Xiaoman Zhang

Chaoyi Wu, Weixiong Lin, Xiaoman Zhang et al. · harvard

Recently, Large Language Models (LLMs) have showcased remarkable capabilities in natural language understanding. While demonstrating proficiency in everyday conversations and question-answering situations, these models frequently struggle in domains that require precision, such as medical applications, due to their lack of domain-specific knowledge. In this paper, we describe the procedure for building a powerful, open-source language model specifically designed for medicine applications, termed as PMC-LLaMA. Our contributions are threefold: (i) we systematically investigate the process of adapting a general-purpose foundation language model towards medical domain, this involves data-centric knowledge injection through the integration of 4.8M biomedical academic papers and 30K medical textbooks, as well as comprehensive fine-tuning for alignment with domain-specific instructions; (ii) we contribute a large-scale, comprehensive dataset for instruction tuning. This dataset encompasses medical question-answering (QA), rationale for reasoning, and conversational dialogues, comprising a total of 202M tokens; (iii) we conduct thorough ablation studies to demonstrate the effectiveness of each proposed component. While evaluating on various public medical question-answering benchmarks, our lightweight PMCLLaMA, which consists of only 13 billion parameters, exhibits superior performance, even surpassing ChatGPT. All models, codes, datasets can be found in https://github.com/chaoyi-wu/PMC-LLaMA.

Chaoyi Wu, Jiayu Lei, Qiaoyu Zheng et al. · harvard

Driven by the large foundation models, the development of artificial intelligence has witnessed tremendous progress lately, leading to a surge of general interest from the public. In this study, we aim to assess the performance of OpenAI's newest model, GPT-4V(ision), specifically in the realm of multimodal medical diagnosis. Our evaluation encompasses 17 human body systems, including Central Nervous System, Head and Neck, Cardiac, Chest, Hematology, Hepatobiliary, Gastrointestinal, Urogenital, Gynecology, Obstetrics, Breast, Musculoskeletal, Spine, Vascular, Oncology, Trauma, Pediatrics, with images taken from 8 modalities used in daily clinic routine, e.g., X-ray, Computed Tomography (CT), Magnetic Resonance Imaging (MRI), Positron Emission Tomography (PET), Digital Subtraction Angiography (DSA), Mammography, Ultrasound, and Pathology. We probe the GPT-4V's ability on multiple clinical tasks with or without patent history provided, including imaging modality and anatomy recognition, disease diagnosis, report generation, disease localisation. Our observation shows that, while GPT-4V demonstrates proficiency in distinguishing between medical image modalities and anatomy, it faces significant challenges in disease diagnosis and generating comprehensive reports. These findings underscore that while large multimodal models have made significant advancements in computer vision and natural language processing, it remains far from being used to effectively support real-world medical applications and clinical decision-making. All images used in this report can be found in https://github.com/chaoyi-wu/GPT-4V_Medical_Evaluation.

Weixiong Lin, Ziheng Zhao, Xiaoman Zhang et al. · harvard

Foundation models trained on large-scale dataset gain a recent surge in CV and NLP. In contrast, development in biomedical domain lags far behind due to data scarcity. To address this issue, we build and release PMC-OA, a biomedical dataset with 1.6M image-caption pairs collected from PubMedCentral's OpenAccess subset, which is 8 times larger than before. PMC-OA covers diverse modalities or diseases, with majority of the image-caption samples aligned at finer-grained level, i.e., subfigure and subcaption. While pretraining a CLIP-style model on PMC-OA, our model named PMC-CLIP achieves state-of-the-art results on various downstream tasks, including image-text retrieval on ROCO, MedMNIST image classification, Medical VQA, i.e. +8.1% R@10 on image-text retrieval, +3.9% accuracy on image classification.

Chaoyi Wu, Xiaoman Zhang, Ya Zhang et al. · harvard

In this study, we aim to initiate the development of Radiology Foundation Model, termed as RadFM. We consider the construction of foundational models from three perspectives, namely, dataset construction, model design, and thorough evaluation. Our contribution can be concluded as follows: (i), we construct a large-scale Medical Multi-modal Dataset, MedMD, which consists of 16M 2D and 3D medical scans with high-quality text descriptions or reports across various data formats, modalities, and tasks, covering over 5000 distinct diseases. To the best of our knowledge, this is the first large-scale, high-quality, medical visual-language dataset, with both 2D and 3D scans; (ii), we propose an architecture that enables visually conditioned generative pre-training, i.e., allowing for integration of text input with 2D or 3D medical scans, and generate responses for diverse radiologic tasks. The model was initially pre-trained on MedMD and subsequently fine-tuned on the domain-specific dataset, which is a radiologic cleaned version of MedMD, containing 3M radiologic visual-language pairs, termed as RadMD; (iii), we propose a new evaluation benchmark, RadBench, that comprises five tasks, including modality recognition, disease diagnosis, visual question answering, report generation and rationale diagnosis, aiming to comprehensively assess the capability of foundation models in handling practical clinical problems. We conduct both automatic and human evaluation on RadBench, in both cases, RadFM outperforms existing multi-modal foundation models, that are publicaly accessible, including Openflamingo, MedFlamingo, MedVInT and GPT-4V. Additionally, we also adapt RadFM for different public benchmarks, surpassing existing SOTAs on diverse datasets. All codes, data, and model checkpoint will all be made publicly available to promote further research and development in the field.

Xiaoman Zhang, Chaoyi Wu, Ya Zhang et al. · harvard

While multi-modal foundation models pre-trained on large-scale data have been successful in natural language understanding and vision recognition, their use in medical domains is still limited due to the fine-grained nature of medical tasks and the high demand for domain knowledge. To address this challenge, we propose a novel approach called Knowledge-enhanced Auto Diagnosis (KAD) which leverages existing medical domain knowledge to guide vision-language pre-training using paired chest X-rays and radiology reports. We evaluate KAD on {four} external X-ray datasets and demonstrate that its zero-shot performance is not only comparable to that of fully-supervised models, but also superior to the average of three expert radiologists for three (out of five) pathologies with statistical significance. Moreover, when few-shot annotation is available, KAD outperforms all existing approaches in fine-tuning settings, demonstrating its potential for application in different clinical scenarios.

Chaoyi Wu, Xiaoman Zhang, Ya Zhang et al. · harvard

In this paper, we consider enhancing medical visual-language pre-training (VLP) with domain-specific knowledge, by exploiting the paired image-text reports from the radiological daily practice. In particular, we make the following contributions: First, unlike existing works that directly process the raw reports, we adopt a novel triplet extraction module to extract the medical-related information, avoiding unnecessary complexity from language grammar and enhancing the supervision signals; Second, we propose a novel triplet encoding module with entity translation by querying a knowledge base, to exploit the rich domain knowledge in medical field, and implicitly build relationships between medical entities in the language embedding space; Third, we propose to use a Transformer-based fusion model for spatially aligning the entity description with visual signals at the image patch level, enabling the ability for medical diagnosis; Fourth, we conduct thorough experiments to validate the effectiveness of our architecture, and benchmark on numerous public benchmarks, e.g., ChestX-ray14, RSNA Pneumonia, SIIM-ACR Pneumothorax, COVIDx CXR-2, COVID Rural, and EdemaSeverity. In both zero-shot and fine-tuning settings, our model has demonstrated strong performance compared with the former methods on disease classification and grounding.

Tengfei Zhang, Ziheng Zhao, Lisong Dai et al.

Medical imaging artificial intelligence has achieved strong performance in isolated image interpretation, but remains poorly aligned with radiological practice, where diagnosis and follow-up rely on comparison across prior studies and analogous reference cases. Here we formulate radiological comparison as an entity-aware cross-image reasoning problem and introduce a framework that supports both reference-case retrieval and temporal comparative interpretation. We construct MedReCo-DB, a large-scale comparative imaging resource derived from routine image-report pairs, comprising more than 690,000 images from over 160,000 patients across eight institutions, four countries and seven imaging modalities. Reports are decomposed into anatomical structures, abnormal findings and pathological conditions to provide supervision for entity-conditioned retrieval and comparative visual question answering. Using this resource, we develop MedReCo, an entity-aware visual encoder for controllable retrieval of clinically analogous cases, and MedReCo-VLM, a vision--language extension for generative interpretation of interval change. Across internal, external and cross-center evaluations, MedReCo achieved the highest Recall@1 in all 12 internal retrieval settings and improved external retrieval by a mean of 6.0 percentage points. In clinically confusable differential groups, it consistently outperformed the strongest baselines. MedReCo-VLM achieved the best performance across all comparative generation evaluations and improved longitudinal follow-up accuracy by 14.5-46.5 percentage points on chest radiographs and 13.0-27.9 percentage points on CT. These findings suggest that entity-aware comparative reasoning can be learned from routine clinical data at scale and may provide a more clinically aligned foundation for medical imaging AI.

Xucheng Wang, Zhizhou Yang, Xiaoman Zhang et al.

This study presents the first evaluation of general-purpose imitation learning for surgeon-robot collaborative assistance in open surgery, targeting suture following: the grab-pull-release motion an assistant performs at every stitch. We collect 160 teleoperated demonstrations (32,374 frames) on an open-source robot arm, benchmark four architecturally diverse imitation learning policies (ACT, Diffusion Policy, SmolVLA, $π_0$) across 28 trained models evaluated in 32 configurations along three clinically motivated dimensions: dataset size, camera viewpoint, and background variation. Our results demonstrate that under ideal conditions, the four policies achieve $50$-$75\%$ task success, with depth error as the dominant failure mode across all architectures. Among all policies, $π_0$ achieves the strongest results with a pretrained vision-language backbone, demonstrating superior data efficiency, greater robustness to background variation, and smoother trajectories compatible with surgical workflow. When deployed in a surgeon-robot suturing trial, $π_0$ yields a $92\%$ stitch completion rate. These findings establish collaborative robotic assistance in open surgery as a feasible target for imitation learning and highlight depth perception and end-effector design as key priorities for clinical translation.

Chaoyi Wu, Xiaoman Zhang, Yanfeng Wang et al. · harvard

In this paper, we consider the problem of disease diagnosis. Unlike the conventional learning paradigm that treats labels independently, we propose a knowledge-enhanced framework, that enables training visual representation with the guidance of medical domain knowledge. In particular, we make the following contributions: First, to explicitly incorporate experts' knowledge, we propose to learn a neural representation for the medical knowledge graph via contrastive learning, implicitly establishing relations between different medical concepts. Second, while training the visual encoder, we keep the parameters of the knowledge encoder frozen and propose to learn a set of prompt vectors for efficient adaptation. Third, we adopt a Transformer-based disease-query module for cross-model fusion, which naturally enables explainable diagnosis results via cross attention. To validate the effectiveness of our proposed framework, we conduct thorough experiments on three x-ray imaging datasets across different anatomy structures, showing our model is able to exploit the implicit relations between diseases/findings, thus is beneficial to the commonly encountered problem in the medical domain, namely, long-tailed and zero-shot recognition, which conventional methods either struggle or completely fail to realize.

Jiayu Lei, Lisong Dai, Haoyun Jiang et al. · harvard

Magnetic resonance imaging~(MRI) have played a crucial role in brain disease diagnosis, with which a range of computer-aided artificial intelligence methods have been proposed. However, the early explorations usually focus on the limited types of brain diseases in one study and train the model on the data in a small scale, yielding the bottleneck of generalization. Towards a more effective and scalable paradigm, we propose a hierarchical knowledge-enhanced pre-training framework for the universal brain MRI diagnosis, termed as UniBrain. Specifically, UniBrain leverages a large-scale dataset of 24,770 imaging-report pairs from routine diagnostics. Different from previous pre-training techniques for the unitary vision or textual feature, or with the brute-force alignment between vision and language information, we leverage the unique characteristic of report information in different granularity to build a hierarchical alignment mechanism, which strengthens the efficiency in feature learning. Our UniBrain is validated on three real world datasets with severe class imbalance and the public BraTS2019 dataset. It not only consistently outperforms all state-of-the-art diagnostic methods by a large margin and provides a superior grounding performance but also shows comparable performance compared to expert radiologists on certain disease types.

Jiayu Lei, Xiaoman Zhang, Chaoyi Wu et al. · harvard

Radiologists are tasked with interpreting a large number of images in a daily base, with the responsibility of generating corresponding reports. This demanding workload elevates the risk of human error, potentially leading to treatment delays, increased healthcare costs, revenue loss, and operational inefficiencies. To address these challenges, we initiate a series of work on grounded Automatic Report Generation (AutoRG), starting from the brain MRI interpretation system, which supports the delineation of brain structures, the localization of anomalies, and the generation of well-organized findings. We make contributions from the following aspects, first, on dataset construction, we release a comprehensive dataset encompassing segmentation masks of anomaly regions and manually authored reports, termed as RadGenome-Brain MRI. This data resource is intended to catalyze ongoing research and development in the field of AI-assisted report generation systems. Second, on system design, we propose AutoRG-Brain, the first brain MRI report generation system with pixel-level grounded visual clues. Third, for evaluation, we conduct quantitative assessments and human evaluations of brain structure segmentation, anomaly localization, and report generation tasks to provide evidence of its reliability and accuracy. This system has been integrated into real clinical scenarios, where radiologists were instructed to write reports based on our generated findings and anomaly segmentation masks. The results demonstrate that our system enhances the report-writing skills of junior doctors, aligning their performance more closely with senior doctors, thereby boosting overall productivity.

Xiaoman Zhang, Julián N. Acosta, Hong-Yu Zhou et al.

Recent advancements in artificial intelligence have significantly improved the automatic generation of radiology reports. However, existing evaluation methods fail to reveal the models' understanding of radiological images and their capacity to achieve human-level granularity in descriptions. To bridge this gap, we introduce a system, named ReXKG, which extracts structured information from processed reports to construct a comprehensive radiology knowledge graph. We then propose three metrics to evaluate the similarity of nodes (ReXKG-NSC), distribution of edges (ReXKG-AMS), and coverage of subgraphs (ReXKG-SCS) across various knowledge graphs. We conduct an in-depth comparative analysis of AI-generated and human-written radiology reports, assessing the performance of both specialist and generalist models. Our study provides a deeper understanding of the capabilities and limitations of current AI models in radiology report generation, offering valuable insights for improving model performance and clinical applicability.

Grace Chang Yuan, Xiaoman Zhang, Sung Eun Kim et al.

Multi-agent large language model (LLM) systems have emerged as a promising approach for clinical diagnosis, leveraging collaboration among agents to refine medical reasoning. However, most existing frameworks rely on single-vendor teams (e.g., multiple agents from the same model family), which risk correlated failure modes that reinforce shared biases rather than correcting them. We investigate the impact of vendor diversity by comparing Single-LLM, Single-Vendor, and Mixed-Vendor Multi-Agent Conversation (MAC) frameworks. Using three doctor agents instantiated with o4-mini, Gemini-2.5-Pro, and Claude-4.5-Sonnet, we evaluate performance on RareBench and DiagnosisArena. Mixed-vendor configurations consistently outperform single-vendor counterparts, achieving state-of-the-art recall and accuracy. Overlap analysis reveals the underlying mechanism: mixed-vendor teams pool complementary inductive biases, surfacing correct diagnoses that individual models or homogeneous teams collectively miss. These results highlight vendor diversity as a key design principle for robust clinical diagnostic systems.

Xucheng Wang, Xiaoman Zhang, Sung Eun Kim et al.

Ultrasound acquisition requires skilled probe manipulation and real-time adjustments. Vision-language models (VLMs) could enable autonomous ultrasound systems, but existing benchmarks evaluate only static images, not dynamic procedural understanding. We introduce ReXSonoVQA, a video QA benchmark with 514 video clips and 514 questions (249 MCQ, 265 free-response) targeting three competencies: Action-Goal Reasoning, Artifact Resolution & Optimization, and Procedure Context & Planning. Zero-shot evaluation of Gemini 3 Pro, Qwen3.5-397B, LLaVA-Video-72B, and Seed 2.0 Pro shows VLMs can extract some procedural information, but troubleshooting questions remain challenging with minimal gains over text-only baselines, exposing limitations in causal reasoning. ReXSonoVQA enables developing perception systems for ultrasound training, guidance, and robotic automation.

Julián N. Acosta, Xiaoman Zhang, Siddhant Dogra et al.

We present Head CT Ontology Normalized Evaluation (HeadCT-ONE), a metric for evaluating head CT report generation through ontology-normalized entity and relation extraction. HeadCT-ONE enhances current information extraction derived metrics (such as RadGraph F1) by implementing entity normalization through domain-specific ontologies, addressing radiological language variability. HeadCT-ONE compares normalized entities and relations, allowing for controllable weighting of different entity types or specific entities. Through experiments on head CT reports from three health systems, we show that HeadCT-ONE's normalization and weighting approach improves the capture of semantically equivalent reports, better distinguishes between normal and abnormal reports, and aligns with radiologists' assessment of clinically significant errors, while offering flexibility to prioritize specific aspects of report content. Our results demonstrate how HeadCT-ONE enables more flexible, controllable, and granular automated evaluation of head CT reports.

Oishi Banerjee, Sung Eun Kim, Alexandra N. Willauer et al.

Everyday photographs taken with ordinary cameras are already widely used in telemedicine and other online health conversations, yet no comprehensive benchmark evaluates whether vision-language models can interpret their medical content. Analyzing these images requires both fine-grained natural image understanding and domain-specific medical reasoning, a combination that challenges both general-purpose and specialized models. We introduce ReXInTheWild, a benchmark of 955 clinician-verified multiple-choice questions spanning seven clinical topics across 484 photographs sourced from the biomedical literature. When evaluated on ReXInTheWild, leading multimodal large language models show substantial performance variation: Gemini-3 achieves 78% accuracy, followed by Claude Opus 4.5 (72%) and GPT-5 (68%), while the medical specialist model MedGemma reaches only 37%. A systematic error analysis also reveals four categories of common errors, ranging from low-level geometric errors to high-level reasoning failures and requiring different mitigation strategies. ReXInTheWild provides a challenging, clinically grounded benchmark at the intersection of natural image understanding and medical reasoning. The dataset is available on HuggingFace.

Pengcheng Qiu, Chaoyi Wu, Xiaoman Zhang et al. · harvard

The development of open-source, multilingual medical language models can benefit a wide, linguistically diverse audience from different regions. To promote this domain, we present contributions from the following: First, we construct a multilingual medical corpus, containing approximately 25.5B tokens encompassing 6 main languages, termed as MMedC, enabling auto-regressive domain adaptation for general LLMs; Second, to monitor the development of multilingual medical LLMs, we propose a multilingual medical multi-choice question-answering benchmark with rationale, termed as MMedBench; Third, we have assessed a number of open-source large language models (LLMs) on our benchmark, along with those further auto-regressive trained on MMedC. Our final model, MMed-Llama 3, with only 8B parameters, achieves superior performance compared to all other open-source models on both MMedBench and English benchmarks, even rivaling GPT-4. In conclusion, in this work, we present a large-scale corpus, a benchmark and a series of models to support the development of multilingual medical LLMs.

Roshan Kenia, Xiaoman Zhang, Pranav Rajpurkar

Autonomous coding agents built on large language models (LLMs) can now solve many general software and machine learning tasks, but they remain ineffective on complex, domain-specific scientific problems. Medical imaging is a particularly demanding domain, requiring long training cycles, high-dimensional data handling, and specialized preprocessing and validation pipelines, capabilities not fully measured in existing agent benchmarks. To address this gap, we introduce ReX-MLE, a benchmark of 20 challenges derived from high-impact medical imaging competitions spanning diverse modalities and task types. Unlike prior ML-agent benchmarks, ReX-MLE evaluates full end-to-end workflows, requiring agents to independently manage data preprocessing, model training, and submission under realistic compute and time constraints. Evaluating state-of-the-art agents (AIDE, ML-Master, R&D-Agent) with different LLM backends (GPT-5, Gemini, Claude), we observe a severe performance gap: most submissions rank in the 0th percentile compared to human experts. Failures stem from domain-knowledge and engineering limitations. ReX-MLE exposes these bottlenecks and provides a foundation for developing domain-aware autonomous AI systems.

Xiaoman Zhang, Chaoyi Wu, Ziheng Zhao et al. · harvard

Developing generalist foundation model has recently attracted tremendous attention among researchers in the field of AI for Medicine (AI4Medicine). A pivotal insight in developing these models is their reliance on dataset scaling, which emphasizes the requirements on developing open-source medical image datasets that incorporate diverse supervision signals across various imaging modalities. In this paper, we introduce RadGenome-Chest CT, a comprehensive, large-scale, region-guided 3D chest CT interpretation dataset based on CT-RATE. Specifically, we leverage the latest powerful universal segmentation and large language models, to extend the original datasets (over 25,692 non-contrast 3D chest CT volume and reports from 20,000 patients) from the following aspects: (i) organ-level segmentation masks covering 197 categories, which provide intermediate reasoning visual clues for interpretation; (ii) 665 K multi-granularity grounded reports, where each sentence of the report is linked to the corresponding anatomical region of CT volume in the form of a segmentation mask; (iii) 1.3 M grounded VQA pairs, where questions and answers are all linked with reference segmentation masks, enabling models to associate visual evidence with textual explanations. All grounded reports and VQA pairs in the validation set have gone through manual verification to ensure dataset quality. We believe that RadGenome-Chest CT can significantly advance the development of multimodal medical foundation models, by training to generate texts based on given segmentation regions, which is unattainable with previous relevant datasets. We will release all segmentation masks, grounded reports, and VQA pairs to facilitate further research and development in this field.

Xiaoman Zhang, Julián N. Acosta, Josh Miller et al.

We present ReXGradient-160K, representing the largest publicly available chest X-ray dataset to date in terms of the number of patients. This dataset contains 160,000 chest X-ray studies with paired radiological reports from 109,487 unique patients across 3 U.S. health systems (79 medical sites). This comprehensive dataset includes multiple images per study and detailed radiology reports, making it particularly valuable for the development and evaluation of AI systems for medical imaging and automated report generation models. The dataset is divided into training (140,000 studies), validation (10,000 studies), and public test (10,000 studies) sets, with an additional private test set (10,000 studies) reserved for model evaluation on the ReXrank benchmark. By providing this extensive dataset, we aim to accelerate research in medical imaging AI and advance the state-of-the-art in automated radiological analysis. Our dataset will be open-sourced at https://huggingface.co/datasets/rajpurkarlab/ReXGradient-160K.

Alice Heiman, Xiaoman Zhang, Emma Chen et al.

Medical vision-language models often struggle with generating accurate quantitative measurements in radiology reports, leading to hallucinations that undermine clinical reliability. We introduce FactCheXcker, a modular framework that de-hallucinates radiology report measurements by leveraging an improved query-code-update paradigm. Specifically, FactCheXcker employs specialized modules and the code generation capabilities of large language models to solve measurement queries generated based on the original report. After extracting measurable findings, the results are incorporated into an updated report. We evaluate FactCheXcker on endotracheal tube placement, which accounts for an average of 78% of report measurements, using the MIMIC-CXR dataset and 11 medical report-generation models. Our results show that FactCheXcker significantly reduces hallucinations, improves measurement precision, and maintains the quality of the original reports. Specifically, FactCheXcker improves the performance of 10/11 models and achieves an average improvement of 135.0% in reducing measurement hallucinations measured by mean absolute error. Code is available at https://github.com/rajpurkarlab/FactCheXcker.

Mohammed Baharoon, Luyang Luo, Michael Moritz et al.

We introduce ReXGroundingCT, the first publicly available dataset linking free-text findings to pixel-level 3D segmentations in chest CT scans. The dataset includes 3,142 non-contrast chest CT scans paired with standardized radiology reports from CT-RATE. Construction followed a structured three-stage pipeline. First, GPT-4 was used to extract and standardize findings, descriptors, and metadata from reports originally written in Turkish and machine-translated into English. Second, GPT-4o-mini categorized each finding into a hierarchical ontology of lung and pleural abnormalities. Third, 3D annotations were produced for all CT volumes: the training set was quality-assured by board-certified radiologists, and the validation and test sets were fully annotated by board-certified radiologists. Additionally, a complementary chain-of-thought dataset was created to provide step-by-step hierarchical anatomical reasoning for localizing findings within the CT volume, using GPT-4o and localization coordinates derived from organ segmentation models. ReXGroundingCT contains 16,301 annotated entities across 8,028 text-to-3D-segmentation pairs, covering diverse radiological patterns from 3,142 non-contrast CT scans. About 79% of findings are focal abnormalities and 21% are non-focal. The dataset includes a public validation set of 50 cases and a private test set of 100 cases, both annotated by board-certified radiologists. The dataset establishes a foundation for enabling free-text finding segmentation and grounded radiology report generation in CT imaging. Model performance on the private test set is hosted on a public leaderboard at https://rexrank.ai/ReXGroundingCT. The dataset is available at https://huggingface.co/datasets/rajpurkarlab/ReXGroundingCT.

Ankit Pal, Jung-Oh Lee, Xiaoman Zhang et al.

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

Sraavya Sambara, Sung Eun Kim, Xiaoman Zhang et al.

Current Vision-Language Models (VLMs) struggle to ground anatomical regions in 3D medical images and reason about them in a step-by-step manner, a key requirement of real-world diagnostic assessment. This ability is essential for aligning model outputs with the diagnostic workflows clinicians use in practice, enabling trustworthy clinician-AI collaboration. Existing 3D datasets provide localization labels, but none support this "grounded reasoning" ability. To address this gap, we introduce 3DReasonKnee, the first 3D grounded reasoning dataset for medical images, which provides 494k high-quality quintuples derived from 7,970 3D knee MRI volumes. Each quintuple includes: (1) the 3D MRI volume, (2) a diagnostic question targeting a specific anatomical region (3) a 3D bounding box localizing the relevant anatomical structures, (4) clinician-generated diagnostic reasoning steps that explicitly detail the 3D reasoning process, and (5) structured severity assessments for the relevant anatomical region. The creation and validation of 3DReasonKnee, involving over 450 hours of expert clinician time for manually segmenting MRIs and generating reasoning chains, ensures its superior quality and clinical relevance. We establish ReasonKnee-Bench to evaluate localization and diagnostic accuracy, providing insight into VLM ability to perform grounding and severity assessment across anatomical regions and diagnostic inquiries. We benchmark five state-of-the-art VLMs, providing baseline performance for ReasonKnee-Bench. By providing this unique resource of expert-annotated 3D reasoning pathways, 3DReasonKnee serves as a repository of orthopedic surgeons' diagnostic expertise and offers a vital testbed for advancing multimodal medical AI systems towards 3D, clinically aligned, localized decision-making capabilities. The dataset can be found in: https://huggingface.co/datasets/rajpurkarlab/3DReasonKnee

Ziheng Zhao, Yao Zhang, Chaoyi Wu et al. · harvard

This paper aims to build a model that can Segment Anything in 3D medical images, driven by medical terminologies as Text prompts, termed as SAT. Our main contributions are three-fold: (i) We construct the first multimodal knowledge tree on human anatomy, including 6502 anatomical terminologies; Then, we build the largest and most comprehensive segmentation dataset for training, collecting over 22K 3D scans from 72 datasets, across 497 classes, with careful standardization on both image and label space; (ii) We propose to inject medical knowledge into a text encoder via contrastive learning and formulate a large-vocabulary segmentation model that can be prompted by medical terminologies in text form; (iii) We train SAT-Nano (110M parameters) and SAT-Pro (447M parameters). SAT-Pro achieves comparable performance to 72 nnU-Nets -- the strongest specialist models trained on each dataset (over 2.2B parameters combined) -- over 497 categories. Compared with the interactive approach MedSAM, SAT-Pro consistently outperforms across all 7 human body regions with +7.1% average Dice Similarity Coefficient (DSC) improvement, while showing enhanced scalability and robustness. On 2 external (cross-center) datasets, SAT-Pro achieves higher performance than all baselines (+3.7% average DSC), demonstrating superior generalization ability.

Qiaoyu Zheng, Weike Zhao, Chaoyi Wu et al. · harvard

Developing a generalist radiology diagnosis system can greatly enhance clinical diagnostics. In this paper, we introduce RadDiag, a foundational model supporting 2D and 3D inputs across various modalities and anatomies, using a transformer-based fusion module for comprehensive disease diagnosis. Due to patient privacy concerns and the lack of large-scale radiology diagnosis datasets, we utilize high-quality, clinician-reviewed radiological images available online with diagnosis labels. Our dataset, RP3D-DiagDS, contains 40,936 cases with 195,010 scans covering 5,568 disorders (930 unique ICD-10-CM codes). Experimentally, our RadDiag achieves 95.14% AUC on internal evaluation with the knowledge-enhancement strategy. Additionally, RadDiag can be zero-shot applied or fine-tuned to external diagnosis datasets sourced from various hospitals, demonstrating state-of-the-art results. In conclusion, we show that publicly shared medical data on the Internet is a tremendous and valuable resource that can potentially support building a generalist AI for healthcare.

Xiaoman Zhang, Hong-Yu Zhou, Xiaoli Yang et al.

AI-driven models have demonstrated significant potential in automating radiology report generation for chest X-rays. However, there is no standardized benchmark for objectively evaluating their performance. To address this, we present ReXrank, https://rexrank.ai, a public leaderboard and challenge for assessing AI-powered radiology report generation. Our framework incorporates ReXGradient, the largest test dataset consisting of 10,000 studies, and three public datasets (MIMIC-CXR, IU-Xray, CheXpert Plus) for report generation assessment. ReXrank employs 8 evaluation metrics and separately assesses models capable of generating only findings sections and those providing both findings and impressions sections. By providing this standardized evaluation framework, ReXrank enables meaningful comparisons of model performance and offers crucial insights into their robustness across diverse clinical settings. Beyond its current focus on chest X-rays, ReXrank's framework sets the stage for comprehensive evaluation of automated reporting across the full spectrum of medical imaging.

Wenke Ren, Hengxiao Guo, Wenwen Zuo et al.

Vision-language models (VLMs) are increasingly proposed as general-purpose tools for scientific data interpretation, yet their reliability on real astronomical observations across diverse modalities remains untested. We present AstroVLBench, a comprehensive benchmark comprising over 4,100 expert-verified instances across five tasks spanning optical imaging, radio interferometry, multi-wavelength photometry, time-domain light curves, and optical spectroscopy. Evaluating six frontier models, we find that performance is strongly modality-dependent: while one model (Gemini 3 Pro) emerges as the most consistently capable across tasks, task-specific strengths vary, and all models substantially underperform domain-specialized methods. Mechanistic ablations reveal that performance depends not only on directing attention to salient visual features but also on grounding those features in physical knowledge. Phenomenological prompts describing what to look for improve accuracy by sharpening model focus, but physical prompts explaining why those features matter perform better overall and yield more balanced classifications with reduced class-specific bias. Consistent with this picture, presenting the underlying one-dimensional measurements directly as numerical tables instead of rendered plots yields up to 13 percentage points improvement. Reasoning quality analysis further demonstrates that, without explicit physical grounding, models may reach correct predictions from phenomenologically plausible cues while providing physically imprecise justifications, establishing that accuracy alone is insufficient for trustworthy scientific deployment. These findings provide the first systematic, multi-modal baselines for VLMs in observational astronomy and identify the specific representation, grounding, and reasoning bottlenecks where current models fail.

Xiao Zhou, Xiaoman Zhang, Chaoyi Wu et al. · harvard

In this paper, we consider the problem of visual representation learning for computational pathology, by exploiting large-scale image-text pairs gathered from public resources, along with the domain-specific knowledge in pathology. Specifically, we make the following contributions: (i) We curate a pathology knowledge tree that consists of 50,470 informative attributes for 4,718 diseases requiring pathology diagnosis from 32 human tissues. To our knowledge, this is the first comprehensive structured pathology knowledge base; (ii) We develop a knowledge-enhanced visual-language pretraining approach, where we first project pathology-specific knowledge into latent embedding space via a language model, and use it to guide the visual representation learning; (iii) We conduct thorough experiments to validate the effectiveness of our proposed components, demonstrating significant performance improvement on various downstream tasks, including cross-modal retrieval, zero-shot classification on pathology patches, and zero-shot tumor subtyping on whole slide images (WSIs).

Weike Zhao, Chaoyi Wu, Yanjie Fan et al. · harvard

Rare diseases collectively affect over 300 million individuals worldwide, yet timely and accurate diagnosis remains a pervasive challenge. This is largely due to their clinical heterogeneity, low individual prevalence, and the limited familiarity most clinicians have with rare conditions. Here, we introduce DeepRare, the first rare disease diagnosis agentic system powered by a large language model (LLM), capable of processing heterogeneous clinical inputs. The system generates ranked diagnostic hypotheses for rare diseases, each accompanied by a transparent chain of reasoning that links intermediate analytic steps to verifiable medical evidence. DeepRare comprises three key components: a central host with a long-term memory module; specialized agent servers responsible for domain-specific analytical tasks integrating over 40 specialized tools and web-scale, up-to-date medical knowledge sources, ensuring access to the most current clinical information. This modular and scalable design enables complex diagnostic reasoning while maintaining traceability and adaptability. We evaluate DeepRare on eight datasets. The system demonstrates exceptional diagnostic performance among 2,919 diseases, achieving 100% accuracy for 1013 diseases. In HPO-based evaluations, DeepRare significantly outperforms other 15 methods, like traditional bioinformatics diagnostic tools, LLMs, and other agentic systems, achieving an average Recall@1 score of 57.18% and surpassing the second-best method (Reasoning LLM) by a substantial margin of 23.79 percentage points. For multi-modal input scenarios, DeepRare achieves 70.60% at Recall@1 compared to Exomiser's 53.20% in 109 cases. Manual verification of reasoning chains by clinical experts achieves 95.40% agreements. Furthermore, the DeepRare system has been implemented as a user-friendly web application http://raredx.cn/doctor.

Emily Alsentzer, Marie-Laure Charpignon, Bill Chen et al.

The 6th Annual Conference on Health, Inference, and Learning (CHIL 2025), hosted by the Association for Health Learning and Inference (AHLI), was held in person on June 25-27, 2025, at the University of California, Berkeley, in Berkeley, California, USA. As part of this year's program, we hosted Research Roundtables to catalyze collaborative, small-group dialogue around critical, timely topics at the intersection of machine learning and healthcare. Each roundtable was moderated by a team of senior and junior chairs who fostered open exchange, intellectual curiosity, and inclusive engagement. The sessions emphasized rigorous discussion of key challenges, exploration of emerging opportunities, and collective ideation toward actionable directions in the field. In total, eight roundtables were held by 19 roundtable chairs on topics of "Explainability, Interpretability, and Transparency," "Uncertainty, Bias, and Fairness," "Causality," "Domain Adaptation," "Foundation Models," "Learning from Small Medical Data," "Multimodal Methods," and "Scalable, Translational Healthcare Solutions."

Dexuan He, Xiao Zhou, Wenbin Guan et al. · harvard

Rare cancers comprise 20-25% of all malignancies but face major diagnostic challenges due to limited expert availability-especially in pediatric oncology, where they represent over 70% of cases. While pathology vision-language (VL) foundation models show promising zero-shot capabilities for common cancer subtyping, their clinical performance for rare cancers remains limited. Existing multi-instance learning (MIL) methods rely only on visual features, overlooking cross-modal knowledge and compromising interpretability critical for rare cancer diagnosis. To address this limitation, we propose PathPT, a novel framework that fully exploits the potential of vision-language pathology foundation models through spatially-aware visual aggregation and task-specific prompt tuning. Unlike conventional MIL, PathPT converts WSI-level supervision into fine-grained tile-level guidance by leveraging the zero-shot capabilities of VL models, thereby preserving localization on cancerous regions and enabling cross-modal reasoning through prompts aligned with histopathological semantics. We benchmark PathPT on eight rare cancer datasets(four adult and four pediatric) spanning 56 subtypes and 2,910 WSIs, as well as three common cancer datasets, evaluating four state-of-the-art VL models and four MIL frameworks under three few-shot settings. Results show that PathPT consistently delivers superior performance, achieving substantial gains in subtyping accuracy and cancerous region grounding ability. This work advances AI-assisted diagnosis for rare cancers, offering a scalable solution for improving subtyping accuracy in settings with limited access to specialized expertise.

Julian Acosta, Scott Adams, Julius Kernbach et al.

We developed a voice-driven artificial intelligence (AI) system that guides anyone - from paramedics to family members - through expert-level stroke evaluations using natural conversation, while also enabling smartphone video capture of key examination components for documentation and potential expert review. This addresses a critical gap in emergency care: current stroke recognition by first responders is inconsistent and often inaccurate, with sensitivity for stroke detection as low as 58%, causing life-threatening delays in treatment. Three non-medical volunteers used our AI system to assess ten simulated stroke patients, including cases with likely large vessel occlusion (LVO) strokes and stroke-like conditions, while we measured diagnostic accuracy, completion times, user confidence, and expert physician review of the AI-generated reports. The AI system correctly identified 84% of individual stroke signs and detected 75% of likely LVOs, completing evaluations in just over 6 minutes. Users reported high confidence (median 4.5/5) and ease of use (mean 4.67/5). The system successfully identified 86% of actual strokes but also incorrectly flagged 2 of 3 non-stroke cases as strokes. When an expert physician reviewed the AI reports with videos, they identified the correct diagnosis in 100% of cases, but felt confident enough to make preliminary treatment decisions in only 40% of cases due to observed AI errors including incorrect scoring and false information. While the current system's limitations necessitate human oversight, ongoing rapid advancements in speech-to-speech AI models suggest that future versions are poised to enable highly accurate assessments. Achieving human-level voice interaction could transform emergency medical care, putting expert-informed assessment capabilities in everyone's hands.

Weike Zhao, Chaoyi Wu, Xiaoman Zhang et al.

This paper introduces a novel, entity-aware metric, termed as Radiological Report (Text) Evaluation (RaTEScore), to assess the quality of medical reports generated by AI models. RaTEScore emphasizes crucial medical entities such as diagnostic outcomes and anatomical details, and is robust against complex medical synonyms and sensitive to negation expressions. Technically, we developed a comprehensive medical NER dataset, RaTE-NER, and trained an NER model specifically for this purpose. This model enables the decomposition of complex radiological reports into constituent medical entities. The metric itself is derived by comparing the similarity of entity embeddings, obtained from a language model, based on their types and relevance to clinical significance. Our evaluations demonstrate that RaTEScore aligns more closely with human preference than existing metrics, validated both on established public benchmarks and our newly proposed RaTE-Eval benchmark.

Xiaoman Zhang, Chaoyi Wu, Ziheng Zhao et al.

Medical Visual Question Answering (MedVQA) presents a significant opportunity to enhance diagnostic accuracy and healthcare delivery by leveraging artificial intelligence to interpret and answer questions based on medical images. In this study, we reframe the problem of MedVQA as a generation task that naturally follows the human-machine interaction and propose a generative-based model for medical visual understanding by aligning visual information from a pre-trained vision encoder with a large language model. We establish a scalable pipeline to construct a large-scale medical visual question-answering dataset, named PMC-VQA, which contains 227k VQA pairs of 149k images that cover various modalities or diseases. We train the proposed model on PMC-VQA and then fine-tune it on multiple public benchmarks, e.g., VQA-RAD, SLAKE, and Image-Clef-2019, significantly outperforming existing MedVQA models in generating relevant, accurate free-form answers. In addition, we propose a test set that has undergone manual verification, which is significantly more challenging, serving to better monitor the development of generative MedVQA methods. To facilitate comprehensive evaluation and comparison, we have maintained a leaderboard at https://paperswithcode.com/paper/pmc-vqa-visual-instruction-tuning-for-medical, offering a centralized resource for tracking progress and benchmarking state-of-the-art approaches. The PMC-VQA dataset emerges as a vital resource for the field of research, and the MedVInT presents a significant breakthrough in the area of MedVQA.

Xiaoman Zhang, Weidi Xie, Chaoqin Huang et al.

In this paper, we target self-supervised representation learning for zero-shot tumor segmentation. We make the following contributions: First, we advocate a zero-shot setting, where models from pre-training should be directly applicable for the downstream task, without using any manual annotations. Second, we take inspiration from "layer-decomposition", and innovate on the training regime with simulated tumor data. Third, we conduct extensive ablation studies to analyse the critical components in data simulation, and validate the necessity of different proxy tasks. We demonstrate that, with sufficient texture randomization in simulation, model trained on synthetic data can effortlessly generalise to segment real tumor data. Forth, our approach achieves superior results for zero-shot tumor segmentation on different downstream datasets, BraTS2018 for brain tumor segmentation and LiTS2017 for liver tumor segmentation. While evaluating the model transferability for tumor segmentation under a low-annotation regime, the proposed approach also outperforms all existing self-supervised approaches, opening up the usage of self-supervised learning in practical scenarios.

Shixiang Feng, Yuhang Zhou, Xiaoman Zhang et al.

Annotating multiple organs in 3D medical images is time-consuming and costly. Meanwhile, there exist many single-organ datasets with one specific organ annotated. This paper investigates how to learn a multi-organ segmentation model leveraging a set of binary-labeled datasets. A novel Multi-teacher Single-student Knowledge Distillation (MS-KD) framework is proposed, where the teacher models are pre-trained single-organ segmentation networks, and the student model is a multi-organ segmentation network. Considering that each teacher focuses on different organs, a region-based supervision method, consisting of logits-wise supervision and feature-wise supervision, is proposed. Each teacher supervises the student in two regions, the organ region where the teacher is considered as an expert and the background region where all teachers agree. Extensive experiments on three public single-organ datasets and a multi-organ dataset have demonstrated the effectiveness of the proposed MS-KD framework.

Yuhang Zhou, Xiaoman Zhang, Shixiang Feng et al.

Most existing approaches to train a unified multi-organ segmentation model from several single-organ datasets require simultaneously access multiple datasets during training. In the real scenarios, due to privacy and ethics concerns, the training data of the organs of interest may not be publicly available. To this end, we investigate a data-free incremental organ segmentation scenario and propose a novel incremental training framework to solve it. We use the pretrained model instead of its own training data for privacy protection. Specifically, given a pretrained $K$ organ segmentation model and a new single-organ dataset, we train a unified $K+1$ organ segmentation model without accessing any data belonging to the previous training stages. Our approach consists of two parts: the background label alignment strategy and the uncertainty-aware guidance strategy. The first part is used for knowledge transfer from the pretained model to the training model. The second part is used to extract the uncertainty information from the pretrained model to guide the whole knowledge transfer process. By combing these two strategies, more reliable information is extracted from the pretrained model without original training data. Experiments on multiple publicly available pretrained models and a multi-organ dataset MOBA have demonstrated the effectiveness of our framework.

Xiaoman Zhang, Shixiang Feng, Yuhang Zhou et al.

Automatic and accurate tumor segmentation on medical images is in high demand to assist physicians with diagnosis and treatment. However, it is difficult to obtain massive amounts of annotated training data required by the deep-learning models as the manual delineation process is often tedious and expertise required. Although self-supervised learning (SSL) scheme has been widely adopted to address this problem, most SSL methods focus only on global structure information, ignoring the key distinguishing features of tumor regions: local intensity variation and large size distribution. In this paper, we propose Scale-Aware Restoration (SAR), a SSL method for 3D tumor segmentation. Specifically, a novel proxy task, i.e. scale discrimination, is formulated to pre-train the 3D neural network combined with the self-restoration task. Thus, the pre-trained model learns multi-level local representations through multi-scale inputs. Moreover, an adversarial learning module is further introduced to learn modality invariant representations from multiple unlabeled source datasets. We demonstrate the effectiveness of our methods on two downstream tasks: i) Brain tumor segmentation, ii) Pancreas tumor segmentation. Compared with the state-of-the-art 3D SSL methods, our proposed approach can significantly improve the segmentation accuracy. Besides, we analyze its advantages from multiple perspectives such as data efficiency, performance, and convergence speed.

Xiaoman Zhang, Ziyuan Zhao, Cen Chen et al.

Bone age assessment is an important clinical trial to measure skeletal child maturity and diagnose of growth disorders. Conventional approaches such as the Tanner-Whitehouse (TW) and Greulich and Pyle (GP) may not perform well due to their large inter-observer and intra-observer variations. In this paper, we propose a finger joint localization strategy to filter out most non-informative parts of images. When combining with the conventional full image-based deep network, we observe a much-improved performance. % Our approach utilizes full hand and specific joints images for skeletal maturity prediction. In this study, we applied powerful deep neural network and explored a process in the forecast of skeletal bone age with the specifically combine joints images to increase the performance accuracy compared with the whole hand images.

Ziyuan Zhao, Xiaoman Zhang, Cen Chen et al.

Segmentation stands at the forefront of many high-level vision tasks. In this study, we focus on segmenting finger bones within a newly introduced semi-supervised self-taught deep learning framework which consists of a student network and a stand-alone teacher module. The whole system is boosted in a life-long learning manner wherein each step the teacher module provides a refinement for the student network to learn with newly unlabeled data. Experimental results demonstrate the superiority of the proposed method over conventional supervised deep learning methods.