Medical image segmentation is a critical task in clinical workflows, particularly for the detection and delineation of pathological regions. While convolutional architectures like U-Net have become standard for such tasks, their limited receptive field restricts global context modeling. Recent efforts integrating transformers have addressed this, but often result in deep, computationally expensive models unsuitable for real-time use. In this work, we present a novel end-to-end lightweight architecture designed specifically for real-time binary medical image segmentation. Our model combines a Swin Transformer-like encoder with a U-Net-like decoder, connected via skip pathways to preserve spatial detail while capturing contextual information. Unlike existing designs such as Swin Transformer or U-Net, our architecture is significantly shallower and competitively efficient. To improve the encoder's ability to learn meaningful features without relying on large amounts of labeled data, we first train it using Barlow Twins, a self-supervised learning method that helps the model focus on important patterns by reducing unnecessary repetition in the learned features. After this pretraining, we fine-tune the entire model for our specific task. Experiments on benchmark binary segmentation tasks demonstrate that our model achieves competitive accuracy with substantially reduced parameter count and faster inference, positioning it as a practical alternative for deployment in real-time and resource-limited clinical environments. The code for our method is available at Github repository: https://github.com/mkianih/Barlow-Swin.

Vision foundation models like DINOv2 demonstrate remarkable potential in medical imaging despite their origin in natural image domains. However, their design inherently works best for uni-modal image analysis, limiting their effectiveness for multi-modal imaging tasks that are common in many medical fields, such as neurology and oncology. While supervised models perform well in this setting, they fail to leverage unlabeled datasets and struggle with missing modalities, a frequent challenge in clinical settings. To bridge these gaps, we introduce MM-DINOv2, a novel and efficient framework that adapts the pre-trained vision foundation model DINOv2 for multi-modal medical imaging. Our approach incorporates multi-modal patch embeddings, enabling vision foundation models to effectively process multi-modal imaging data. To address missing modalities, we employ full-modality masking, which encourages the model to learn robust cross-modality relationships. Furthermore, we leverage semi-supervised learning to harness large unlabeled datasets, enhancing both the accuracy and reliability of medical predictions. Applied to glioma subtype classification from multi-sequence brain MRI, our method achieves a Matthews Correlation Coefficient (MCC) of 0.6 on an external test set, surpassing state-of-the-art supervised approaches by +11.1%. Our work establishes a scalable and robust solution for multi-modal medical imaging tasks, leveraging powerful vision foundation models pre-trained on natural images while addressing real-world clinical challenges such as missing data and limited annotations.

Maternal-fetal Ultrasound is the primary modality for monitoring fetal development, yet automated segmentation remains challenging due to the scarcity of high-quality annotations. To address this limitation, we propose a semi-supervised learning framework that leverages information divergence for robust fetal ultrasound segmentation. Our method employs a lightweight convolutional network (1.47M parameters) and a Transformer-based network, trained jointly with labelled data through standard supervision and with unlabelled data via cross-supervision. To encourage consistent and confident predictions, we introduce an information divergence loss that combines per-pixel Kullback-Leibler divergence and Mutual Information Gap, effectively reducing prediction disagreement between the two models. In addition, we apply mixup on unlabelled samples to further enhance robustness. Experiments on two fetal ultrasound datasets demonstrate that our approach consistently outperforms seven state-of-the-art semi-supervised methods. When only 5% of training data is labelled, our framework improves the Dice score by 2.39%, reduces the 95% Hausdorff distance by 14.90, and decreases the Average Surface Distance by 4.18. These results highlight the effectiveness of leveraging information divergence for annotation-efficient and robust medical image segmentation. Our code is publicly available on GitHub.

The advent of large-scale vision foundation models, pre-trained on diverse natural images, has marked a paradigm shift in computer vision. However, how the frontier vision foundation models' efficacies transfer to specialized domains remains such as medical imaging remains an open question. This report investigates whether DINOv3, a state-of-the-art self-supervised vision transformer (ViT) that features strong capability in dense prediction tasks, can directly serve as a powerful, unified encoder for medical vision tasks without domain-specific pre-training. To answer this, we benchmark DINOv3 across common medical vision tasks, including 2D/3D classification and segmentation on a wide range of medical imaging modalities. We systematically analyze its scalability by varying model sizes and input image resolutions. Our findings reveal that DINOv3 shows impressive performance and establishes a formidable new baseline. Remarkably, it can even outperform medical-specific foundation models like BiomedCLIP and CT-Net on several tasks, despite being trained solely on natural images. However, we identify clear limitations: The model's features degrade in scenarios requiring deep domain specialization, such as in Whole-Slide Pathological Images (WSIs), Electron Microscopy (EM), and Positron Emission Tomography (PET). Furthermore, we observe that DINOv3 does not consistently obey scaling law in the medical domain; performance does not reliably increase with larger models or finer feature resolutions, showing diverse scaling behaviors across tasks. Ultimately, our work establishes DINOv3 as a strong baseline, whose powerful visual features can serve as a robust prior for multiple complex medical tasks. This opens promising future directions, such as leveraging its features to enforce multiview consistency in 3D reconstruction.

Brain Metastases (BM) are a large contributor to mortality of patients with cancer. They are treated with Stereotactic Radiosurgery (SRS) and monitored with Magnetic Resonance Imaging (MRI) at regular follow-up intervals according to treatment guidelines. Analyzing and quantifying this longitudinal imaging represents an intractable workload for clinicians. As a result, follow-up images are not annotated and merely assessed by observation. Response to treatment in longitudinal imaging is being studied, to better understand growth trajectories and ultimately predict treatment success or toxicity as early as possible. In this study, we implement an automated pipeline to curate a large longitudinal dataset of SRS treatment data, resulting in a cohort of 896 BMs in 177 patients who were monitored for >360 days at approximately two-month intervals at Lausanne University Hospital (CHUV). We use a data-driven clustering to identify characteristic trajectories. In addition, we predict 12 months lesion-level response using classical as well as graph machine learning Graph Machine Learning (GML). Clustering revealed 5 dominant growth trajectories with distinct final response categories. Response prediction reaches up to 0.90 AUC (CI95%=0.88-0.92) using only pre-treatment and first follow-up MRI with gradient boosting. Similarly, robust predictive performance of up to 0.88 AUC (CI95%=0.86-0.90) was obtained using GML, offering more flexibility with a single model for multiple input time-points configurations. Our results suggest potential automation and increased precision for the comprehensive assessment and prediction of BM response to SRS in longitudinal MRI. The proposed pipeline facilitates scalable data curation for the investigation of BM growth patterns, and lays the foundation for clinical decision support systems aiming at optimizing personalized care.

Accurate evaluation of the response of glioblastoma to therapy is crucial for clinical decision-making and patient management. The Response Assessment in Neuro-Oncology (RANO) criteria provide a standardized framework to assess patients' clinical response, but their application can be complex and subject to observer variability. This paper presents an automated method for classifying the intervention response from longitudinal MRI scans, developed to predict tumor response during therapy as part of the BraTS 2025 challenge. We propose a novel hybrid framework that combines deep learning derived feature extraction and an extensive set of radiomics and clinically chosen features. Our approach utilizes a fine-tuned ResNet-18 model to extract features from 2D regions of interest across four MRI modalities. These deep features are then fused with a rich set of more than 4800 radiomic and clinically driven features, including 3D radiomics of tumor growth and shrinkage masks, volumetric changes relative to the nadir, and tumor centroid shift. Using the fused feature set, a CatBoost classifier achieves a mean ROC AUC of 0.81 and a Macro F1 score of 0.50 in the 4-class response prediction task (Complete Response, Partial Response, Stable Disease, Progressive Disease). Our results highlight that synergizing learned image representations with domain-targeted radiomic features provides a robust and effective solution for automated treatment response assessment in neuro-oncology.

We aimed to develop and validate a radiomics-based machine learning nomogram using multiparametric magnetic resonance imaging to preoperatively predict substantial lymphovascular space invasion in patients with endometrial cancer. This retrospective dual-center study included patients with histologically confirmed endometrial cancer who underwent preoperative magnetic resonance imaging (MRI). The patients were divided into training and test sets. Radiomic features were extracted from multiparametric magnetic resonance imaging to generate radiomic scores using a support vector machine algorithm. Three predictive models were constructed: clinical (Model^C), radiomics-only (Model^R), and fusion (Model^N). The models' performances were evaluated by analyzing their receiver operating characteristic curves, and pairwise comparisons of the models' areas under the curves were conducted using DeLong's test and adjusted using the Bonferroni correction. Decision curve analysis with integrated discrimination improvement was used for net benefit comparison. This study enrolled 283 women (training set: n = 198; test set: n = 85). The lymphovascular space invasion groups (substantial and no/focal) had significantly different radiomic scores (P < 0.05). Model^N achieved an area under the curve of 0.818 (95% confidence interval: 0.757-0.869) and 0.746 (95% confidence interval: 0.640-0.835) for the training and test sets, respectively, demonstrating a good fit according to the Hosmer-Lemeshow test (P > 0.05). The DeLong test with Bonferroni correction indicated that Model^N demonstrated better diagnostic efficiency than Model^C in predicting substantial lymphovascular space invasion in the two sets (adjusted P < 0.05). In addition, decision curve analysis demonstrated a higher net benefit for Model^N, with integrated discrimination improvements of 0.043 and 0.732 (P < 0.01) in the training and test sets, respectively. The multiparametric magnetic resonance imaging-based radiomics machine learning nomogram showed moderate diagnostic performance for substantial lymphovascular space invasion in patients with endometrial cancer.

Point-of-care ultrasonography has become a valuable tool for assessing diaphragmatic function in critically ill patients receiving invasive mechanical ventilation. However, conventional diaphragm ultrasound assessment remains highly operator-dependent and subjective. Previous research introduced automatic measurement of diaphragmatic excursion and velocity using 2D speckle-tracking technology. This study aimed to develop an artificial intelligence-multimodal learning framework to improve the prediction of weaning failure and guide individualized weaning strategies. This prospective study enrolled critically ill patients older than 18 years who received mechanical ventilation for more than 48 hours and were eligible for a spontaneous breathing trial in 2 intensive care units in Guangzhou, China. Before the spontaneous breathing trial, diaphragm ultrasound videos were collected using a standardized protocol, and automatic measurements of excursion and velocity were obtained. A total of 88 patients were included, with 50 successfully weaned and 38 experiencing weaning failure. Each patient record included 27 clinical and 6 diaphragmatic indicators, selected based on previous literature and phenotyping studies. Clinical variables were preprocessed using OneHotEncoder, normalization, and scaling. Ultrasound videos were interpolated to a uniform resolution of 224×224×96. Artificial intelligence-multimodal learning based on clinical characteristics, laboratory parameters, and diaphragm ultrasonic videos was established. Four experiments were conducted in an ablation setting to evaluate model performance using different combinations of input data: (1) diaphragmatic excursion only, (2) clinical and diaphragmatic indicators, (3) ultrasound videos only, and (4) all modalities combined (multimodal). Metrics for evaluation included classification accuracy, area under the receiver operating characteristic curve (AUC), average precision in the precision-recall curve, and calibration curve. Variable importance was assessed using SHAP (Shapley Additive Explanation) to interpret feature contributions and understand model predictions. The multimodal co-learning model outperformed all single-modal approaches. The accuracy improved when predicted through diaphragm ultrasound video data using Video Vision Transformer (accuracy=0.8095, AUC=0.852), clinical or ultrasound indicators (accuracy=0.7381, AUC=0.746), and the multimodal co-learning (accuracy=0.8331, AUC=0.894). The proposed co-learning model achieved the highest score (average precision=0.91) among the 4 experiments. Furthermore, calibration curve analysis demonstrated that the proposed colearning model was well calibrated, as the curve was closest to the perfectly calibrated line. Combining ultrasound and clinical data for colearning improved the accuracy of the weaning outcome prediction. Multimodal learning based on automatic measurement of point-of-care ultrasonography and automated collection of objective clinical indicators greatly enhanced the practical operability and user-friendliness of the system. The proposed model offered promising potential for widespread clinical application in intensive care settings.

Fat fraction (FF) quantification in individual muscles using quantitative MRI is of major importance for monitoring disease progression and assessing disease severity in neuromuscular diseases. Undersampling of MRI acquisitions is commonly used to reduce scanning time. The present paper introduces novel unrolled neural networks for the reconstruction of undersampled MRI acquisitions. These networks are designed with the aim of maintaining accurate FF quantification while reducing reconstruction time and memory usage. The proposed approach relies on a combination of a simplified architecture (Half U-Net) with unrolled networks that achieved high performance in the well-known FastMRI challenge (variational network [VarNet] and densely interconnected residual cascading network [DIRCN]). The algorithms were trained and evaluated using 3D MRI Dixon acquisitions of the thigh from controls and patients with neuromuscular diseases. The study was performed by applying a retrospective undersampling with acceleration factors of 4 and 8. Reconstructed images were used to computed FF maps. Results disclose that the novel unrolled neural networks were able to maintain reconstruction, biomarker assessment, and segmentation quality while reducing memory usage by 24% to 16% and reducing reconstruction time from 21% to 17%. Using an acceleration factor of 8, the proposed algorithms, HalfVarNet and HalfDIRCN, achieved structural similarity index (SSIM) scores of 93.76 ± 0.38 and 94.95 ± 0.32, mean squared error (MSE) values of 12.76 ± 1.08 × 10^-2 and 10.25 ± 0.87 × 10^-2, and a relative FF quadratic error of 0.23 ± 0.02% and 0.17 ± 0.02%, respectively. The proposed method enables time and memory-efficient reconstruction of undersampled 3D MRI data, supporting its potential for clinical application.

This retrospective-prospective study evaluated whether a deep learning-based MRI reconstruction algorithm can preserve diagnostic quality in brain MRI scans accelerated up to fourfold, using both public and prospective clinical data. The study included 18 healthy volunteers (scans acquired at 3T, January 2024-March 2025), as well as selected fastMRI public datasets with diverse pathologies. Phase-encoding-undersampled 2D/3D T1, T2, and FLAIR sequences were reconstructed with DeepFoqus-Accelerate and compared with standard-of-care (SOC). Three board-certified neuroradiologists and two MRI technologists independently reviewed 36 paired SOC/AI reconstructions from both datasets using a 5-point Likert scale, while quantitative similarity was assessed for 408 scans and 1224 datasets using Structural Similarity Index (SSIM), Peak Signal-to-Noise Ratio (PSNR), and Haar wavelet-based Perceptual Similarity Index (HaarPSI). No AI-reconstructed scan scored below 3 (minimally acceptable), and 95% scored $\geq 4$. Mean SSIM was 0.95 $\pm$ 0.03 (90% cases >0.90), PSNR >41.0 dB, and HaarPSI >0.94. Inter-rater agreement was slight to moderate. Rare artifacts did not affect diagnostic interpretation. These findings demonstrate that DeepFoqus-Accelerate enables robust fourfold brain MRI acceleration with 75% reduced scan time, while preserving diagnostic image quality and supporting improved workflow efficiency.