While deep convolutional neural networks (DCNNs) have achieved remarkable performance in chest X-ray interpretation, their success typically depends on access to large-scale, expertly annotated datasets. However, collecting such data in real-world clinical settings can be difficult because of limited labeling resources, privacy concerns, and patient variability. In this study, we applied a multimodal Transformer pretrained on free-text reports and their paired CXRs to evaluate the effectiveness of this method in settings with limited labeled data. Our dataset consisted of more than 1 million CXRs, each accompanied by reports from board-certified radiologists and 31 structured labels. The results indicated that a linear model trained on embeddings from the pretrained model achieved AUCs of 0.907 and 0.903 on internal and external test sets, respectively, using only 128 cases and 384 controls; the results were comparable those of DenseNet trained on the entire dataset, whose AUCs were 0.908 and 0.903, respectively. Additionally, we demonstrated similar results by extending the application of this approach to a subset annotated with structured echocardiographic reports. Furthermore, this multimodal model exhibited excellent small sample learning capabilities when tested on external validation sets such as CheXpert and ChestX-ray14. This research significantly reduces the sample size necessary for future artificial intelligence advancements in CXR interpretation.

<i>Objective.</i>Alzheimer's disease (AD) is the most prevalent form of dementia worldwide, encompassing a prodromal stage known as mild cognitive impairment (MCI), where patients may either progress to AD or remain stable. The objective of the work was to capture structural and functional modulations of brain structure and function relying on multimodal MRI data and single nucleotide polymorphisms, also in case of missing views, with the twofold goal of classifying AD patients versus healthy controls and detecting MCI converters.<i>Approach.</i>We propose a multimodal deep learning (DL)-based classification framework where a generative module employing cycle generative adversarial networks was introduced in the latent space for imputing missing data (a common issue of multimodal approaches). Explainable AI method was then used to extract input features' relevance allowing for post-hoc validation and enhancing the interpretability of the learned representations.<i>Main results.</i>Experimental results on two tasks, AD detection and MCI conversion, showed that our framework reached competitive performance in the state-of-the-art with an accuracy of0.926±0.02(CI [0.90, 0.95]) and0.711±0.01(CI [0.70, 0.72]) in the two tasks, respectively. The interpretability analysis revealed gray matter modulations in cortical and subcortical brain areas typically associated with AD. Moreover, impairments in sensory-motor and visual resting state networks along the disease continuum, as well as genetic mutations defining biological processes linked to endocytosis, amyloid-beta, and cholesterol, were identified.<i>Significance.</i>Our integrative and interpretable DL approach shows promising performance for AD detection and MCI prediction while shedding light on important biological insights.

Out-of-distribution (OOD) generalization remains a central challenge in deploying deep learning models to real-world scenarios, particularly in domains such as biomedical images, where distribution shifts are both subtle and pervasive. While existing methods often pursue domain invariance through complex generative models or adversarial training, these approaches may overlook the underlying causal mechanisms of generalization.In this work, we propose Causally-Guided Gaussian Perturbations (CGP)-a lightweight framework that enhances OOD generalization by injecting spatially varying noise into input images, guided by soft causal masks derived from Vision Transformers. By applying stronger perturbations to background regions and weaker ones to foreground areas, CGP encourages the model to rely on causally relevant features rather than spurious correlations.Experimental results on the challenging WILDS benchmark Camelyon17 demonstrate consistent performance gains over state-of-the-art OOD baselines, highlighting the potential of causal perturbation as a tool for reliable and interpretable generalization.

We introduce mpLLM, a prompt-conditioned hierarchical mixture-of-experts (MoE) architecture for visual question answering over multi-parametric 3D brain MRI (mpMRI). mpLLM routes across modality-level and token-level projection experts to fuse multiple interrelated 3D modalities, enabling efficient training without image--report pretraining. To address limited image-text paired supervision, mpLLM integrates a synthetic visual question answering (VQA) protocol that generates medically relevant VQA from segmentation annotations, and we collaborate with medical experts for clinical validation. mpLLM outperforms strong medical VLM baselines by 5.3% on average across multiple mpMRI datasets. Our study features three main contributions: (1) the first clinically validated VQA dataset for 3D brain mpMRI, (2) a novel multimodal LLM that handles multiple interrelated 3D modalities, and (3) strong empirical results that demonstrate the medical utility of our methodology. Ablations highlight the importance of modality-level and token-level experts and prompt-conditioned routing. We have included our source code in the supplementary materials and will release our dataset upon publication.

Ultrasound is crucial in modern medicine but faces challenges like operator dependence, image noise, and real-time scanning, hindering AI integration. While large multimodal models excel in other medical imaging areas, they struggle with ultrasound's complexities. To address this, we introduce Dolphin v1.0 (V1) and its reasoning-augmented version, Dolphin R1-the first large-scale multimodal ultrasound foundation models unifying diverse clinical tasks in a single vision-language framework.To tackle ultrasound variability and noise, we curated a 2-million-scale multimodal dataset, combining textbook knowledge, public data, synthetic samples, and general corpora. This ensures robust perception, generalization, and clinical adaptability.The Dolphin series employs a three-stage training strategy: domain-specialized pretraining, instruction-driven alignment, and reinforcement-based refinement. Dolphin v1.0 delivers reliable performance in classification, detection, regression, and report generation. Dolphin R1 enhances diagnostic inference, reasoning transparency, and interpretability through reinforcement learning with ultrasound-specific rewards.Evaluated on U2-Bench across eight ultrasound tasks, Dolphin R1 achieves a U2-score of 0.5835-over twice the second-best model (0.2968) setting a new state of the art. Dolphin v1.0 also performs competitively, validating the unified framework. Comparisons show reasoning-enhanced training significantly improves diagnostic accuracy, consistency, and interpretability, highlighting its importance for high-stakes medical AI.

The error is caused by special characters that arXiv's system doesn't recognize. Here's the cleaned version with all problematic characters replaced: Breast magnetic resonance imaging is a critical tool for cancer detection and treatment planning, but its clinical utility is hindered by poor specificity, leading to high false-positive rates and unnecessary biopsies. This study introduces a transformer-based framework for automated classification of breast lesions in dynamic contrast-enhanced MRI, addressing the challenge of distinguishing benign from malignant findings. We implemented a SegFormer architecture that achieved an AUC of 0.92 for lesion-level classification, with 100% sensitivity and 67% specificity at the patient level - potentially eliminating one-third of unnecessary biopsies without missing malignancies. The model quantifies malignant pixel distribution via semantic segmentation, producing interpretable spatial predictions that support clinical decision-making. To establish reproducible benchmarks, we curated BreastDCEDL_AMBL by transforming The Cancer Imaging Archive's AMBL collection into a standardized deep learning dataset with 88 patients and 133 annotated lesions (89 benign, 44 malignant). This resource addresses a key infrastructure gap, as existing public datasets lack benign lesion annotations, limiting benign-malignant classification research. Training incorporated an expanded cohort of over 1,200 patients through integration with BreastDCEDL datasets, validating transfer learning approaches despite primary tumor-only annotations. Public release of the dataset, models, and evaluation protocols provides the first standardized benchmark for DCE-MRI lesion classification, enabling methodological advancement toward clinical deployment.

Liver fibrosis represents the accumulation of excessive extracellular matrix caused by sustained hepatic injury. It disrupts normal lobular architecture and function, increasing the chances of cirrhosis and liver failure. Precise staging of fibrosis for early diagnosis and intervention is often invasive, which carries risks and complications. To address this challenge, recent advances in artificial intelligence-based liver segmentation and fibrosis staging offer a non-invasive alternative. As a result, the CARE 2025 Challenge aimed for automated methods to quantify and analyse liver fibrosis in real-world scenarios, using multi-centre, multi-modal, and multi-phase MRI data. This challenge included tasks of precise liver segmentation (LiSeg) and fibrosis staging (LiFS). In this study, we developed an automated pipeline for both tasks across all the provided MRI modalities. This pipeline integrates pseudo-labelling based on multi-modal co-registration, liver segmentation using deep neural networks, and liver fibrosis staging based on shape, textural, appearance, and directional (STAD) features derived from segmentation masks and MRI images. By solely using the released data with limited annotations, our proposed pipeline demonstrated excellent generalisability for all MRI modalities, achieving top-tier performance across all competition subtasks. This approach provides a rapid and reproducible framework for quantitative MRI-based liver fibrosis assessment, supporting early diagnosis and clinical decision-making. Code is available at https://github.com/YangForever/care2025_liver_biodreamer.

Vision-grounded medical report generation aims to produce clinically accurate descriptions of medical images, anchored in explicit visual evidence to improve interpretability and facilitate integration into clinical workflows. However, existing methods often rely on separately trained detection modules that require extensive expert annotations, introducing high labeling costs and limiting generalizability due to pathology distribution bias across datasets. To address these challenges, we propose Self-Supervised Anatomical Consistency Learning (SS-ACL) -- a novel and annotation-free framework that aligns generated reports with corresponding anatomical regions using simple textual prompts. SS-ACL constructs a hierarchical anatomical graph inspired by the invariant top-down inclusion structure of human anatomy, organizing entities by spatial location. It recursively reconstructs fine-grained anatomical regions to enforce intra-sample spatial alignment, inherently guiding attention maps toward visually relevant areas prompted by text. To further enhance inter-sample semantic alignment for abnormality recognition, SS-ACL introduces a region-level contrastive learning based on anatomical consistency. These aligned embeddings serve as priors for report generation, enabling attention maps to provide interpretable visual evidence. Extensive experiments demonstrate that SS-ACL, without relying on expert annotations, (i) generates accurate and visually grounded reports -- outperforming state-of-the-art methods by 10\% in lexical accuracy and 25\% in clinical efficacy, and (ii) achieves competitive performance on various downstream visual tasks, surpassing current leading visual foundation models by 8\% in zero-shot visual grounding.

Knee Osteoarthritis (K-OA) is characterized as a progressive joint condition with global prevalence, exhibiting deterioration over time and impacting a significant portion of the population. It happens because joints wear out slowly. The main reason for osteoarthritis is the wearing away of the cushion in the joints, which makes the bones rub together. This causes feelings of stiffness, unease, and difficulty moving. Persons with osteoarthritis find it hard to do simple things like walking, standing, or going up stairs. Besides that, it can also make people feel sad or worried because of the ongoing pain and trouble it causes. Knee osteoarthritis exerts a sustained impact on both the economy and society. Typically, radiologists assess knee health through MRI or X-ray images, assigning KL-grades. MRI excels in visualizing soft tissues like cartilage, menisci, and ligaments, directly revealing cartilage degeneration and joint inflammation crucial for osteoarthritis (OA) diagnosis. In contrast, X-rays primarily show bone, only inferring cartilage loss through joint space narrowing-a late indicator of OA. This makes MRI superior for detecting early changes and subtle lesions often missed by X-rays. However, manual diagnosis of Knee osteoarthritis is laborious and time-consuming. In response, deep learning methodologies such as vision transformer (ViT) has been implemented to enhance efficiency and streamline workflows in clinical settings. This research leverages ViT for Knee Osteoarthritis KL grading, achieving an accuracy of 88%. It illustrates that employing a simple transfer learning technique with this model yields superior performance compared to more intricate architectures.

Amyloid PET/CT is essential for quantifying amyloid-beta (Aβ) deposition in Alzheimer's disease (AD), with the Centiloid (CL) scale standardizing measurements across imaging centers. However, MRI-based CL pipelines face challenges: high cost, contraindications, and patient burden. To address these challenges, we developed a deep learning-based CT parcellation pipeline calibrated to the standard CL scale using CT images from PET/CT scans and evaluated its performance relative to standard pipelines. A total of 306 participants (23 young controls [YCs] and 283 patients) underwent 18 F-florbetaben (FBB) PET/CT and MRI. Based on visual assessment, 207 patients were classified as Aβ-positive and 76 as Aβ-negative. PET images were processed using the CT parcellation pipeline and compared to FreeSurfer (FS) and standard pipelines. Agreement was assessed via regression analyses. Effect size, variance, and ROC analyses were used to compare pipelines and determine the optimal CL threshold relative to visual Aβ assessment. The CT parcellation showed high concordance with the FS and provided reliable CL quantification (R² = 0.99). Both pipelines demonstrated similar variance in YCs and effect sizes between YCs and ADCI. ROC analyses confirmed comparable accuracy and similar CL thresholds, supporting CT parcellation as a viable MRI-free alternative. Our findings indicate that the CT parcellation pipeline achieves a level of accuracy similar to FS in CL quantification, demonstrating its reliability as an MRI-free alternative. In PET/CT, CT and PET are acquired sequentially within the same session on a shared bed and headrest, which helps maintain consistent positioning and adequate spatial alignment, reducing registration errors and supporting more reliable and precise quantification.