AI-based neural decoding reconstructs visual perception by leveraging generative models to map brain activity measured through functional MRI (fMRI) into the observed visual stimulus. Traditionally, ridge linear models transform fMRI into a latent space, which is then decoded using variational autoencoders (VAE) or latent diffusion models (LDM). Owing to the complexity and noisiness of fMRI data, newer approaches split the reconstruction into two sequential stages, the first one providing a rough visual approximation using a VAE, the second one incorporating semantic information through the adoption of LDM guided by contrastive language-image pre-training (CLIP) embeddings. This work addressed some key scientific and technical gaps of the two-stage neural decoding by: 1) implementing a gated recurrent unit (GRU)-based architecture to establish a non-linear mapping between the fMRI signal and the VAE latent space, 2) optimizing the dimensionality of the VAE latent space, 3) systematically evaluating the contribution of the first reconstruction stage, and 4) analyzing the impact of different brain regions of interest (ROIs) on reconstruction quality. Experiments on the Natural Scenes Dataset, containing 73,000 unique natural images, along with fMRI of eight subjects, demonstrated that the proposed architecture maintained competitive performance while reducing the complexity of its first stage by 85%. The sensitivity analysis showcased that the first reconstruction stage is essential for preserving high structural similarity in the final reconstructions. Restricting analysis to semantic ROIs, while excluding early visual areas, diminished visual coherence, preserving semantics though. The inter-subject repeatability across ROIs was about 92 and 98% for visual and sematic metrics, respectively. This study represents a key step toward optimized neural decoding architectures leveraging non-linear models for stimulus prediction. Sensitivity analysis highlighted the interplay between the two reconstruction stages, while ROI-based analysis provided strong evidence that the two-stage AI model reflects the brain's hierarchical processing of visual information.

With the rapid development of artificial intelligence (AI), AI-assisted medical imaging analysis demonstrates remarkable performance in early lung cancer screening. However, the costly annotation process and privacy concerns limit the construction of large-scale medical datasets, hampering the further application of AI in healthcare. To address the data scarcity in lung cancer screening, we propose Lung-DDPM, a thoracic CT image synthesis approach that effectively generates high-fidelity 3D synthetic CT images, which prove helpful in downstream lung nodule segmentation tasks. Our method is based on semantic layout-guided denoising diffusion probabilistic models (DDPM), enabling anatomically reasonable, seamless, and consistent sample generation even from incomplete semantic layouts. Our results suggest that the proposed method outperforms other state-of-the-art (SOTA) generative models in image quality evaluation and downstream lung nodule segmentation tasks. Specifically, Lung-DDPM achieved superior performance on our large validation cohort, with a Fréchet inception distance (FID) of 0.0047, maximum mean discrepancy (MMD) of 0.0070, and mean squared error (MSE) of 0.0024. These results were 7.4×, 3.1×, and 29.5× better than the second-best competitors, respectively. Furthermore, the lung nodule segmentation model, trained on a dataset combining real and Lung-DDPM-generated synthetic samples, attained a Dice Coefficient (Dice) of 0.3914 and sensitivity of 0.4393. This represents 8.8% and 18.6% improvements in Dice and sensitivity compared to the model trained solely on real samples. The experimental results highlight Lung-DDPM's potential for a broader range of medical imaging applications, such as general tumor segmentation, cancer survival estimation, and risk prediction. The code and pretrained models are available at https://github.com/Manem-Lab/Lung-DDPM/.

Multi-Modal Medical Image Fusion (MMMIF) has become increasingly important in clinical applications, as it enables the integration of complementary information from different imaging modalities to support more accurate diagnosis and treatment planning. The primary objective of Medical Image Fusion (MIF) is to generate a fused image that retains the most informative features from the Source Images (SI), thereby enhancing the reliability of clinical decision-making systems. However, due to inherent limitations in individual imaging modalities-such as poor spatial resolution in functional images or low contrast in anatomical scans-fused images can suffer from information degradation or distortion. To address these limitations, this study proposes a novel fusion framework that integrates the Non-Subsampled Shearlet Transform (NSST) with a Convolutional Neural Network (CNN) for effective sub-band enhancement and image reconstruction. Initially, each source image is decomposed into Low-Frequency Coefficients (LFC) and multiple High-Frequency Coefficients (HFC) using NSST. The proposed Concurrent Denoising and Enhancement Network (CDEN) is then applied to these sub-bands to suppress noise and enhance critical structural details. The enhanced LFCs are fused using an AlexNet-based activity-level fusion model, while the enhanced HFCs are combined using a Pulse Coupled Neural Network (PCNN) guided by a Novel Sum-Modified Laplacian (NSML) metric. Finally, the fused image is reconstructed via Inverse-NSST (I-NSST). Experimental results prove that the proposed method outperforms existing fusion algorithms, achieving approximately 16.5% higher performance in terms of the QAB/F (edge preservation) metric, along with strong results across both subjective visual assessments and objective quality indices.

Multimodal medical image fusion (MMIF) aims to integrate images from different modalities to produce a comprehensive image that enhances medical diagnosis by accurately depicting organ structures, tissue textures, and metabolic information. Capturing both the unique and complementary information across multiple modalities simultaneously is a key research challenge in MMIF. To address this challenge, this paper proposes a novel image fusion method, MMIF-AMIN, which features a new architecture that can effectively extract these unique and complementary features. Specifically, an Invertible Dense Network (IDN) is employed for lossless feature extraction from individual modalities. To extract complementary information between modalities, a Multi-scale Complementary Feature Extraction Module (MCFEM) is designed, which incorporates a hybrid attention mechanism, convolutional layers of varying sizes, and Transformers. An adaptive loss function is introduced to guide model learning, addressing the limitations of traditional manually-designed loss functions and enhancing the depth of data mining. Extensive experiments demonstrate that MMIF-AMIN outperforms nine state-of-the-art MMIF methods, delivering superior results in both quantitative and qualitative analyses. Ablation experiments confirm the effectiveness of each component of the proposed method. Additionally, extending MMIF-AMIN to other image fusion tasks also achieves promising performance.

Generative artificial intelligence (AI) has been playing an important role in various domains. Leveraging its high capability to generate high-fidelity and diverse synthetic data, generative AI is widely applied in diagnostic tasks, such as lung cancer diagnosis using computed tomography (CT). However, existing generative models for lung cancer diagnosis suffer from low efficiency and anatomical imprecision, which limit their clinical applicability. To address these drawbacks, we propose Lung-DDPM+, an improved version of our previous model, Lung-DDPM. This novel approach is a denoising diffusion probabilistic model (DDPM) guided by nodule semantic layouts and accelerated by a pulmonary DPM-solver, enabling the method to focus on lesion areas while achieving a better trade-off between sampling efficiency and quality. Evaluation results on the public LIDC-IDRI dataset suggest that the proposed method achieves 8$\times$ fewer FLOPs (floating point operations per second), 6.8$\times$ lower GPU memory consumption, and 14$\times$ faster sampling compared to Lung-DDPM. Moreover, it maintains comparable sample quality to both Lung-DDPM and other state-of-the-art (SOTA) generative models in two downstream segmentation tasks. We also conducted a Visual Turing Test by an experienced radiologist, showing the advanced quality and fidelity of synthetic samples generated by the proposed method. These experimental results demonstrate that Lung-DDPM+ can effectively generate high-quality thoracic CT images with lung nodules, highlighting its potential for broader applications, such as general tumor synthesis and lesion generation in medical imaging. The code and pretrained models are available at https://github.com/Manem-Lab/Lung-DDPM-PLUS.

We propose a deep learning-based approach that integrates MRI sequence parameters to improve the accuracy and generalizability of quantitative image synthesis from clinical weighted MRI. Our physics-driven neural network embeds MRI sequence parameters -- repetition time (TR), echo time (TE), and inversion time (TI) -- directly into the model via parameter embedding, enabling the network to learn the underlying physical principles of MRI signal formation. The model takes conventional T1-weighted, T2-weighted, and T2-FLAIR images as input and synthesizes T1, T2, and proton density (PD) quantitative maps. Trained on healthy brain MR images, it was evaluated on both internal and external test datasets. The proposed method achieved high performance with PSNR values exceeding 34 dB and SSIM values above 0.92 for all synthesized parameter maps. It outperformed conventional deep learning models in accuracy and robustness, including data with previously unseen brain structures and lesions. Notably, our model accurately synthesized quantitative maps for these unseen pathological regions, highlighting its superior generalization capability. Incorporating MRI sequence parameters via parameter embedding allows the neural network to better learn the physical characteristics of MR signals, significantly enhancing the performance and reliability of quantitative MRI synthesis. This method shows great potential for accelerating qMRI and improving its clinical utility.

Despite significant progress in generative modelling, existing diffusion models often struggle to produce anatomically precise female pelvic images, limiting their application in gynaecological imaging, where data scarcity and patient privacy concerns are critical. To overcome these barriers, we introduce a novel diffusion-based framework for uterine MRI synthesis, integrating both unconditional and conditioned Denoising Diffusion Probabilistic Models (DDPMs) and Latent Diffusion Models (LDMs) in 2D and 3D. Our approach generates anatomically coherent, high fidelity synthetic images that closely mimic real scans and provide valuable resources for training robust diagnostic models. We evaluate generative quality using advanced perceptual and distributional metrics, benchmarking against standard reconstruction methods, and demonstrate substantial gains in diagnostic accuracy on a key classification task. A blinded expert evaluation further validates the clinical realism of our synthetic images. We release our models with privacy safeguards and a comprehensive synthetic uterine MRI dataset to support reproducible research and advance equitable AI in gynaecology.

Atrial fibrillation (AF) is a prevalent cardiac arrhythmia often treated with catheter ablation procedures, but procedural outcomes are highly variable. Evaluating and improving ablation efficacy is challenging due to the complex interaction between patient-specific tissue and procedural factors. This paper asks two questions: Can AF recurrence be predicted by simulating the effects of procedural parameters? How should we ablate to reduce AF recurrence? We propose SOFA (Simulating and Optimizing Atrial Fibrillation Ablation), a novel deep-learning framework that addresses these questions. SOFA first simulates the outcome of an ablation strategy by generating a post-ablation image depicting scar formation, conditioned on a patient's pre-ablation LGE-MRI and the specific procedural parameters used (e.g., ablation locations, duration, temperature, power, and force). During this simulation, it predicts AF recurrence risk. Critically, SOFA then introduces an optimization scheme that refines these procedural parameters to minimize the predicted risk. Our method leverages a multi-modal, multi-view generator that processes 2.5D representations of the atrium. Quantitative evaluations show that SOFA accurately synthesizes post-ablation images and that our optimization scheme leads to a 22.18\% reduction in the model-predicted recurrence risk. To the best of our knowledge, SOFA is the first framework to integrate the simulation of procedural effects, recurrence prediction, and parameter optimization, offering a novel tool for personalizing AF ablation.

Acute ischemic stroke (AIS) is a leading cause of death and long-term disability worldwide, where rapid reperfusion remains critical for salvaging brain tissue. Although CT perfusion (CTP) imaging provides essential hemodynamic information, its limitations-including extended processing times, additional radiation exposure, and variable software outputs-can delay treatment. In contrast, non-contrast head CT (NCHCT) is ubiquitously available in acute stroke settings. This study explores a generative artificial intelligence approach to predict key perfusion parameters (relative cerebral blood flow [rCBF] and time-to-maximum [Tmax]) directly from NCHCT, potentially streamlining stroke imaging workflows and expanding access to critical perfusion data. We retrospectively identified patients evaluated for AIS who underwent NCHCT, CT angiography, and CTP. Ground truth perfusion maps (rCBF and Tmax) were extracted from VIZ.ai post-processed CTP studies. A modified pix2pix-turbo generative adversarial network (GAN) was developed to translate co-registered NCHCT images into corresponding perfusion maps. The network was trained using paired NCHCT-CTP data, with training, validation, and testing splits of 80%:10%:10%. Performance was assessed on the test set using quantitative metrics including the structural similarity index measure (SSIM), peak signal-to-noise ratio (PSNR), and Fréchet inception distance (FID). Out of 120 patients, studies from 99 patients fitting our inclusion and exclusion criteria were used as the primary cohort (mean age 73.3 ± 13.5 years; 46.5% female). Cerebral occlusions were predominantly in the middle cerebral artery. GAN-generated Tmax maps achieved an SSIM of 0.827, PSNR of 16.99, and FID of 62.21, while the rCBF maps demonstrated comparable performance (SSIM 0.79, PSNR 16.38, FID 59.58). These results indicate that the model approximates ground truth perfusion maps to a moderate degree and successfully captures key cerebral hemodynamic features. Our findings demonstrate the feasibility of generating functional perfusion maps directly from widely available NCHCT images using a modified GAN. This cross-modality approach may serve as a valuable adjunct in AIS evaluation, particularly in resource-limited settings or when traditional CTP provides limited diagnostic information. Future studies with larger, multicenter datasets and further model refinements are warranted to enhance clinical accuracy and utility.

Generative artificial intelligence models facilitate open-data sharing by proposing synthetic data as surrogates of real patient data. Despite the promise for healthcare, some of these models are susceptible to patient data memorization, where models generate patient data copies instead of novel synthetic samples, resulting in patient re-identification. Here we assess memorization in unconditional latent diffusion models by training them on a variety of datasets for synthetic data generation and detecting memorization with a self-supervised copy detection approach. We show a high degree of patient data memorization across all datasets, with approximately 37.2% of patient data detected as memorized and 68.7% of synthetic samples identified as patient data copies. Latent diffusion models are more susceptible to memorization than autoencoders and generative adversarial networks, and they outperform non-diffusion models in synthesis quality. Augmentation strategies during training, small architecture size and increasing datasets can reduce memorization, while overtraining the models can enhance it. These results emphasize the importance of carefully training generative models on private medical imaging datasets and examining the synthetic data to ensure patient privacy.