Noise2Void: unsupervised denoising of PET images

Deep learning-based PET image denoising and reconstruction: a review

This review focuses on positron emission tomography (PET) imaging algorithms and traces the evolution of PET image reconstruction methods. First, we provide an overview of conventional PET image reconstruction methods from filtered backprojection through to recent iterative PET image reconstruction algorithms, and then review deep learning methods for PET data up to the latest innovations within three main categories. The first category involves post-processing methods for PET image denoising. The second category comprises direct image reconstruction methods that learn mappings from sinograms to the reconstructed images in an end-to-end manner. The third category comprises iterative reconstruction methods that combine conventional iterative image reconstruction with neural-network enhancement. We discuss future perspectives on PET imaging and deep learning technology.

PIDNET: Polar Transformation Based Implicit Disentanglement Network for Truncation Artifacts

The interior problem, a persistent ill-posed challenge in CT imaging, gives rise to truncation artifacts capable of distorting CT values, thereby significantly impacting clinical diagnoses. Traditional methods have long struggled to effectively solve this issue until the advent of supervised models built on deep neural networks. However, supervised models are constrained by the need for paired data, limiting their practical application. Therefore, we propose a simple and efficient unsupervised method based on the Cycle-GAN framework. Introducing an implicit disentanglement strategy, we aim to separate truncation artifacts from content information. The separated artifact features serve as complementary constraints and the source of generating simulated paired data to enhance the training of the sub-network dedicated to removing truncation artifacts. Additionally, we incorporate polar transformation and an innovative constraint tailored specifically for truncation artifact features, further contributing to the effectiveness of our approach. Experiments conducted on multiple datasets demonstrate that our unsupervised network outperforms the traditional Cycle-GAN model significantly. When compared to state-of-the-art supervised models trained on paired datasets, our model achieves comparable visual results and closely aligns with quantitative evaluation metrics.

A Review on Low-Dose Emission Tomography Post-Reconstruction Denoising with Neural Network Approaches

Low-dose emission tomography (ET) plays a crucial role in medical imaging, enabling the acquisition of functional information for various biological processes while minimizing the patient dose. However, the inherent randomness in the photon counting process is a source of noise which is amplified in low-dose ET. This review article provides an overview of existing post-processing techniques, with an emphasis on deep neural network (NN) approaches. Furthermore, we explore future directions in the field of NN-based low-dose ET. This comprehensive examination sheds light on the potential of deep learning in enhancing the quality and resolution of low-dose ET images, ultimately advancing the field of medical imaging.

Artificial Intelligence for PET and SPECT Image Enhancement

Nuclear medicine imaging modalities such as PET and SPECT are confounded by high noise levels and low spatial resolution, necessitating postreconstruction image enhancement to improve their quality and quantitative accuracy. Artificial intelligence (AI) models such as convolutional neural networks, U-Nets, and generative adversarial networks have shown promising outcomes in enhancing PET and SPECT images. This review article presents a comprehensive survey of state-of-the-art AI methods for PET and SPECT image enhancement and seeks to identify emerging trends in this field. We focus on recent breakthroughs in AI-based PET and SPECT image denoising and deblurring. Supervised deep-learning models have shown great potential in reducing radiotracer dose and scan times without sacrificing image quality and diagnostic accuracy. However, the clinical utility of these methods is often limited by their need for paired clean and corrupt datasets for training. This has motivated research into unsupervised alternatives that can overcome this limitation by relying on only corrupt inputs or unpaired datasets to train models. This review highlights recently published supervised and unsupervised efforts toward AI-based PET and SPECT image enhancement. We discuss cross-scanner and cross-protocol training efforts, which can greatly enhance the clinical translatability of AI-based image enhancement tools. We also aim to address the looming question of whether the improvements in image quality generated by AI models lead to actual clinical benefit. To this end, we discuss works that have focused on task-specific objective clinical evaluation of AI models for image enhancement or incorporated clinical metrics into their loss functions to guide the image generation process. Finally, we discuss emerging research directions, which include the exploration of novel training paradigms, curation of larger task-specific datasets, and objective clinical evaluation that will enable the realization of the full translation potential of these models in the future.

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.

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.

Self-supervised deep learning for joint 3D low-dose PET/CT image denoising

Deep learning (DL)-based denoising of low-dose positron emission tomography (LDPET) and low-dose computed tomography (LDCT) has been widely explored. However, previous methods have focused only on single modality denoising, neglecting the possibility of simultaneously denoising LDPET and LDCT using only one neural network, i.e., joint LDPET/LDCT denoising. Moreover, DL-based denoising methods generally require plenty of well-aligned LD-normal-dose (LD-ND) sample pairs, which can be difficult to obtain. To this end, we propose a self-supervised two-stage training framework named MAsk-then-Cycle (MAC), to achieve self-supervised joint LDPET/LDCT denoising. The first stage of MAC is masked autoencoder (MAE)-based pre-training and the second stage is self-supervised denoising training. Specifically, we propose a self-supervised denoising strategy named cycle self-recombination (CSR), which enables denoising without well-aligned sample pairs. Unlike other methods that treat noise as a homogeneous whole, CSR disentangles noise into signal-dependent and independent noises. This is more in line with the actual imaging process and allows for flexible recombination of noises and signals to generate new samples. These new samples contain implicit constraints that can improve the network's denoising ability. Based on these constraints, we design multiple loss functions to enable self-supervised training. Then we design a CSR-based denoising network to achieve joint 3D LDPET/LDCT denoising. Existing self-supervised methods generally lack pixel-level constraints on networks, which can easily lead to additional artifacts. Before denoising training, we perform MAE-based pre-training to indirectly impose pixel-level constraints on networks. Experiments on an LDPET/LDCT dataset demonstrate its superiority over existing methods. Our method is the first self-supervised joint LDPET/LDCT denoising method. It does not require any prior assumptions and is therefore more robust.

Federated Transfer Learning for Low-dose PET Denoising: A Pilot Study with Simulated Heterogeneous Data

Positron emission tomography (PET) with a reduced injection dose, i.e., low-dose PET, is an efficient way to reduce radiation dose. However, low-dose PET reconstruction suffers from a low signal-to-noise ratio (SNR), affecting diagnosis and other PET-related applications. Recently, deep learning-based PET denoising methods have demonstrated superior performance in generating high-quality reconstruction. However, these methods require a large amount of representative data for training, which can be difficult to collect and share due to medical data privacy regulations. Moreover, low-dose PET data at different institutions may use different low-dose protocols, leading to non-identical data distribution. While previous federated learning (FL) algorithms enable multi-institution collaborative training without the need of aggregating local data, it is challenging for previous methods to address the large domain shift caused by different low-dose PET settings, and the application of FL to PET is still under-explored. In this work, we propose a federated transfer learning (FTL) framework for low-dose PET denoising using heterogeneous low-dose data. Our experimental results on simulated multi-institutional data demonstrate that our method can efficiently utilize heterogeneous low-dose data without compromising data privacy for achieving superior low-dose PET denoising performance for different institutions with different low-dose settings, as compared to previous FL methods.

Virtual high-count PET image generation using a deep learning method

PURPOSE

Recently, deep learning-based methods have been established to denoise the low-count positron emission tomography (PET) images and predict their standard-count image counterparts, which could achieve reduction of injected dosage and scan time, and improve image quality for equivalent lesion detectability and clinical diagnosis. In clinical settings, the majority scans are still acquired using standard injection dose with standard scan time. In this work, we applied a 3D U-Net network to reduce the noise of standard-count PET images to obtain the virtual-high-count (VHC) PET images for identifying the potential benefits of the obtained VHC PET images.

METHODS

The training datasets, including down-sampled standard-count PET images as the network input and high-count images as the desired network output, were derived from 27 whole-body PET datasets, which were acquired using 90-min dynamic scan. The down-sampled standard-count PET images were rebinned with matched noise level of 195 clinical static PET datasets, by matching the normalized standard derivation (NSTD) inside 3D liver region of interests (ROIs). Cross-validation was performed on 27 PET datasets. Normalized mean square error (NMSE), peak signal to noise ratio (PSNR), structural similarity index (SSIM), and standard uptake value (SUV) bias of lesions were used for evaluation on standard-count and VHC PET images, with real-high-count PET image of 90 min as the gold standard. In addition, the network trained with 27 dynamic PET datasets was applied to 195 clinical static datasets to obtain VHC PET images. The NSTD and mean/max SUV of hypermetabolic lesions in standard-count and VHC PET images were evaluated. Three experienced nuclear medicine physicians evaluated the overall image quality of randomly selected 50 out of 195 patients' standard-count and VHC images and conducted 5-score ranking. A Wilcoxon signed-rank test was used to compare differences in the grading of standard-count and VHC images.

RESULTS

The cross-validation results showed that VHC PET images had improved quantitative metrics scores than the standard-count PET images. The mean/max SUVs of 35 lesions in the standard-count and true-high-count PET images did not show significantly statistical difference. Similarly, the mean/max SUVs of VHC and true-high-count PET images did not show significantly statistical difference. For the 195 clinical data, the VHC PET images had a significantly lower NSTD than the standard-count images. The mean/max SUVs of 215 hypermetabolic lesions in the VHC and standard-count images showed no statistically significant difference. In the image quality evaluation by three experienced nuclear medicine physicians, standard-count images and VHC images received scores with mean and standard deviation of 3.34±0.80 and 4.26 ± 0.72 from Physician 1, 3.02 ± 0.87 and 3.96 ± 0.73 from Physician 2, and 3.74 ± 1.10 and 4.58 ± 0.57 from Physician 3, respectively. The VHC images were consistently ranked higher than the standard-count images. The Wilcoxon signed-rank test also indicated that the image quality evaluation between standard-count and VHC images had significant difference.

CONCLUSIONS

A DL method was proposed to convert the standard-count images to the VHC images. The VHC images had reduced noise level. No significant difference in mean/max SUV to the standard-count images was observed. VHC images improved image quality for better lesion detectability and clinical diagnosis.

Deep Learning-Based Denoising in Brain Tumor CHO PET: Comparison with Traditional Approaches

18F-choline (CHO) PET image remains noisy despite minimum physiological activity in the normal brain, and this study developed a deep learning-based denoising algorithm for brain tumor CHO PET. Thirty-nine presurgical CHO PET/CT data were retrospectively collected for patients with pathological confirmed primary diffuse glioma. Two conventional denoising methods, namely, block-matching and 3D filtering (BM3D) and non-local means (NLM), and two deep learning-based approaches, namely, Noise2Noise (N2N) and Noise2Void (N2V), were established for imaging denoising, and the methods were developed without paired data. All algorithms improved the image quality to a certain extent, with the N2N demonstrating the best contrast-to-noise ratio (CNR) (4.05 ± 3.45), CNR improvement ratio (13.60% ± 2.05%) and the lowest entropy (1.68 ± 0.17), compared with other approaches. Little changes were identified in traditional tumor PET features including maximum standard uptake value (SUVmax), SUVmean and total lesion activity (TLA), while the tumor-to-normal (T/N ratio) increased thanks to smaller noise. These results suggested that the N2N algorithm can acquire sufficient denoising performance while preserving the original features of tumors, and may be generalized for abundant brain tumor PET images.

Signal-to-Noise Ratio Comparison of Several Filters against Phantom Image

Image denoising methods are important in order to diminish various kinds of noises, which are presented either capturing the image or distorted during image transmission. Signal-to-noise ratio (SNR) is one of the main barriers which avoids the theoretical observations to be accomplished in practice. In this study, we have utilized various kinds of filtering operators against three various noises, which are the signal-to-noise ratio comparison against the phantom image in spatial and frequency domain. In frequency domain, the average filter is used to smooth the image and frequency domain, and Gaussian low-pass filter is applied with empirically determined cutoff frequency. This work has six major parts such as applying average filter, determining the SNR of region of interest, transforming the image in frequency domain by discrete Fourier transform, obtaining the rectangular Gaussian low-pass filter along with a cutoff frequency, multiplying them, and carrying out the inverse Fourier transform. These steps are repeated accordingly until the resulting image SNR is equal to or greater than the spatial domain SNR. In order to achieve the goal of this study, we have analyzed the proposed approach against some of complex phantom images. The performances of these filters are compared against signal-to-noise ratio.

Effect of Denoising and Deblurring 18F-Fluorodeoxyglucose Positron Emission Tomography Images on a Deep Learning Model’s Classification Performance for Alzheimer’s Disease

Alzheimer’s disease (AD) is the most common progressive neurodegenerative disease. 18F-fluorodeoxyglucose positron emission tomography (18F-FDG PET) is widely used to predict AD using a deep learning model. However, the effects of noise and blurring on 18F-FDG PET images were not considered. The performance of a classification model trained using raw, deblurred (by the fast total variation deblurring method), or denoised (by the median modified Wiener filter) 18F-FDG PET images without or with cropping around the limbic system area using a 3D deep convolutional neural network was investigated. The classification model trained using denoised whole-brain 18F-FDG PET images achieved classification performance (0.75/0.65/0.79/0.39 for sensitivity/specificity/F1-score/Matthews correlation coefficient (MCC), respectively) higher than that with raw and deblurred 18F-FDG PET images. The classification model trained using cropped raw 18F-FDG PET images achieved higher performance (0.78/0.63/0.81/0.40 for sensitivity/specificity/F1-score/MCC) than the whole-brain 18F-FDG PET images (0.72/0.32/0.71/0.10 for sensitivity/specificity/F1-score/MCC, respectively). The 18F-FDG PET image deblurring and cropping (0.89/0.67/0.88/0.57 for sensitivity/specificity/F1-score/MCC) procedures were the most helpful for improving performance. For this model, the right middle frontal, middle temporal, insula, and hippocampus areas were the most predictive of AD using the class activation map. Our findings demonstrate that 18F-FDG PET image preprocessing and cropping improves the explainability and potential clinical applicability of deep learning models.