This paper develops dimension reduction techniques for accelerating diffusion model inference in the context of synthetic data generation. The idea is to integrate compressed sensing into diffusion models: (i) compress the data into a latent space, (ii) train a diffusion model in the latent space, and (iii) apply a compressed sensing algorithm to the samples generated in the latent space, facilitating the efficiency of both model training and inference. Under suitable sparsity assumptions on data, the proposed algorithm is proved to enjoy faster convergence by combining diffusion model inference with sparse recovery. As a byproduct, we obtain an optimal value for the latent space dimension. We also conduct numerical experiments on a range of datasets, including image data (handwritten digits, medical images, and climate data) and financial time series for stress testing.

Accurate quantification of tau pathology via tau positron emission tomography (PET) scan is crucial for diagnosing and monitoring Alzheimer's disease (AD). However, the high cost and limited availability of tau PET restrict its widespread use. In contrast, structural magnetic resonance imaging (MRI) and plasma-based biomarkers provide non-invasive and widely available complementary information related to brain anatomy and disease progression. In this work, we propose a text-guided 3D diffusion model for 3D tau PET image synthesis, leveraging multimodal conditions from both structural MRI and plasma measurement. Specifically, the textual prompt is from the plasma p-tau217 measurement, which is a key indicator of AD progression, while MRI provides anatomical structure constraints. The proposed framework is trained and evaluated using clinical AV1451 tau PET data from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database. Experimental results demonstrate that our approach can generate realistic, clinically meaningful 3D tau PET across a range of disease stages. The proposed framework can help perform tau PET data augmentation under different settings, provide a non-invasive, cost-effective alternative for visualizing tau pathology, and support the simulation of disease progression under varying plasma biomarker levels and cognitive conditions.

Generating realistic MRIs to accurately predict future changes in the structure of brain is an invaluable tool for clinicians in assessing clinical outcomes and analysing the disease progression at the patient level. However, current existing methods present some limitations: (i) some approaches fail to explicitly capture the relationship between structural changes and time intervals, especially when trained on age-imbalanced datasets; (ii) others rely only on scan interpolation, which lack clinical utility, as they generate intermediate images between timepoints rather than future pathological progression; and (iii) most approaches rely on 2D slice-based architectures, thereby disregarding full 3D anatomical context, which is essential for accurate longitudinal predictions. We propose a 3D Temporally-Aware Diffusion Model (TADM-3D), which accurately predicts brain progression on MRI volumes. To better model the relationship between time interval and brain changes, TADM-3D uses a pre-trained Brain-Age Estimator (BAE) that guides the diffusion model in the generation of MRIs that accurately reflect the expected age difference between baseline and generated follow-up scans. Additionally, to further improve the temporal awareness of TADM-3D, we propose the Back-In-Time Regularisation (BITR), by training TADM-3D to predict bidirectionally from the baseline to follow-up (forward), as well as from the follow-up to baseline (backward). Although predicting past scans has limited clinical applications, this regularisation helps the model generate temporally more accurate scans. We train and evaluate TADM-3D on the OASIS-3 dataset, and we validate the generalisation performance on an external test set from the NACC dataset. The code will be available upon acceptance.

Magnetic resonance imaging of fetal and neonatal brains reveals rapid neurodevelopment marked by substantial anatomical changes unfolding within days. Studying this critical stage of the developing human brain, therefore, requires accurate brain models-referred to as atlases-of high spatial and temporal resolution. To meet these demands, established traditional atlases and recently proposed deep learning-based methods rely on large and comprehensive datasets. This poses a major challenge for studying brains in the presence of pathologies for which data remains scarce. We address this limitation with CINeMA (Conditional Implicit Neural Multi-Modal Atlas), a novel framework for creating high-resolution, spatio-temporal, multimodal brain atlases, suitable for low-data settings. Unlike established methods, CINeMA operates in latent space, avoiding compute-intensive image registration and reducing atlas construction times from days to minutes. Furthermore, it enables flexible conditioning on anatomical features including gestational age, birth age, and pathologies like agenesis of the corpus callosum and ventriculomegaly of varying degree. CINeMA supports downstream tasks such as tissue segmentation and age prediction whereas its generative properties enable synthetic data creation and anatomically informed data augmentation. Surpassing state-of-the-art methods in accuracy, efficiency, and versatility, CINeMA represents a powerful tool for advancing brain research. We release the code and atlases at https://github.com/m-dannecker/CINeMA.

Low-field portable magnetic resonance imaging (pMRI) devices address a crucial requirement in the realm of healthcare by offering the capability for on-demand and timely access to MRI, especially in the context of routine stroke emergency. Nevertheless, images acquired by these devices often exhibit poor clarity and low resolution, resulting in their reduced potential to support precise diagnostic evaluations and lesion quantification. In this paper, we propose a 3D deep learning based model, named Stroke-Aware CycleGAN (SA-CycleGAN), to enhance the quality of low-field images for further improving diagnosis of routine stroke. Firstly, based on traditional CycleGAN, SA-CycleGAN incorporates a prior of stroke lesions by applying a novel spatial feature transform mechanism. Secondly, gradient difference losses are combined to deal with the problem that the synthesized images tend to be overly smooth. We present a dataset comprising 101 paired high-field and low-field diffusion-weighted imaging (DWI), which were acquired through dual scans of the same patient in close temporal proximity. Our experiments demonstrate that SA-CycleGAN is capable of generating images with higher quality and greater clarity compared to the original low-field DWI. Additionally, in terms of quantifying stroke lesions, SA-CycleGAN outperforms existing methods. The lesion volume exhibits a strong correlation between the generated images and the high-field images, with R=0.852. In contrast, the lesion volume correlation between the low-field images and the high-field images is notably lower, with R=0.462. Furthermore, the mean absolute difference in lesion volumes between the generated images and high-field images (1.73±2.03 mL) was significantly smaller than the difference between the low-field images and high-field images (2.53±4.24 mL). It shows that the synthesized images not only exhibit superior visual clarity compared to the low-field acquired images, but also possess a high degree of consistency with high-field images. In routine clinical practice, the proposed SA-CycleGAN offers an accessible and cost-effective means of rapidly obtaining higher-quality images, holding the potential to enhance the efficiency and accuracy of stroke diagnosis in routine clinical settings. The code and trained models will be released on GitHub: SA-CycleGAN.

The early and precise diagnosis of stroke plays an important role in its treatment planning. Computed Tomography (CT) is utilised as a first diagnostic tool for quick diagnosis and to rule out haemorrhage. Diffusion Magnetic Resonance Imaging (MRI) provides superior sensitivity in comparison to CT for detecting early acute ischaemia and small lesions. However, the long scan time and limited availability of MRI make it not feasible for emergency settings. To deal with this problem, this study presents a brain mask-guided and fidelity-constrained cycle-consistent generative adversarial network for translating CT images into diffusion MRI images for stroke diagnosis. A brain mask is concatenated with the input CT image and given as input to the generator to encourage more focus on the critical foreground areas. A fidelity-constrained loss is utilised to preserve details for better translation results. A publicly available dataset, A Paired CT-MRI Dataset for Ischemic Stroke Segmentation (APIS) is utilised to train and test the models. The proposed method yields MSE 197.45 [95% CI: 180.80, 214.10], PSNR 25.50 [95% CI: 25.10, 25.92], and SSIM 88.50 [95% CI: 87.50, 89.50] on a testing set. The proposed method significantly improves techniques based on UNet, cycle-consistent generative adversarial networks (CycleGAN) and Attention generative adversarial networks (GAN). Furthermore, an ablation study was performed, which demonstrates the effectiveness of incorporating fidelity-constrained loss and brain mask information as a soft guide in translating CT images into diffusion MRI images. The experimental results demonstrate that the proposed approach has the potential to support faster and precise diagnosis of stroke.

To overcome the scarcity of annotated dental X-ray datasets, this study presents a novel pipeline for generating high-resolution synthetic orthopantomography (OPG) images using customized generative adversarial networks (GANs). A total of 4777 real OPG images were collected from clinical centres in Pakistan, Thailand, and the U.S., covering diverse anatomical features. Twelve GAN models were initially trained, with four top-performing variants selected for further training on both combined and region-specific datasets. Synthetic images were generated at 2048 × 1024 pixels, maintaining fine anatomical detail. The evaluation was conducted using (1) a YOLO-based object detection model trained on real OPGs to assess feature representation via mean average precision, and (2) expert dentist scoring for anatomical and diagnostic realism. All selected models produced realistic synthetic OPGs. The YOLO detector achieved strong performance on these images, indicating accurate structural representation. Expert evaluations confirmed high anatomical plausibility, with models M1 and M3 achieving over 50% of the reference scores assigned to real OPGs. The developed GAN-based pipeline enables the ethical and scalable creation of synthetic OPG images, suitable for augmenting datasets used in artificial intelligence-driven dental diagnostics. This method provides a practical solution to data limitations in dental artificial intelligence, supporting model development in privacy-sensitive or low-resource environments.

This study explored a generative image synthesis method based on diffusion models, potentially providing a low-cost and high-efficiency training data augmentation strategy for medical artificial intelligence (AI) applications. The MedMNIST v2 dataset was utilized as a small-volume training dataset under low-performance computing conditions. Based on the characteristics of existing samples, new medical images were synthesized using the proposed annotated diffusion model. In addition to observational assessment, quantitative evaluation was performed based on the gradient descent of the loss function during the generation process and the Fréchet Inception Distance (FID), using various loss functions and feature vector dimensions. Compared to the original data, the proposed diffusion model successfully generated medical images of similar styles but with dramatically varied anatomic details. The model trained with the Huber loss function achieved a higher FID of 15.2 at a feature vector dimension of 2048, compared with the model trained with the L2 loss function, which achieved the best FID of 0.85 at a feature vector dimension of 64. The use of the Huber loss enhanced model robustness, while FID values indicated acceptable similarity between generated and real images. Future work should explore the application of these models to more complex datasets and clinical scenarios. This study demonstrated that diffusion model-based medical image synthesis is potentially applicable as an augmentation strategy for AI, particularly in situations where access to real clinical data is limited. Optimal training parameters were also proposed by evaluating the dimensionality of feature vectors in FID calculations and the complexity of loss functions.

This study aims to explore the feasibility of employing generative adversarial networks (GAN) to generate synthetic contrast-enhanced ultrasound (CEUS) from grayscale ultrasound images of patients with thyroid nodules while dispensing with the need for ultrasound contrast agents injection. Patients who underwent preoperative thyroid CEUS examinations between January 2020 and July 2022 were collected retrospectively. The cycle-GAN framework integrated paired and unpaired learning modules was employed to develop the non-invasive image generation process. The synthetic CEUS images was generated in three phases: pre-arterial, plateau, and venous. The evaluation included quantitative similarity metrics, classification performance, and qualitative assessment by radiologists. CEUS videos of 360 thyroid nodules from 314 patients (45 years ± 12 [SD]; 272 women) in the internal dataset and 202 thyroid nodules from 183 patients (46 years ± 13 [SD]; 148 women) in the external dataset were included. In the external testing dataset, quantitative analysis revealed a significant degree of similarity between real and synthetic CEUS images (structure similarity index, 0.89 ± 0.04; peak signal-to-noise ratio, 28.17 ± 2.42). Radiologists deemed 126 of 132 [95%] synthetic CEUS images diagnostically useful. The accuracy of radiologists in distinguishing between real and synthetic images was 55.6% (95% CI: 0.49, 0.63), with an AUC of 61.0% (95% CI: 0.65, 0.68). No statistically significant difference (p > 0.05) was observed when radiologists assessed peak intensity and enhancement patterns using real CEUS and synthetic CEUS. Both quantitative analysis and radiologist evaluations exhibited that synthetic CEUS images generated by generative adversarial networks were similar to real CEUS images. QuestionIt is feasible to generate synthetic thyroid contrast-enhanced ultrasound images using generative adversarial networks without ultrasound contrast agents injection. FindingsCompared to real contrast-enhanced ultrasound images, synthetic contrast-enhanced ultrasound images exhibited high similarity and image quality. Clinical relevanceThis non-invasive and intelligent transformation may reduce the requirement for ultrasound contrast agents in certain cases, particularly in scenarios where ultrasound contrast agents administration is contraindicated, such as in patients with allergies, poor tolerance, or limited access to resources.

Reducing PET scan acquisition time to minimize motion-related artifacts and improving patient comfort is always demanding. This study proposes a deep-learning framework for synthesizing diagnostic-quality PET images from ultrafast scans in multi-tracer total-body PET imaging. A retrospective analysis was conducted on clinical uEXPLORER PET/CT datasets from a single institution, including [¹⁸F]FDG (N=50), [¹⁸F]FAPI (N=45) and [⁶⁸Ga]FAPI (N=60) studies. Standard 300-s acquisitions were performed for each patient, with ultrafast scan PET images (3, 6, 15, 30, and 40 s) generated through list-mode data truncation. We developed two variants of a 3D SwinUNETR-V2 architecture: Model 1 (PET-only input) and Model 2 (PET+CT fusion input). The proposed methodology was trained and tested on all three datasets using 5-fold cross-validation. The proposed Model 1 and Model 2 significantly enhanced subjective image quality and lesion detectability in multi-tracer PET images compared to the original ultrafast scans. Model 1 and Model 2 also improved objective image quality metrics. For the [¹⁸F]FDG datasets, both approaches improved peak signal-to-noise ratio (PSNR) metrics across ultra-short acquisitions: 3 s: 48.169±6.121 (Model 1) vs. 48.123±6.103 (Model 2) vs. 44.092±7.508 (ultrafast), p < 0.001; 6 s: 48.997±5.960 vs. 48.461±5.897 vs. 46.503±7.190, p < 0.001; 15 s: 50.310±5.674 vs. 50.042±5.734 vs. 49.331±6.732, p < 0.001. The proposed Model 1 and Model 2 effectively enhance image quality of multi-tracer total-body PET scans with ultrafast acquisition times. The predicted PET images demonstrate comparable performance in terms of image quality and lesion detectability.