The growing heterogeneity of cardiac patient data from hospitals and wearables necessitates predictive models that are tailored, comprehensible, and safeguard privacy. This study introduces PerFed-Cardio, a lightweight and interpretable semi-federated learning (Semi-FL) system for real-time cardiovascular risk stratification utilizing multimodal data, including cardiac imaging, physiological signals, and electronic health records (EHR). In contrast to conventional federated learning, where all clients engage uniformly, our methodology employs a personalized Semi-FL approach that enables high-capacity nodes (e.g., hospitals) to conduct comprehensive training, while edge devices (e.g., wearables) refine shared models via modality-specific subnetworks. Cardiac MRI and echocardiography pictures are analyzed via lightweight convolutional neural networks enhanced with local attention modules to highlight diagnostically significant areas. Physiological characteristics (e.g., ECG, activity) and EHR data are amalgamated through attention-based fusion layers. Model transparency is attained using Local Interpretable Model-agnostic Explanations (LIME) and Grad-CAM, which offer spatial and feature-level elucidations for each prediction. Assessments on authentic multimodal datasets from 123 patients across five simulated institutions indicate that PerFed-Cardio attains an AUC-ROC of 0.972 with an inference latency of 130 ms. The customized model calibration and targeted training diminish communication load by 28%, while maintaining an F1-score over 92% in noisy situations. These findings underscore PerFed-Cardio as a privacy-conscious, adaptive, and interpretable system for scalable cardiac risk assessment.

ST-elevation myocardial infarction (STEMI) is a major cardiac event that requires rapid reperfusion therapy. The same reperfusion mechanism that minimizes infarct size and mortality may paradoxically exacerbate further cardiac damage-a condition known as reperfusion injury. Oxidative stress, calcium excess, mitochondrial malfunction, and programmed cell death mechanisms make myocardial dysfunction worse. Even with the best revascularization techniques, reperfusion damage still jeopardizes the long-term prognosis and myocardial healing. A thorough narrative review was carried out using some of the most well-known scientific databases, including ScienceDirect, PubMed, and Google Scholar. With an emphasis on pathophysiological causes, clinical manifestations, innovative biomarkers, imaging modalities, artificial intelligence applications, and developing treatment methods related to reperfusion injury, peer-reviewed publications published between 2015 and 2025 were highlighted. The review focuses on the molecular processes that underlie cardiac reperfusion injury, such as reactive oxygen species, calcium dysregulation, opening of the mitochondrial permeability transition pore, and several types of programmed cell death. Clinical syndromes such as myocardial stunning, coronary no-reflow, and intramyocardial hemorrhage are thoroughly studied-all of which lead to negative consequences like heart failure and left ventricular dysfunction. Cardiac magnetic resonance imaging along with coronary angiography and significant biomarkers like N-terminal proBNP and soluble ST2 aid in risk stratification and prognosis. In addition to mechanical techniques like ischemia postconditioning and remote ischemic conditioning, pharmacological treatments are also examined. Despite promising research findings, the majority of therapies have not yet proven consistently effective in extensive clinical studies. Consideration of sex-specific risk factors, medicines that target the mitochondria, tailored therapies, and the use of artificial intelligence for risk assessment and early diagnosis are some potential future avenues. Reperfusion damage continues to be a significant obstacle to the best possible recovery after STEMI, even with improvements in revascularization. The management of STEMI still relies heavily on early reperfusion, although adjuvant medicines that target reperfusion injury specifically are desperately needed. Molecular-targeted approaches, AI-driven risk assessment, and precision medicine advancements have the potential to reduce cardiac damage and enhance long-term outcomes for patients with STEMI.

MRI assessment for extraprostatic extension (EPE) of prostate cancer (PCa) is challenging due to limited accuracy and interobserver agreement. To develop an interpretable Tabular Prior-data Fitted Network (TabPFN)-based radiomics model to evaluate EPE using MRI and explore its integration with radiologists' assessments. Retrospective. Five hundred and thirteen consecutive patients who underwent radical prostatectomy. Four hundred and eleven patients from center 1 (mean age 67 ± 7 years) formed training (287 patients) and internal test (124 patients) sets, and 102 patients from center 2 (mean age 66 ± 6 years) were assigned as an external test set. Three Tesla, fast spin echo T2-weighted imaging (T2WI) and diffusion-weighted imaging using single-shot echo planar imaging. Radiomics features were extracted from T2WI and apparent diffusion coefficient maps, and the TabRadiomics model was developed using TabPFN. Three machine learning models served as baseline comparisons: support vector machine, random forest, and categorical boosting. Two radiologists (with > 1500 and > 500 prostate MRI interpretations, respectively) independently evaluated EPE grade on MRI. Artificial intelligence (AI)-modified EPE grading algorithms incorporating the TabRadiomics model with radiologists' interpretations of curvilinear contact length and frank EPE were simulated. Receiver operating characteristic curve (AUC), Delong test, and McNemar test. p < 0.05 was considered significant. The TabRadiomics model performed comparably to machine learning models in both internal and external tests, with AUCs of 0.806 (95% CI, 0.727-0.884) and 0.842 (95% CI, 0.770-0.912), respectively. AI-modified algorithms showed significantly higher accuracies compared with the less experienced reader in internal testing, with up to 34.7% of interpretations requiring no radiologist input. However, no difference was observed in both readers in the external test set. The TabRadiomics model demonstrated high performance in EPE assessment and may improve clinical assessment in PCa. 4. Stage 2.

Predicting tumor regression grade (TRG) after neoadjuvant chemoradiotherapy (NCRT) in patients with locally advanced rectal cancer (LARC) preoperatively accurately is crucial for providing individualized treatment plans. This study aims to develop transrectal contrast-enhanced ultrasound-based (TR-CEUS) radiomics models for predicting TRG. A total of 190 LARC patients undergoing NCRT and subsequent total mesorectal excision were categorized into good and poor response groups based on pathological TRG. TR-CEUS examinations were conducted before and after NCRT. Machine learning (ML) models for predicting TRG were developed by employing pre- and post-NCRT TR-CEUS image series, based on seven classifiers, including random forest (RF), multi-layer perceptron (MLP) and so on. The predictive performance of models was evaluated using receiver operating characteristic curve analysis and Delong test. A total of 1525 TR-CEUS images were included for analysis, and 3360 ML models were constructed using image series before and after NCRT, respectively. The optimal pre-NCRT ML model, constructed from imaging series before NCRT, was RF; whereas the optimal post-NCRT model, derived from imaging series after NCRT, was MLP. The areas under the curve for the optimal RF and MLP models demonstrated values of 0.609 and 0.857, respectively, in the cross-validation cohort, with corresponding values of 0.659 and 0.841 observed in the independent test cohort. Delong tests showed that the predictive efficacy of the post-NCRT model was statistically higher than that of the pre-NCRT model (p < 0.05). Radiomics model developed by TR-CEUS images after NCRT demonstrated high predictive performance for TRG, thereby facilitating precise evaluation of therapeutic response to NCRT in LARC patients.

Intracranial aneurysms (IAs) are common vascular pathologies with a risk of fatal rupture. Human assessment of rupture risk is error prone, and treatment decision for unruptured IAs often rely on expert opinion and institutional policy. Therefore, we aimed to develop a computer-assisted aneurysm rupture prediction framework to help guide the decision-making process and create future decision criteria. This retrospective study included 335 patients with 500 IAs, of the 500 IAs studied, 250 were labeled as ruptured and 250 as unruptured. A skilled radiologist and a neurosurgeon visually examined the computed tomography angiography (CTA) images and labeled the IAs. For external validation we included 24 IAs, 10 ruptured and 15 unruptured, imaged with 3D rotational angiography (3D-RA) from the Aneurisk dataset. The pretrained nnU-net model was used for automated vessel segmentation, which was fed to pretrained PointNet++ models for vessel labeling and aneurysm segmentation. From these the latent keypoint representations were extracted as vessel shape and aneurysm shape features, respectively. Additionally, conventional features such as IAs morphological measurements, location and patient data, such as age, sex, were used for training and testing eight machine learning models for rupture status classification. The top-performing model, a random forest with feature selection, achieved an area under the receiver operating curve of 0.851, an accuracy of 0.782, a sensitivity of 0.804, and a specificity of 0.760. This model used 14 aneurysm shape features, seven conventional features, and one vessel shape feature. On the external dataset, it achieved an AUC of 0.805. While aneurysm shape features consistently contributed significantly across the classification models, vessel shape features contributed a small portion. Our proposed automated artificial intelligence framework could assist in clinical decision-making by assessing aneurysm rupture risk using screening tests, such as CTA and 3D-RA.

Vision foundation models have demonstrated vast potential in achieving generalist medical segmentation capability, providing a versatile, task-agnostic solution through a single model. However, current generalist models involve simple pre-training on various medical data containing irrelevant information, often resulting in the negative transfer phenomenon and degenerated performance. Furthermore, the practical applicability of foundation models across diverse open-world scenarios, especially in out-of-distribution (OOD) settings, has not been extensively evaluated. Here we construct a publicly accessible database, MedSegDB, based on a tree-structured hierarchy and annotated from 129 public medical segmentation repositories and 5 in-house datasets. We further propose a Generalist Medical Segmentation model (MedSegX), a vision foundation model trained with a model-agnostic Contextual Mixture of Adapter Experts (ConMoAE) for open-world segmentation. We conduct a comprehensive evaluation of MedSegX across a range of medical segmentation tasks. Experimental results indicate that MedSegX achieves state-of-the-art performance across various modalities and organ systems in in-distribution (ID) settings. In OOD and real-world clinical settings, MedSegX consistently maintains its performance in both zero-shot and data-efficient generalization, outperforming other foundation models.

In radiomics, feature selection methods are primarily used to eliminate redundant features and identify relevant ones. Feature projection methods, such as principal component analysis (PCA), are often avoided due to concerns that recombining features may compromise interpretability. However, since most radiomic features lack inherent semantic meaning, prioritizing interpretability over predictive performance may not be justified. This study investigates whether feature projection methods can improve predictive performance compared to feature selection, as measured by the area under the receiver operating characteristic curve (AUC), the area under the precision-recall curve (AUPRC), and the F1, F0.5 and F2 scores. Models were trained on a large collection of 50 binary classification radiomic datasets derived from CT and MRI of various organs and representing different clinical outcomes. Evaluation was performed using nested, stratified 5-fold cross-validation with 10 repeats. Nine feature projection methods, including PCA, Kernel PCA, and Non-Negative Matrix Factorization (NMF), were compared to nine selection methods, such as Minimum Redundancy Maximum Relevance (MRMRe), Extremely Randomized Trees (ET), and LASSO, using four classifiers. The results showed that selection methods, particularly ET, MRMRe, Boruta, and LASSO, achieved the highest overall performance. Importantly, performance varied considerably across datasets, and some projection methods, such as NMF, occasionally outperformed all selection methods on individual datasets, indicating their potential utility. However, the average difference between selection methods and projection methods across all datasets was negligible and statistically insignificant, suggesting that both perform similarly based solely on methodological considerations. These findings support the notion that, in a typical radiomics study, selection methods should remain the primary approach but also emphasize the importance of considering projection methods in order to achieve the highest performance.

This study systematically evaluates the diagnostic performance of artificial intelligence (AI)-driven and conventional radiomics models in detecting muscle-invasive bladder cancer (MIBC) through meta-analytical approaches. Furthermore, it investigates their potential synergistic value with the Vesical Imaging-Reporting and Data System (VI-RADS) and assesses clinical translation prospects. This study adhered to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. We conducted a comprehensive systematic search of PubMed, Web of Science, Embase, and Cochrane Library databases up to May 13, 2025, and manually screened the references of included studies. The quality and risk of bias of the selected studies were assessed using the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) and Radiomics Quality Score (RQS) tools. We pooled the area under the curve (AUC), sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), and their 95% confidence intervals (95% CI). Additionally, meta-regression and subgroup analyses were performed to identify potential sources of heterogeneity. This meta-analysis incorporated 43 studies comprising 9624 patients. The majority of included studies demonstrated low risk of bias, with a mean RQS of 18.89. Pooled analysis yielded an AUC of 0.92 (95% CI: 0.89-0.94). The aggregate sensitivity and specificity were both 0.86 (95% CI: 0.84-0.87), with heterogeneity indices of I² = 43.58 and I² = 72.76, respectively. The PLR was 5.97 (95% CI: 5.28-6.75, I² = 64.04), while the NLR was 0.17 (95% CI: 0.15-0.19, I² = 37.68). The DOR reached 35.57 (95% CI: 29.76-42.51, I² = 99.92). Notably, all included studies exhibited significant heterogeneity (P < 0.1). Meta-regression and subgroup analyses identified several significant sources of heterogeneity, including: study center type (single-center vs. multi-center), sample size (<100 vs. ≥100 patients), dataset classification (training, validation, testing, or ungrouped), imaging modality (computed tomography [CT] vs. magnetic resonance imaging [MRI]), modeling algorithm (deep learning vs. machine learning vs. other), validation methodology (cross-validation vs. cohort validation), segmentation method (manual vs. [semi]automated), regional differences (China vs. other countries), and risk of bias (high vs. low vs. unclear). AI-driven and traditional radiomic models have exhibited robust diagnostic performance for MIBC. Nevertheless, substantial heterogeneity across studies necessitates validation through multinational, multicenter prospective cohort studies to establish external validity.

Vision-language models and their adaptations to image segmentation tasks present enormous potential for producing highly accurate and interpretable results. However, implementations based on CLIP and BiomedCLIP are still lagging behind more sophisticated architectures such as CRIS. In this work, instead of focusing on text prompt engineering as is the norm, we attempt to narrow this gap by showing how to ensemble vision-language segmentation models (VLSMs) with a low-complexity CNN. By doing so, we achieve a significant Dice score improvement of 6.3% on the BKAI polyp dataset using the ensembled BiomedCLIPSeg, while other datasets exhibit gains ranging from 1% to 6%. Furthermore, we provide initial results on additional four radiology and non-radiology datasets. We conclude that ensembling works differently across these datasets (from outperforming to underperforming the CRIS model), indicating a topic for future investigation by the community. The code is available at https://github.com/juliadietlmeier/VLSM-Ensemble.

Background: To systematically review and perform a meta-analysis of artificial intelligence (AI)-driven methods for detecting and correcting magnetic resonance imaging (MRI) motion artifacts, assessing current developments, effectiveness, challenges, and future research directions. Methods: A comprehensive systematic review and meta-analysis were conducted, focusing on deep learning (DL) approaches, particularly generative models, for the detection and correction of MRI motion artifacts. Quantitative data were extracted regarding utilized datasets, DL architectures, and performance metrics. Results: DL, particularly generative models, show promise for reducing motion artifacts and improving image quality; however, limited generalizability, reliance on paired training data, and risk of visual distortions remain key challenges that motivate standardized datasets and reporting. Conclusions: AI-driven methods, particularly DL generative models, show significant potential for improving MRI image quality by effectively addressing motion artifacts. However, critical challenges must be addressed, including the need for comprehensive public datasets, standardized reporting protocols for artifact levels, and more advanced, adaptable DL techniques to reduce reliance on extensive paired datasets. Addressing these aspects could substantially enhance MRI diagnostic accuracy, reduce healthcare costs, and improve patient care outcomes.