Clinical decision-making relies heavily on understanding relative positions of anatomical structures and anomalies. Therefore, for Vision-Language Models (VLMs) to be applicable in clinical practice, the ability to accurately determine relative positions on medical images is a fundamental prerequisite. Despite its importance, this capability remains highly underexplored. To address this gap, we evaluate the ability of state-of-the-art VLMs, GPT-4o, Llama3.2, Pixtral, and JanusPro, and find that all models fail at this fundamental task. Inspired by successful approaches in computer vision, we investigate whether visual prompts, such as alphanumeric or colored markers placed on anatomical structures, can enhance performance. While these markers provide moderate improvements, results remain significantly lower on medical images compared to observations made on natural images. Our evaluations suggest that, in medical imaging, VLMs rely more on prior anatomical knowledge than on actual image content for answering relative position questions, often leading to incorrect conclusions. To facilitate further research in this area, we introduce the MIRP , Medical Imaging Relative Positioning, benchmark dataset, designed to systematically evaluate the capability to identify relative positions in medical images.

Classification models that provide human-interpretable explanations enhance clinicians' trust and usability in medical image diagnosis. One research focus is the integration and prediction of pathology-related visual attributes used by radiologists alongside the diagnosis, aligning AI decision-making with clinical reasoning. Radiologists use attributes like shape and texture as established diagnostic criteria and mirroring these in AI decision-making both enhances transparency and enables explicit validation of model outputs. However, the adoption of such models is limited by the scarcity of large-scale medical image datasets annotated with these attributes. To address this challenge, we propose synthesizing attribute-annotated data using a generative model. We enhance the Diffusion Model with attribute conditioning and train it using only 20 attribute-labeled lung nodule samples from the LIDC-IDRI dataset. Incorporating its generated images into the training of an explainable model boosts performance, increasing attribute prediction accuracy by 13.4% and target prediction accuracy by 1.8% compared to training with only the small real attribute-annotated dataset. This work highlights the potential of synthetic data to overcome dataset limitations, enhancing the applicability of explainable models in medical image analysis.

Estimating the elasticity of soft tissue can provide useful information for various diagnostic applications. Ultrasound shear wave elastography offers a non-invasive approach. However, its generalizability and standardization across different systems and processing pipelines remain limited. Considering the influence of image processing on ultrasound based diagnostics, recent literature has discussed the impact of different image processing steps on reliable and reproducible elasticity analysis. In this work, we investigate the need of ultrasound pre-processing steps for deep learning-based ultrasound shear wave elastography. We evaluate the performance of a 3D convolutional neural network in predicting shear wave velocities from spatio-temporal ultrasound images, studying different degrees of pre-processing on the input images, ranging from fully beamformed and filtered ultrasound images to raw radiofrequency data. We compare the predictions from our deep learning approach to a conventional time-of-flight method across four gelatin phantoms with different elasticity levels. Our results demonstrate statistically significant differences in the predicted shear wave velocity among all elasticity groups, regardless of the degree of pre-processing. Although pre-processing slightly improves performance metrics, our results show that the deep learning approach can reliably differentiate between elasticity groups using raw, unprocessed radiofrequency data. These results show that deep learning-based approaches could reduce the need for and the bias of traditional ultrasound pre-processing steps in ultrasound shear wave elastography, enabling faster and more reliable clinical elasticity assessments.

The lumbar spine and pelvis are critical stress-bearing structures of the human body, and their rapid and accurate segmentation plays a vital role in clinical diagnosis and intervention. However, conventional CT imaging poses significant challenges due to the low contrast of sacral and bilateral hip tissues and the complex and highly similar intervertebral space structures within the lumbar spine. To address these challenges, we propose a general-purpose segmentation network that integrates a cross-scale feature fusion strategy with a linear self-attention mechanism. The proposed network effectively extracts multi-scale features and fuses them along the channel dimension, enabling both structural and boundary information of lumbar and pelvic regions to be captured within the encoder-decoder architecture.Furthermore, we introduce a linear mapping strategy to approximate the traditional attention matrix with a low-rank representation, allowing the linear attention mechanism to significantly reduce computational complexity while maintaining segmentation accuracy for vertebrae and pelvic bones. Comparative and ablation experiments conducted on the CTSpine1K and CTPelvic1K datasets demonstrate that our method achieves improvements of 1.5% in Dice Similarity Coefficient (DSC) and 2.6% in Hausdorff Distance (HD) over state-of-the-art models, validating the effectiveness of our approach in enhancing boundary segmentation quality and segmentation accuracy in homogeneous anatomical regions.

Accurate segmentation of small lesions in Breast Dynamic Contrast-Enhanced MRI (DCE-MRI) is critical for early cancer detection, especially in high-risk patients. While recent deep learning methods have advanced lesion segmentation, they primarily target large lesions and neglect valuable longitudinal and clinical information routinely used by radiologists. In real-world screening, detecting subtle or emerging lesions requires radiologists to compare across timepoints and consider previous radiology assessments, such as the BI-RADS score. We propose LesiOnTime, a novel 3D segmentation approach that mimics clinical diagnostic workflows by jointly leveraging longitudinal imaging and BIRADS scores. The key components are: (1) a Temporal Prior Attention (TPA) block that dynamically integrates information from previous and current scans; and (2) a BI-RADS Consistency Regularization (BCR) loss that enforces latent space alignment for scans with similar radiological assessments, thus embedding domain knowledge into the training process. Evaluated on a curated in-house longitudinal dataset of high-risk patients with DCE-MRI, our approach outperforms state-of-the-art single-timepoint and longitudinal baselines by 5% in terms of Dice. Ablation studies demonstrate that both TPA and BCR contribute complementary performance gains. These results highlight the importance of incorporating temporal and clinical context for reliable early lesion segmentation in real-world breast cancer screening. Our code is publicly available at https://github.com/cirmuw/LesiOnTime

Subvariants of testicular germ cell tumor (TGCT) significantly affect therapeutic strategies and patient prognosis. However, preoperatively distinguishing seminoma (SE) from non-seminoma (n-SE) remains a challenge. This study aimed to evaluate the performance of a deep learning-based super-resolution (SR) US radiomics model for SE/n-SE differentiation. This international multicenter retrospective study recruited patients with confirmed TGCT between 2015 and 2023. A pre-trained SR reconstruction algorithm was applied to enhance native resolution (NR) images. NR and SR radiomics models were constructed, and the superior model was then integrated with clinical features to construct clinical-radiomics models. Diagnostic performance was evaluated by ROC analysis (AUC) and compared with radiologists' assessments using the DeLong test. A total of 486 male patients were enrolled for training (n = 338), domestic (n = 92), and international (n = 59) validation sets. The SR radiomics model achieved AUCs of 0.90, 0.82, and 0.91, respectively, in the training, domestic, and international validation sets, significantly surpassing the NR model (p < 0.001, p = 0.031, and p = 0.001, respectively). The clinical-radiomics model exhibited a significantly higher across both domestic and international validation sets compared to the SR radiomics model alone (0.95 vs 0.82, p = 0.004; 0.97 vs 0.91, p = 0.031). Moreover, the clinical-radiomics model surpassed the performance of experienced radiologists in both domestic (AUC, 0.95 vs 0.85, p = 0.012) and international (AUC, 0.97 vs 0.77, p < 0.001) validation cohorts. The SR-based clinical-radiomics model can effectively differentiate between SE and n-SE. This international multicenter study demonstrated that a radiomics model of deep learning-based SR reconstructed US images enabled effective differentiation between SE and n-SE. Clinical parameters and radiologists' assessments exhibit limited diagnostic accuracy for SE/n-SE differentiation in TGCT. Based on scrotal US images of TGCT, the SR radiomics models performed better than the NR radiomics models. The SR-based clinical-radiomics model outperforms both the radiomics model and radiologists' assessment, enabling accurate, non-invasive preoperative differentiation between SE and n-SE.

Breast MRI provides high-resolution volumetric imaging critical for tumor assessment and treatment planning, yet manual interpretation of 3D scans remains labor-intensive and subjective. While AI-powered tools hold promise for accelerating medical image analysis, adoption of commercial medical AI products remains limited in low- and middle-income countries due to high license costs, proprietary software, and infrastructure demands. In this work, we investigate whether the Segment Anything Model 2 (SAM2) can be adapted for low-cost, minimal-input 3D tumor segmentation in breast MRI. Using a single bounding box annotation on one slice, we propagate segmentation predictions across the 3D volume using three different slice-wise tracking strategies: top-to-bottom, bottom-to-top, and center-outward. We evaluate these strategies across a large cohort of patients and find that center-outward propagation yields the most consistent and accurate segmentations. Despite being a zero-shot model not trained for volumetric medical data, SAM2 achieves strong segmentation performance under minimal supervision. We further analyze how segmentation performance relates to tumor size, location, and shape, identifying key failure modes. Our results suggest that general-purpose foundation models such as SAM2 can support 3D medical image analysis with minimal supervision, offering an accessible and affordable alternative for resource-constrained settings.

Teacher-Student (TS) networks have become the mainstream frameworks of semi-supervised deep learning, and are widely used in medical image segmentation. However, traditional TSs based on single or homogeneous encoders often struggle to capture the rich semantic details required for complex, fine-grained tasks. To address this, we propose a novel semi-supervised medical image segmentation framework (IHE-Net), which makes good use of the feature discrepancies of two heterogeneous encoders to improve segmentation performance. The two encoders are instantiated by different learning paradigm networks, namely CNN and Transformer/Mamba, respectively, to extract richer and more robust context representations from unlabeled data. On this basis, we propose a simple yet powerful multi-level feature discrepancy fusion module (MFDF), which effectively integrates different modal features and their discrepancies from two heterogeneous encoders. This design enhances the representational capacity of the model through efficient fusion without introducing additional computational overhead. Furthermore, we introduce a triple consistency learning strategy to improve predictive stability by setting dual decoders and adding mixed output consistency. Extensive experimental results on three skin lesion segmentation datasets, ISIC2017, ISIC2018, and PH2, demonstrate the superiority of our framework. Ablation studies further validate the rationale and effectiveness of the proposed method. Code is available at: https://github.com/joey-AI-medical-learning/IHE-Net.

Chest X-ray (CXR) imaging is one of the most widely used diagnostic modalities in clinical practice, encompassing a broad spectrum of diagnostic tasks. Recent advancements have seen the extensive application of reasoning-based multimodal large language models (MLLMs) in medical imaging to enhance diagnostic efficiency and interpretability. However, existing multimodal models predominantly rely on "one-time" diagnostic approaches, lacking verifiable supervision of the reasoning process. This leads to challenges in multi-task CXR diagnosis, including lengthy reasoning, sparse rewards, and frequent hallucinations. To address these issues, we propose CX-Mind, the first generative model to achieve interleaved "think-answer" reasoning for CXR tasks, driven by curriculum-based reinforcement learning and verifiable process rewards (CuRL-VPR). Specifically, we constructed an instruction-tuning dataset, CX-Set, comprising 708,473 images and 2,619,148 samples, and generated 42,828 high-quality interleaved reasoning data points supervised by clinical reports. Optimization was conducted in two stages under the Group Relative Policy Optimization framework: initially stabilizing basic reasoning with closed-domain tasks, followed by transfer to open-domain diagnostics, incorporating rule-based conditional process rewards to bypass the need for pretrained reward models. Extensive experimental results demonstrate that CX-Mind significantly outperforms existing medical and general-domain MLLMs in visual understanding, text generation, and spatiotemporal alignment, achieving an average performance improvement of 25.1% over comparable CXR-specific models. On real-world clinical dataset (Rui-CXR), CX-Mind achieves a mean recall@1 across 14 diseases that substantially surpasses the second-best results, with multi-center expert evaluations further confirming its clinical utility across multiple dimensions.

Early and accurate detection of the bone fracture is paramount to initiating treatment as early as possible and avoiding any delay in patient treatment and outcomes. Interpretation of X-ray image is a time consuming and error prone task, especially when resources for such interpretation are limited by lack of radiology expertise. Additionally, deep learning approaches used currently, typically suffer from misclassifications and lack interpretable explanations to clinical use. In order to overcome these challenges, we propose an automated framework of bone fracture detection using a VGG-19 model modified to our needs. It incorporates sophisticated preprocessing techniques that include Contrast Limited Adaptive Histogram Equalization (CLAHE), Otsu's thresholding, and Canny edge detection, among others, to enhance image clarity as well as to facilitate the feature extraction. Therefore, we use Grad-CAM, an Explainable AI method that can generate visual heatmaps of the model's decision making process, as a type of model interpretability, for clinicians to understand the model's decision making process. It encourages trust and helps in further clinical validation. It is deployed in a real time web application, where healthcare professionals can upload X-ray images and get the diagnostic feedback within 0.5 seconds. The performance of our modified VGG-19 model attains 99.78\% classification accuracy and AUC score of 1.00, making it exceptionally good. The framework provides a reliable, fast, and interpretable solution for bone fracture detection that reasons more efficiently for diagnoses and better patient care.