Automated 3D CT diagnosis empowers clinicians to make timely, evidence-based decisions by enhancing diagnostic accuracy and workflow efficiency. While multimodal large language models (MLLMs) exhibit promising performance in visual-language understanding, existing methods mainly focus on 2D medical images, which fundamentally limits their ability to capture complex 3D anatomical structures. This limitation often leads to misinterpretation of subtle pathologies and causes diagnostic hallucinations. In this paper, we present Hybrid Spatial Encoding Network (HSENet), a framework that exploits enriched 3D medical visual cues by effective visual perception and projection for accurate and robust vision-language understanding. Specifically, HSENet employs dual-3D vision encoders to perceive both global volumetric contexts and fine-grained anatomical details, which are pre-trained by dual-stage alignment with diagnostic reports. Furthermore, we propose Spatial Packer, an efficient multimodal projector that condenses high-resolution 3D spatial regions into a compact set of informative visual tokens via centroid-based compression. By assigning spatial packers with dual-3D vision encoders, HSENet can seamlessly perceive and transfer hybrid visual representations to LLM's semantic space, facilitating accurate diagnostic text generation. Experimental results demonstrate that our method achieves state-of-the-art performance in 3D language-visual retrieval (39.85% of R@100, +5.96% gain), 3D medical report generation (24.01% of BLEU-4, +8.01% gain), and 3D visual question answering (73.60% of Major Class Accuracy, +1.99% gain), confirming its effectiveness. Our code is available at https://github.com/YanzhaoShi/HSENet.

Different medical imaging modalities capture diagnostic information at varying spatial resolutions, from coarse global patterns to fine-grained localized structures. However, most existing vision-language frameworks in the medical domain apply a uniform strategy for local feature extraction, overlooking the modality-specific demands. In this work, we present MedMoE, a modular and extensible vision-language processing framework that dynamically adapts visual representation based on the diagnostic context. MedMoE incorporates a Mixture-of-Experts (MoE) module conditioned on the report type, which routes multi-scale image features through specialized expert branches trained to capture modality-specific visual semantics. These experts operate over feature pyramids derived from a Swin Transformer backbone, enabling spatially adaptive attention to clinically relevant regions. This framework produces localized visual representations aligned with textual descriptions, without requiring modality-specific supervision at inference. Empirical results on diverse medical benchmarks demonstrate that MedMoE improves alignment and retrieval performance across imaging modalities, underscoring the value of modality-specialized visual representations in clinical vision-language systems.

The growing integration of vision-language models (VLMs) in medical applications offers promising support for diagnostic reasoning. However, current medical VLMs often face limitations in generalization, transparency, and computational efficiency-barriers that hinder deployment in real-world, resource-constrained settings. To address these challenges, we propose a Reasoning-Aware Reinforcement Learning framework, \textbf{RARL}, that enhances the reasoning capabilities of medical VLMs while remaining efficient and adaptable to low-resource environments. Our approach fine-tunes a lightweight base model, Qwen2-VL-2B-Instruct, using Low-Rank Adaptation and custom reward functions that jointly consider diagnostic accuracy and reasoning quality. Training is performed on a single NVIDIA A100-PCIE-40GB GPU, demonstrating the feasibility of deploying such models in constrained environments. We evaluate the model using an LLM-as-judge framework that scores both correctness and explanation quality. Experimental results show that RARL significantly improves VLM performance in medical image analysis and clinical reasoning, outperforming supervised fine-tuning on reasoning-focused tasks by approximately 7.78%, while requiring fewer computational resources. Additionally, we demonstrate the generalization capabilities of our approach on unseen datasets, achieving around 27% improved performance compared to supervised fine-tuning and about 4% over traditional RL fine-tuning. Our experiments also illustrate that diversity prompting during training and reasoning prompting during inference are crucial for enhancing VLM performance. Our findings highlight the potential of reasoning-guided learning and reasoning prompting to steer medical VLMs toward more transparent, accurate, and resource-efficient clinical decision-making. Code and data are publicly available.

We present ReXVQA, the largest and most comprehensive benchmark for visual question answering (VQA) in chest radiology, comprising approximately 696,000 questions paired with 160,000 chest X-rays studies across training, validation, and test sets. Unlike prior efforts that rely heavily on template based queries, ReXVQA introduces a diverse and clinically authentic task suite reflecting five core radiological reasoning skills: presence assessment, location analysis, negation detection, differential diagnosis, and geometric reasoning. We evaluate eight state-of-the-art multimodal large language models, including MedGemma-4B-it, Qwen2.5-VL, Janus-Pro-7B, and Eagle2-9B. The best-performing model (MedGemma) achieves 83.24% overall accuracy. To bridge the gap between AI performance and clinical expertise, we conducted a comprehensive human reader study involving 3 radiology residents on 200 randomly sampled cases. Our evaluation demonstrates that MedGemma achieved superior performance (83.84% accuracy) compared to human readers (best radiology resident: 77.27%), representing a significant milestone where AI performance exceeds expert human evaluation on chest X-ray interpretation. The reader study reveals distinct performance patterns between AI models and human experts, with strong inter-reader agreement among radiologists while showing more variable agreement patterns between human readers and AI models. ReXVQA establishes a new standard for evaluating generalist radiological AI systems, offering public leaderboards, fine-grained evaluation splits, structured explanations, and category-level breakdowns. This benchmark lays the foundation for next-generation AI systems capable of mimicking expert-level clinical reasoning beyond narrow pathology classification. Our dataset will be open-sourced at https://huggingface.co/datasets/rajpurkarlab/ReXVQA

BackgroundDifferent Natural Language Processing (NLP) techniques have demonstrated promising results for data extraction from radiological reports. Both traditional rule-based methods like regular expressions (Regex) and modern Large Language Models (LLMs) can extract structured information. However, comparison between these approaches for extraction of specific radiological data elements has not been widely conducted. MethodsWe compared accuracy and processing time between Regex and LLM-based approaches for extracting BI-RADS scores from 7,764 radiology reports (mammography, ultrasound, MRI, and biopsy). We developed a rule-based algorithm using Regex patterns and implemented an LLM-based extraction using the Rombos-LLM-V2.6-Qwen-14b model. A ground truth dataset of 199 manually classified reports was used for evaluation. ResultsThere was no statistically significant difference in the accuracy in extracting BI-RADS scores between Regex and an LLM-based method (accuracy of 89.20% for Regex versus 87.69% for the LLM-based method; p=0.56). Compared to the LLM-based method, Regex processing was more efficient, completing the task 28,120 times faster (0.06 seconds vs. 1687.20 seconds). Further analysis revealed LLMs favored common classifications (particularly BI-RADS value of 2) while Regex more frequently returned "unclear" values. We also could confirm in our sample an already known laterality bias for breast cancer (BI-RADS 6) and detected a slight laterality skew for suspected breast cancer (BI-RADS 5) as well. ConclusionFor structured, standardized data like BI-RADS, traditional NLP techniques seem to be superior, though future work should explore hybrid approaches combining Regex precision for standardized elements with LLM contextual understanding for more complex information extraction tasks.

The escalating demand for medical image interpretation underscores the critical need for advanced artificial intelligence solutions to enhance the efficiency and accuracy of radiological diagnoses. This paper introduces CXR-PathFinder, a novel Large Language Model (LLM)-centric foundation model specifically engineered for automated chest X-ray (CXR) report generation. We propose a unique training paradigm, Clinician-Guided Adversarial Fine-Tuning (CGAFT), which meticulously integrates expert clinical feedback into an adversarial learning framework to mitigate factual inconsistencies and improve diagnostic precision. Complementing this, our Knowledge Graph Augmentation Module (KGAM) acts as an inference-time safeguard, dynamically verifying generated medical statements against authoritative knowledge bases to minimize hallucinations and ensure standardized terminology. Leveraging a comprehensive dataset of millions of paired CXR images and expert reports, our experiments demonstrate that CXR-PathFinder significantly outperforms existing state-of-the-art medical vision-language models across various quantitative metrics, including clinical accuracy (Macro F1 (14): 46.5, Micro F1 (14): 59.5). Furthermore, blinded human evaluation by board-certified radiologists confirms CXR-PathFinder's superior clinical utility, completeness, and accuracy, establishing its potential as a reliable and efficient aid for radiological practice. The developed method effectively balances high diagnostic fidelity with computational efficiency, providing a robust solution for automated medical report generation.

Artificial intelligence (AI)-based chest X-ray (CXR) interpretation assistants have demonstrated significant progress and are increasingly being applied in clinical settings. However, contemporary medical AI models often adhere to a simplistic input-to-output paradigm, directly processing an image and an instruction to generate a result, where the instructions may be integral to the model's architecture. This approach overlooks the modeling of the inherent diagnostic reasoning in chest X-ray interpretation. Such reasoning is typically sequential, where each interpretive stage considers the images, the current task, and the contextual information from previous stages. This oversight leads to several shortcomings, including misalignment with clinical scenarios, contextless reasoning, and untraceable errors. To fill this gap, we construct CXRTrek, a new multi-stage visual question answering (VQA) dataset for CXR interpretation. The dataset is designed to explicitly simulate the diagnostic reasoning process employed by radiologists in real-world clinical settings for the first time. CXRTrek covers 8 sequential diagnostic stages, comprising 428,966 samples and over 11 million question-answer (Q&A) pairs, with an average of 26.29 Q&A pairs per sample. Building on the CXRTrek dataset, we propose a new vision-language large model (VLLM), CXRTrekNet, specifically designed to incorporate the clinical reasoning flow into the VLLM framework. CXRTrekNet effectively models the dependencies between diagnostic stages and captures reasoning patterns within the radiological context. Trained on our dataset, the model consistently outperforms existing medical VLLMs on the CXRTrek benchmarks and demonstrates superior generalization across multiple tasks on five diverse external datasets. The dataset and model can be found in our repository (https://github.com/guanjinquan/CXRTrek).

The differential diagnosis of neurodegenerative dementias is a challenging clinical task, mainly because of the overlap in symptom presentation and the similarity of patterns observed in structural neuroimaging. To improve diagnostic efficiency and accuracy, deep learning-based methods such as Convolutional Neural Networks and Vision Transformers have been proposed for the automatic classification of brain MRIs. However, despite their strong predictive performance, these models find limited clinical utility due to their opaque decision making. In this work, we propose a framework that integrates two core components to enhance diagnostic transparency. First, we introduce a modular pipeline for converting 3D T1-weighted brain MRIs into textual radiology reports. Second, we explore the potential of modern Large Language Models (LLMs) to assist clinicians in the differential diagnosis between Frontotemporal dementia subtypes, Alzheimer's disease, and normal aging based on the generated reports. To bridge the gap between predictive accuracy and explainability, we employ reinforcement learning to incentivize diagnostic reasoning in LLMs. Without requiring supervised reasoning traces or distillation from larger models, our approach enables the emergence of structured diagnostic rationales grounded in neuroimaging findings. Unlike post-hoc explainability methods that retrospectively justify model decisions, our framework generates diagnostic rationales as part of the inference process-producing causally grounded explanations that inform and guide the model's decision-making process. In doing so, our framework matches the diagnostic performance of existing deep learning methods while offering rationales that support its diagnostic conclusions.

Computed Tomography (CT) scan, which produces 3D volumetric medical data that can be viewed as hundreds of cross-sectional images (a.k.a. slices), provides detailed anatomical information for diagnosis. For radiologists, creating CT radiology reports is time-consuming and error-prone. A visual question answering (VQA) system that can answer radiologists' questions about some anatomical regions on the CT scan and even automatically generate a radiology report is urgently needed. However, existing VQA systems cannot adequately handle the CT radiology question answering (CTQA) task for: (1) anatomic complexity makes CT images difficult to understand; (2) spatial relationship across hundreds slices is difficult to capture. To address these issues, this paper proposes CT-Agent, a multimodal agentic framework for CTQA. CT-Agent adopts anatomically independent tools to break down the anatomic complexity; furthermore, it efficiently captures the across-slice spatial relationship with a global-local token compression strategy. Experimental results on two 3D chest CT datasets, CT-RATE and RadGenome-ChestCT, verify the superior performance of CT-Agent.

Large language models (LLMs) are increasingly used by physicians for diagnostic support. A key advantage of LLMs is the ability to generate explanations that can help physicians understand the reasoning behind a diagnosis. However, the best-suited format for LLM-generated explanations remains unclear. In this large-scale study, we examined the effect of different formats for LLM explanations on clinical decision-making. For this, we conducted a randomized experiment with radiologists reviewing patient cases with radiological images (N = 2020 assessments). Participants received either no LLM support (control group) or were supported by one of three LLM-generated explanations: (1) a standard output providing the diagnosis without explanation; (2) a differential diagnosis comparing multiple possible diagnoses; or (3) a chain-of-thought explanation offering a detailed reasoning process for the diagnosis. We find that the format of explanations significantly influences diagnostic accuracy. The chain-of-thought explanations yielded the best performance, improving the diagnostic accuracy by 12.2% compared to the control condition without LLM support (P = 0.001). The chain-of-thought explanations are also superior to the standard output without explanation (+7.2%; P = 0.040) and the differential diagnosis format (+9.7%; P = 0.004). We further assessed the robustness of these findings across case difficulty and different physician backgrounds such as general vs. specialized radiologists. Evidently, explaining the reasoning for a diagnosis helps physicians to identify and correct potential errors in LLM predictions and thus improve overall decisions. Altogether, the results highlight the importance of how explanations in medical LLMs are generated to maximize their utility in clinical practice. By designing explanations to support the reasoning processes of physicians, LLMs can improve diagnostic performance and, ultimately, patient outcomes.