This study presents a novel approach using graph neural networks to predict the risk of internal bleeding using vessel maps derived from patient CT and MRI scans, aimed at enhancing the realism of surgical simulators for emergency scenarios such as trauma, where rapid detection of internal bleeding can be lifesaving. First, medical images are segmented and converted into graph representations of the vasculature, where nodes represent vessel branching points with spatial coordinates and edges encode vessel features such as length and radius. Due to no existing dataset directly labeling bleeding risks, we calculate the bleeding probability for each vessel node using a physics-based heuristic, peripheral vascular resistance via the Hagen-Poiseuille equation. A graph attention network is then trained to regress these probabilities, effectively learning to predict hemorrhage risk from the graph-structured imaging data. The model is trained using a tenfold cross-validation on a combined dataset of 1708 vessel graphs extracted from four public image datasets (MSD, KiTS, AbdomenCT, CT-ORG) with optimization via the Adam optimizer, mean squared error loss, early stopping, and L2 regularization. Our model achieves a mean R-squared of 0.86, reaching up to 0.9188 in optimal configurations and low mean training and validation losses of 0.0069 and 0.0074, respectively, in predicting bleeding risk, with higher performance on well-connected vascular graphs. Finally, we integrate the trained model into an immersive virtual reality environment to simulate intra-abdominal bleeding scenarios for immersive surgical training. The model demonstrates robust predictive performance despite the inherent sparsity of real-life datasets.

Accurate risk stratification is crucial for determining the optimal treatment plan for patients with colorectal cancer (CRC). However, existing deep learning models perform poorly in the preoperative diagnosis of CRC and exhibit limited generalizability, primarily due to insufficient annotated data. To address these issues, CRCFound, a self-supervised learning-based CT image foundation model for CRC is proposed. After pretraining on 5137 unlabeled CRC CT images, CRCFound can learn universal feature representations and provide efficient and reliable adaptability for various clinical applications. Comprehensive benchmark tests are conducted on six different diagnostic tasks and two prognosis tasks to validate the performance of the pretrained model. Experimental results demonstrate that CRCFound can easily transfer to most CRC tasks and exhibit outstanding performance and generalization ability. Overall, CRCFound can solve the problem of insufficient annotated data and perform well in a wide range of downstream tasks of CRC, making it a promising solution for accurate diagnosis and personalized treatment of CRC patients.

Ultrasound (US) imaging is increasingly used in spinal procedures due to its real-time, radiation-free capabilities; however, its effectiveness is hindered by shadowing artifacts that obscure deeper tissue structures. Traditional approaches, such as CT-to-US registration, incorporate anatomical information from preoperative CT scans to guide interventions, but they are limited by complex registration requirements, differences in spine curvature, and the need for recent CT imaging. Recent shape completion methods can offer an alternative by reconstructing spinal structures in US data, while being pretrained on large set of publicly available CT scans. However, these approaches are typically offline and have limited reproducibility. In this work, we introduce a novel integrated system that combines robotic ultrasound with real-time shape completion to enhance spinal visualization. Our robotic platform autonomously acquires US sweeps of the lumbar spine, extracts vertebral surfaces from ultrasound, and reconstructs the complete anatomy using a deep learning-based shape completion network. This framework provides interactive, real-time visualization with the capability to autonomously repeat scans and can enable navigation to target locations. This can contribute to better consistency, reproducibility, and understanding of the underlying anatomy. We validate our approach through quantitative experiments assessing shape completion accuracy and evaluations of multiple spine acquisition protocols on a phantom setup. Additionally, we present qualitative results of the visualization on a volunteer scan.

Most prior unsupervised domain adaptation approaches for medical image segmentation are narrowly tailored to either the source-accessible setting, where adaptation is guided by source-target alignment, or the source-free setting, which typically resorts to implicit supervision mechanisms such as pseudo-labeling and model distillation. This substantial divergence in methodological designs between the two settings reveals an inherent flaw: the lack of an explicit, structured construction of anatomical knowledge that naturally generalizes across domains and settings. To bridge this longstanding divide, we introduce a unified, semantically grounded framework that supports both source-accessible and source-free adaptation. Fundamentally distinct from all prior works, our framework's adaptability emerges naturally as a direct consequence of the model architecture, without the need for any handcrafted adaptation strategies. Specifically, our model learns a domain-agnostic probabilistic manifold as a global space of anatomical regularities, mirroring how humans establish visual understanding. Thus, the structural content in each image can be interpreted as a canonical anatomy retrieved from the manifold and a spatial transformation capturing individual-specific geometry. This disentangled, interpretable formulation enables semantically meaningful prediction with intrinsic adaptability. Extensive experiments on challenging cardiac and abdominal datasets show that our framework achieves state-of-the-art results in both settings, with source-free performance closely approaching its source-accessible counterpart, a level of consistency rarely observed in prior works. Beyond quantitative improvement, we demonstrate strong interpretability of the proposed framework via manifold traversal for smooth shape manipulation.

Accelerating Magnetic Resonance Imaging (MRI) reduces scan time but often degrades image quality. While Implicit Neural Representations (INRs) show promise for MRI reconstruction, they struggle at high acceleration factors due to weak prior constraints, leading to structural loss and aliasing artefacts. To address this, we propose PrIINeR, an INR-based MRI reconstruction method that integrates prior knowledge from pre-trained deep learning models into the INR framework. By combining population-level knowledge with instance-based optimization and enforcing dual data consistency, PrIINeR aligns both with the acquired k-space data and the prior-informed reconstruction. Evaluated on the NYU fastMRI dataset, our method not only outperforms state-of-the-art INR-based approaches but also improves upon several learning-based state-of-the-art methods, significantly improving structural preservation and fidelity while effectively removing aliasing artefacts.PrIINeR bridges deep learning and INR-based techniques, offering a more reliable solution for high-quality, accelerated MRI reconstruction. The code is publicly available on https://github.com/multimodallearning/PrIINeR.

Accurately grounding regions of interest (ROIs) is critical for diagnosis and treatment planning in medical imaging. While multimodal large language models (MLLMs) combine visual perception with natural language, current medical-grounding pipelines still rely on supervised fine-tuning with explicit spatial hints, making them ill-equipped to handle the implicit queries common in clinical practice. This work makes three core contributions. We first define Unified Medical Reasoning Grounding (UMRG), a novel vision-language task that demands clinical reasoning and pixel-level grounding. Second, we release U-MRG-14K, a dataset of 14K samples featuring pixel-level masks alongside implicit clinical queries and reasoning traces, spanning 10 modalities, 15 super-categories, and 108 specific categories. Finally, we introduce MedReasoner, a modular framework that distinctly separates reasoning from segmentation: an MLLM reasoner is optimized with reinforcement learning, while a frozen segmentation expert converts spatial prompts into masks, with alignment achieved through format and accuracy rewards. MedReasoner achieves state-of-the-art performance on U-MRG-14K and demonstrates strong generalization to unseen clinical queries, underscoring the significant promise of reinforcement learning for interpretable medical grounding.

The core role of medical images in disease diagnosis makes their quality directly affect the accuracy of clinical judgment. However, due to factors such as low-dose scanning, equipment limitations and imaging artifacts, medical images are often accompanied by non-uniform noise interference, which seriously affects structure recognition and lesion detection. This paper proposes a medical image adaptive denoising model (MI-ND) that integrates multi-scale convolutional and Transformer architecture, introduces a noise level estimator (NLE) and a noise adaptive attention module (NAAB), and realizes channel-spatial attention regulation and cross-modal feature fusion driven by noise perception. Systematic testing is carried out on multimodal public datasets. Experiments show that this method significantly outperforms the comparative methods in image quality indicators such as PSNR, SSIM, and LPIPS, and improves the F1 score and ROC-AUC in downstream diagnostic tasks, showing strong prac-tical value and promotional potential. The model has outstanding benefits in structural recovery, diagnostic sensitivity, and cross-modal robustness, and provides an effective solution for medical image enhancement and AI-assisted diagnosis and treatment.

Body surface area (BSA)-based chemotherapy dosing remains standard despite its limitations in predicting toxicity. Variations in body composition, particularly skeletal muscle and adipose tissue, influence drug metabolism and toxicity risk. This study aims to investigate the mediating role of body composition in the relationship between BSA-based dosing and dose-limiting toxicities (DLTs) in colorectal cancer patients receiving oxaliplatin-based chemotherapy. We retrospectively analyzed 483 stage III colorectal cancer patients treated at National Cheng Kung University Hospital (2013-2021). An artificial intelligence (AI)-driven algorithm quantified skeletal muscle and adipose tissue compartments from lumbar 3 (L3) vertebral-level computed tomography (CT) scans. Mediation analysis evaluated body composition's role in chemotherapy-related toxicities. Among the cohort, 18.2% (n = 88) experienced DLTs. While BSA alone was not significantly associated with DLTs (OR = 0.473, p = 0.376), increased intramuscular adipose tissue (IMAT) significantly predicted higher DLT risk (OR = 1.047, p = 0.038), whereas skeletal muscle area was protective. Mediation analysis confirmed that IMAT partially mediated the relationship between BSA and DLTs (indirect effect: 0.05, p = 0.040), highlighting adipose infiltration's role in chemotherapy toxicity. BSA-based dosing inadequately accounts for interindividual variations in chemotherapy tolerance. AI-assisted body composition analysis provides a precision oncology framework for identifying high-risk patients and optimizing chemotherapy regimens. Prospective validation is warranted to integrate body composition into routine clinical decision-making.

Deep learning has made notable strides in the ultrasonic diagnosis of lymph nodes, yet it faces three primary challenges: a limited number of lymph node images and a scarcity of annotated data; difficulty in comprehensively learning both local and global semantic information; and obstacles in collaborative learning for both image segmentation and classification to achieve accurate diagnosis. To address these issues, we propose the Cross-organ Cross-modality Cswin-transformer Coupled Convolutional Network (C⁵-Net). First, we design a cross-organ and cross-modality transfer learning strategy to leverage skin lesion dermoscopic images, which have abundant annotations and share similarities in fields of view and morphology with the lymph node ultrasound images. Second, we couple Transformer and convolutional network to comprehensively learn both local details and global information. Third, the encoder weights in the C⁵-Net are shared between segmentation and classification tasks to exploit the synergistic knowledge, enhancing overall performance in ultrasound lymph node diagnosis. Our study leverages 690 lymph node ultrasound images and 1000 skin lesion dermoscopic images. Experimental results show that our C⁵-Net achieves the best segmentation and classification performance for lymph nodes among advanced methods, with the Dice coefficient of segmentation equaling 0.854, and the accuracy of classification equaling 0.874. Our method has consistently shown accuracy and robustness in the segmentation and classification of lymph nodes, contributing to the early and accurate detection of lymph nodal malignancy, which is potentially essential for effective treatment planning in clinical oncology.

Epilepsy affects over 50 million people worldwide, with anti-seizure medications (ASMs) as the primary treatment for seizure control. However, ASM selection remains a "trial and error" process due to the lack of reliable predictors of effectiveness and tolerability. While machine learning approaches have been explored, existing models are limited to predicting outcomes only for ASMs encountered during training and have not leveraged recent biomedical foundation models for this task. This work investigates ASM outcome prediction using only patient MRI scans and reports. Specifically, we leverage biomedical vision-language foundation models and introduce a novel contextualized instruction-tuning framework that integrates expert-built knowledge trees of MRI entities to enhance their performance. Additionally, by training only on the four most commonly prescribed ASMs, our framework enables generalization to predicting outcomes and effectiveness for unseen ASMs not present during training. We evaluate our instruction-tuning framework on two retrospective epilepsy patient datasets, achieving an average AUC of 71.39 and 63.03 in predicting outcomes for four primary ASMs and three completely unseen ASMs, respectively. Our approach improves the AUC by 5.53 and 3.51 compared to standard report-based instruction tuning for seen and unseen ASMs, respectively. Our code, MRI knowledge tree, prompting templates, and TREE-TUNE generated instruction-answer tuning dataset are available at the link.