count PET images at various noise levels.Approach.We first investigated the impact of the noise level in the training images on the model performance. Five 3D U-Net models were trained on five groups of images at different noise levels, and a one-size-fits-all model was trained on images covering a wider range of noise levels. We then developed a personalized weighting method by linearly blending the results from two models trained on 20%-count level images and 60%-count level images to balance the trade-off between noise reduction and spatial blurring. By adjusting the weighting factor, denoising can be conducted in a personalized and task-dependent way.Main results.The evaluation results of the six models showed that models trained on noisier images had better performance in denoising but introduced more spatial blurriness, and the one-size-fits-all model did not generalize well when deployed for testing images with a wide range of noise levels. The personalized denoising results showed that noisier images require higher weights on noise reduction to maximize the structural similarity and mean squared error. And model trained on 20%-count level images can produce the best liver lesion detectability.Significance.Our study demonstrated that in deep learning-based low dose PET denoising, noise levels in the training input images have a substantial impact on the model performance. The proposed personalized denoising strategy utilized two training sets to overcome the drawbacks introduced by each individual network and provided a series of denoised results for clinical reading.

Positron Range Corrections and Denoising Techniques for Gallium-68 PET Imaging: A Literature Review.

Gallium-68 (68Ga) is characterized by relatively high positron energy compared to Fluorine-18 (18F), causing substantial image quality degradation. Furthermore, the presence of statistical noise can further degrade image quality. The aim of this literature review is to identify the recently developed positron range correction techniques for 68Ga, as well as noise reduction methods to enhance the image quality of low count 68Ga PET imaging. The search engines PubMed and Scopus were employed, and we limited our research to published results from January 2010 until 1 August 2022. Positron range correction was achieved by using either deblurring or deep learning approaches. The proposed techniques improved the image quality and, in some cases, achieved an image quality comparable to 18F PET. However, none of these techniques was validated in clinical studies. PET denoising for 68Ga-labeled radiotracers was reported using either reconstruction-based techniques or deep learning approaches. It was demonstrated that both approaches can substantially enhance the image quality by reducing the noise levels of low count 68Ga PET imaging. The combination of 68Ga-specific positron range correction techniques and image denoising approaches may enable the application of low-count, high-quality 68Ga PET imaging in a clinical setting.

An improved patch-based regularization method for PET image reconstruction.

BACKGROUND: Statistical reconstruction methods based on penalized maximum likelihood (PML) are being increasingly used in positron emission tomography (PET) imaging to reduce noise and improve image quality. Wang and Qi proposed a patch-based edge-preserving penalties algorithm that can be implemented in three simple steps: a maximum-likelihood expectation-maximization (MLEM) image update, an image smoothing step, and a pixel-by-pixel image fusion step. The pixel-by-pixel image fusion step, which fuses the MLEM updated image and the smoothed image, involves a trade-off between preserving the fine structural features of an image and suppressing noise. Particularly when reconstructing images from low-count data, this step cannot preserve fine structural features in detail. To better preserve these features and accelerate the algorithm convergence, we proposed to improve the patch-based regularization reconstruction method. METHODS: Our improved method involved adding a total variation (TV) regularization step following the MLEM image update in the patch-based algorithm. A feature refinement (FR) step was then used to extract the lost fine structural features from the residual image between the TV regularized image and the fused image based on patch regularization. These structural features would then be added back to the fused image. With the addition of these steps, each iteration of the image should gain more structural information. A brain phantom simulation experiment and a mouse study were conducted to evaluate our proposed improved method. Brain phantom simulation with added noise were used to determine the feasibility of the proposed algorithm and its acceleration of convergence. Data obtained from the mouse study were divided into event count sets to validate the performance of the proposed algorithm when reconstructing images from low-count data. Five criteria were used for quantitative evaluation: signal-to-noise ratio (SNR), covariance (COV), contrast recovery coefficient (CRC), regional relative bias, and relative variance. RESULTS: The bias and variance of the phantom brain image reconstructed using the patch-based method were 0.421 and 5.035, respectively, and this process took 83.637 seconds. The bias and variance of the image reconstructed by the proposed improved method, however, were 0.396 and 4.568, respectively, and this process took 41.851 seconds. This demonstrates that the proposed algorithm accelerated the reconstruction convergence. The CRC of the phantom brain image reconstructed using the patch-based method was iterated 20 times and reached 0.284, compared with the proposed method, which reached 0.446. When using a count of 5,000 K data obtained from the mouse study, both the patch-based method and the proposed method reconstructed images similar to the ground truth image. The intensity of the ground truth image was 88.3, and it was located in the 102nd row and the 116th column. However, when the count was reduced to below 40 K, and the patch-based method was used, image quality was significantly reduced. This effect was not observed when the proposed method was used. When a count of 40 K was used, the image intensity was 58.79 when iterated 100 times by the patch-based method, and it was located in the 102nd row and the 116th column, while the intensity when iterated 50 times by the proposed method was 63.83. This suggests that the proposed method improves image reconstruction from low-count data. CONCLUSIONS: This improved method of PET image reconstruction could potentially improve the quality of PET images faster than other methods and also produce better reconstructions from low-count data.

Reducing Radiation Exposure to Paediatric Patients Undergoing [18F]FDG-PET/CT Imaging.

PURPOSE: To investigate the possibility of reducing the injected activity for whole-body [18F]FDG-PET/CT studies of paediatric oncology patients and to assess the usefulness of time-of-flight (TOF) acquisition on PET image quality at reduced count levels. PROCEDURES: Twenty-nine paediatric oncology patients (12F/17M, 3-18 years old (median age 13y), weight 45±20 kg, BMI 19±4 kg/m2), who underwent routine whole-body PET/CT examinations on a Siemens Biograph mCT TrueV system with TOF capability (555ps) were included in this study. The mean injected activity was 156 ± 45 MBq (3.8 ± 0.8 kg/MBq) and scaled to patient weight. The raw data was collected in listmode (LM) format and pre-processed to simulate reduced levels of [18F]FDG activity (75, 50, 35, 20 and 10% of the original counts) by randomly removing events from the original LM data. All data were reconstructed using the vendor-specific e7-tools with standard OSEM only, with OSEM plus resolution recovery (PSF). The reconstructions were repeated with added TOF (TOF) and PSF+TOF. The benefit of TOF together with the reduced count levels was evaluated by calculating the gains in signal-to-noise ratio (SNR) in the liver and contrast-to-noise ratio (CNR) in all PET-positive lesions before and after TOF employed at every simulated reduced count level. Finally, the PSF+TOF images at 50, 75 and 100% of counts were evaluated clinically on a 5-point scale by three nuclear medicine physicians. RESULTS: The visual inspection of the reconstructed images did not reveal significant differences in image quality between 75 and 100% count levels for PSF+TOF. The improvements in SNR and CNR were the greatest for TOF reconstruction and PSF combined. Both SNR and CNR gains did increase linearly with the patients BMI for both OSEM only and PSF reconstruction. These benefits were observed until reducing the counts to 50 and 35% for SNR and CNR, respectively. CONCLUSIONS: The benefit of using TOF was noticeable when using 50% or greater of the counts when evaluating the CNR and SNR. For [18F]FDG-PET/CT, whole-body paediatric imaging the injected activity can be reduced to 75% of the original dose without compromising PET image quality.

A Monte Carlo method for calculation of the characteristic limits decision threshold and detection limit in low-level radioactivity measurements.

In this study, calculation of decision threshold and detection limit expressed in counts for low-level radioactivity measurements were evaluated and compared to a Monte Carlo method for the case of paired Poisson-distributed observations, i.e. for discrete variables. The calculated characteristic limits obtained from Monte Carlo calculations were compared with analytical expressions given in literature. The results in this study show that the equations given by Currie are in good agreement with the results from the Monte Carlo calculations simulating nuclear counting applications with a low number of observed counts. An exception is observed for a background corresponding to zero counts. This study also shows that at a low number of counts, the specific boundary conditions of the interval that represents counts corresponding to the presence of the analyte (>or ≥), have an impact on the false positives and negatives rates as defined by the parameters α and ß.