Chaoyi Wu

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.

Chaoyi Wu, Pengcheng Qiu, Jinxin Liu et al.

In this study, we present MedS-Bench, a comprehensive benchmark designed to evaluate the performance of large language models (LLMs) in clinical contexts. Unlike existing benchmarks that focus on multiple-choice question answering, MedS-Bench spans 11 high-level clinical tasks, including clinical report summarization, treatment recommendations, diagnosis, named entity recognition, and medical concept explanation, among others. We evaluated six leading LLMs, e.g., MEDITRON, Mistral, InternLM 2, Llama 3, GPT-4, and Claude-3.5 using few-shot prompting, and found that even the most sophisticated models struggle with these complex tasks. To address these limitations, we developed MedS-Ins, a large-scale instruction tuning dataset for medicine. MedS-Ins comprises 58 medically oriented language corpora, totaling 13.5 million samples across 122 tasks. To demonstrate the dataset's utility, we conducted a proof-of-concept experiment by performing instruction tuning on a lightweight, open-source medical language model. The resulting model, MMedIns-Llama 3, significantly outperformed existing models across nearly all clinical tasks. To promote further advancements in the application of LLMs to clinical challenges, we have made the MedS-Ins dataset fully accessible and invite the research community to contribute to its expansion.Additionally, we have launched a dynamic leaderboard for MedS-Bench, which we plan to regularly update the test set to track progress and enhance the adaptation of general LLMs to the medical domain. Leaderboard: https://henrychur.github.io/MedS-Bench/. Github: https://github.com/MAGIC-AI4Med/MedS-Ins.

Cheng Liang, Pengcheng Qiu, Ya Zhang et al.

Large language models (LLMs) are increasingly proposed as clinical agents, yet static, single-turn benchmarks cannot capture how a model dynamically delivers care across an encounter: gathering information, planning treatment, and adapting longitudinal management across successive patient states. Medical education has long addressed an analogous challenge through standardized patients (SPs): trained actors who consistently portray clinical cases, enabling realistic practice and objective, scripted assessment. Here we introduce MedSP1000, an SP-derived interactive benchmark for clinical-agent evaluation, including 1,638 SP cases with 24,602 trajectory-level peer-reviewed rubrics. MedSP1000 converts peer-reviewed SP teaching cases into executable scenarios with defined SP case scripts, clinical environment contexts, and human-validated structured rubric. In each simulation evaluation run, a clinical agent interacts in closed loop with a patient agent and an environment controller, and its behaviour is scored throughout the encounter against expert criteria specified in the original materials. Applying MedSP1000 to a range of general-purpose and medically specialized LLMs, we find that performance on static benchmarks does not reliably translate to such educational scenarios. The best-performing model, GPT-5.5, completes only 60.4% of expert-defined rubric items, whereas the strongest medically specialized model reaches 40.0%; increasing test-time compute produces no measurable gain. These results suggest that current LLMs, including agentic systems tuned for medicine, are not yet reliable enough to be safely integrated into actual clinical practice. More broadly, MedSP1000 shows how process-level, SP-style evaluation can reveal clinically relevant failure modes that single-turn benchmarks miss.

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.

Chaoyi Wu, Feng Chang, Xiao Su et al.

Metastasis on lymph nodes (LNs), the most common way of spread for primary tumor cells, is a sign of increased mortality. However, metastatic LNs are time-consuming and challenging to detect even for professional radiologists due to their small sizes, high sparsity, and ambiguity in appearance. It is desired to leverage recent development in deep learning to automatically detect metastatic LNs. Besides a two-stage detection network, we here introduce an additional branch to leverage information about LN stations, an important reference for radiologists during metastatic LN diagnosis, as supplementary information for metastatic LN detection. The branch targets to solve a closely related task on the LN station level, i.e., classifying whether an LN station contains metastatic LN or not, so as to learn representations for LN stations. Considering that a metastatic LN station is expected to significantly affect the nearby ones, a GCN-based structure is adopted by the branch to model the relationship among different LN stations. At the classification stage of metastatic LN detection, the above learned LN station features, as well as the features reflecting the distance between the LN candidate and the LN stations, are integrated with the LN features. We validate our method on a dataset containing 114 intravenous contrast-enhanced Computed Tomography (CT) images of oral squamous cell carcinoma (OSCC) patients and show that it outperforms several state-of-the-art methods on the mFROC, maxF1, and AUC scores,respectively.

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.

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.

Cheng Liang, Chaoyi Wu, Weike Zhao et al.

Recent progress in large-scale CLIP-like vision-language models(VLMs) has greatly advanced medical image analysis. However, most existing medical VLMs still rely on coarse image-text contrastive objectives and fail to capture the systematic visual knowledge encoded in well-defined medical phenotype ontologies. To address this gap, we construct PhenoKG, the first large-scale, phenotype-centric multimodal knowledge graph that encompasses over 520K high-quality image-text pairs linked to more than 3,000 phenotypes. Building upon PhenoKG, we propose PhenoLIP, a novel pretraining framework that explicitly incorporates structured phenotype knowledge into medical VLMs through a two-stage process. We first learn a knowledge-enhanced phenotype embedding space from textual ontology data and then distill this structured knowledge into multimodal pretraining via a teacher-guided knowledge distillation objective. To support evaluation, we further introduce PhenoBench, an expert-verified benchmark designed for phenotype recognition, comprising over 7,800 image--caption pairs covering more than 1,000 phenotypes. Extensive experiments demonstrate that PhenoLIP outperforms previous state-of-the-art baselines, improving upon BiomedCLIP in phenotype classification accuracy by 8.85\% and BIOMEDICA in cross-modal retrieval by 15.03%, underscoring the value of integrating phenotype-centric priors into medical VLMs for structured and interpretable medical image understanding.

Pengcheng Qiu, Chaoyi Wu, Shuyu Liu et al.

Recent advancements in reasoning-enhanced large language models (LLMs), such as DeepSeek-R1 and OpenAI-o3, have demonstrated significant progress. However, their application in professional medical contexts remains underexplored, particularly in evaluating the quality of their reasoning processes alongside final outputs. Here, we introduce MedR-Bench, a benchmarking dataset of 1,453 structured patient cases, annotated with reasoning references derived from clinical case reports. Spanning 13 body systems and 10 specialties, it includes both common and rare diseases. To comprehensively evaluate LLM performance, we propose a framework encompassing three critical examination recommendation, diagnostic decision-making, and treatment planning, simulating the entire patient care journey. To assess reasoning quality, we present the Reasoning Evaluator, a novel automated system that objectively scores free-text reasoning responses based on efficiency, actuality, and completeness using dynamic cross-referencing and evidence checks. Using this benchmark, we evaluate five state-of-the-art reasoning LLMs, including DeepSeek-R1, OpenAI-o3-mini, and Gemini-2.0-Flash Thinking, etc. Our results show that current LLMs achieve over 85% accuracy in relatively simple diagnostic tasks when provided with sufficient examination results. However, performance declines in more complex tasks, such as examination recommendation and treatment planning. While reasoning outputs are generally reliable, with factuality scores exceeding 90%, critical reasoning steps are frequently missed. These findings underscore both the progress and limitations of clinical LLMs. Notably, open-source models like DeepSeek-R1 are narrowing the gap with proprietary systems, highlighting their potential to drive accessible and equitable advancements in healthcare.

Ziqing Fan, Cheng Liang, Chaoyi Wu et al.

Recent advances in reasoning-enhanced large language models (LLMs) and multimodal LLMs (MLLMs) have significantly improved performance in complex tasks, yet medical AI models often overlook the structured reasoning processes inherent in clinical practice. In this work, we present ChestX-Reasoner, a radiology diagnosis MLLM designed to leverage process supervision mined directly from clinical reports, reflecting the step-by-step reasoning followed by radiologists. We construct a large dataset by extracting and refining reasoning chains from routine radiology reports. Our two-stage training framework combines supervised fine-tuning and reinforcement learning guided by process rewards to better align model reasoning with clinical standards. We introduce RadRBench-CXR, a comprehensive benchmark featuring 59K visual question answering samples with 301K clinically validated reasoning steps, and propose RadRScore, a metric evaluating reasoning factuality, completeness, and effectiveness. ChestX-Reasoner outperforms existing medical and general-domain MLLMs in both diagnostic accuracy and reasoning ability, achieving 16%, 5.9%, and 18% improvements in reasoning ability compared to the best medical MLLM, the best general MLLM, and its base model, respectively, as well as 3.3%, 24%, and 27% improvements in outcome accuracy. All resources are open-sourced to facilitate further research in medical reasoning MLLMs.

Qiaoyu Zheng, Chaoyi Wu, Pengcheng Qiu et al.

We introduce RadA-BenchPlat, an evaluation platform that benchmarks the performance of large language models (LLMs) act as agent cores in radiology environments using 2,200 radiologist-verified synthetic patient records covering six anatomical regions, five imaging modalities, and 2,200 disease scenarios, resulting in 24,200 question-answer pairs that simulate diverse clinical situations. The platform also defines ten categories of tools for agent-driven task solving and evaluates seven leading LLMs, revealing that while models like Claude-3.7-Sonnet can achieve a 67.1% task completion rate in routine settings, they still struggle with complex task understanding and tool coordination, limiting their capacity to serve as the central core of automated radiology systems. By incorporating four advanced prompt engineering strategies--where prompt-backpropagation and multi-agent collaboration contributed 16.8% and 30.7% improvements, respectively--the performance for complex tasks was enhanced by 48.2% overall. Furthermore, automated tool building was explored to improve robustness, achieving a 65.4% success rate, thereby offering promising insights for the future integration of fully automated radiology applications into clinical practice. All of our code and data are openly available at https://github.com/MAGIC-AI4Med/RadABench.

Qiaoyu Zheng, Yuze Sun, Chaoyi Wu et al.

Accurate diagnosis with medical large language models is hindered by knowledge gaps and hallucinations. Retrieval and tool-augmented methods help, but their impact is limited by weak use of external knowledge and poor feedback-reasoning traceability. To address these challenges, We introduce Deep-DxSearch, an agentic RAG system trained end-to-end with reinforcement learning (RL) that enables steer tracebale retrieval-augmented reasoning for medical diagnosis. In Deep-DxSearch, we first construct a large-scale medical retrieval corpus comprising patient records and reliable medical knowledge sources to support retrieval-aware reasoning across diagnostic scenarios. More crutially, we frame the LLM as the core agent and the retrieval corpus as its environment, using tailored rewards on format, retrieval, reasoning structure, and diagnostic accuracy, thereby evolving the agentic RAG policy from large-scale data through RL. Experiments demonstrate that our end-to-end agentic RL training framework consistently outperforms prompt-engineering and training-free RAG approaches across multiple data centers. After training, Deep-DxSearch achieves substantial gains in diagnostic accuracy, surpassing strong diagnostic baselines such as GPT-4o, DeepSeek-R1, and other medical-specific frameworks for both common and rare disease diagnosis under in-distribution and out-of-distribution settings. Moreover, ablation studies on reward design and retrieval corpus components confirm their critical roles, underscoring the uniqueness and effectiveness of our approach compared with traditional implementations. Finally, case studies and interpretability analyses highlight improvements in Deep-DxSearch's diagnostic policy, providing deeper insight into its performance gains and supporting clinicians in delivering more reliable and precise preliminary diagnoses. See https://github.com/MAGIC-AI4Med/Deep-DxSearch.

Yusheng Liao, Chaoyi Wu, Junwei Liu et al.

Electronic Health Records (EHRs) contain rich yet complex information, and their automated analysis is critical for clinical decision-making. Despite recent advances of large language models (LLMs) in clinical workflows, their ability to analyze EHRs remains limited due to narrow task coverage and lack of EHR-oriented reasoning capabilities. This paper aims to bridge the gap, specifically, we present EHR-Ins, a large-scale, comprehensive EHR reasoning instruction dataset, comprising 300k high-quality reasoning cases and 4M non-reasoning cases across 42 distinct EHR tasks. Its core innovation is a thinking-graph-driven framework that enables to generate high-quality reasoning data at scale. Based on it, we develop EHR-R1, a series of reasoning-enhanced LLMs with up to 72B parameters tailored for EHR analysis. Through a multi-stage training paradigm, including domain adaptation, reasoning enhancement, and reinforcement learning, EHR-R1 systematically acquires domain knowledge and diverse reasoning capabilities, enabling accurate and robust EHR analysis. Lastly, we introduce EHR-Bench, a new benchmark curated from MIMIC-IV, spanning 42 tasks, to comprehensively assess reasoning and prediction across EHR scenarios. In experiments, we show that the resulting EHR-R1 consistently outperforms state-of-the-art commercial and open-source LLMs (including DeepSeek-V3 and GPT-4o), surpassing GPT-4o by over 30 points on MIMIC-Bench and achieving a 10\% higher zero-shot AUROC on EHRSHOT. Collectively, EHR-Ins, EHR-R1, and EHR-Bench have significantly advanced the development for more reliable and clinically relevant EHR analysis.

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.

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.

Pengcheng Qiu, Chaoyi Wu, Junwei Liu et al.

In this paper, we present a framework for training large language models (LLMs) as diagnostic agents with reinforcement learning, enabling them to manage multi-turn diagnostic processes, adaptively select examinations, and commit to final diagnoses. Unlike instruction-tuned models trained on static case summaries, our method acquires diagnostic strategies through interactive exploration and outcome-based feedback. Our contributions are fourfold: (i) We present DiagGym, a diagnostics world model trained with electronic health records that emits examination outcomes conditioned on patient history and recommended examination, serving as a virtual clinical environment for realistic diagnosis training and evaluation; (ii) We train DiagAgent via end-to-end, multi-turn reinforcement learning to learn diagnostic policies that optimize both information yield and diagnostic accuracy; (iii) We introduce DiagBench, a diagnostic benchmark comprising 750 cases with physician-validated examination recommendations and 99 cases annotated with 973 physician-written rubrics on diagnosis process; (iv) we demonstrate superior performance across diverse diagnostic settings. DiagAgent significantly outperforms 10 state-of-the-art LLMs, including DeepSeek-v3 and GPT-4o, as well as two prompt-engineered agents. In single-turn settings, DiagAgent achieves 9.34% higher diagnostic accuracy and 44.03% improvement in examination recommendation hit ratio. In end-to-end settings, it delivers 15.12% increase in diagnostic accuracy and 23.09% boost in examination recommendation F1 score. In rubric-based evaluation, it surpasses the next-best model, Claude-sonnet-4, by 7.1% in weighted rubric score. These findings indicate that learning policies in interactive clinical environments confers dynamic and clinically meaningful diagnostic management abilities unattainable through passive training alone.

Tengfei Zhang, Ziheng Zhao, Chaoyi Wu et al.

Developing advanced medical imaging retrieval systems is challenging due to the varying definitions of `similar images' across different medical contexts. This challenge is compounded by the lack of large-scale, high-quality medical imaging retrieval datasets and benchmarks. In this paper, we propose a novel methodology that leverages dense radiology reports to define image-wise similarity ordering at multiple granularities in a scalable and fully automatic manner. Using this approach, we construct two comprehensive medical imaging retrieval datasets: MIMIC-IR for Chest X-rays and CTRATE-IR for CT scans, providing detailed image-image ranking annotations conditioned on diverse anatomical structures. Furthermore, we develop two retrieval systems, RadIR-CXR and model-ChestCT, which demonstrate superior performance in traditional image-image and image-report retrieval tasks. These systems also enable flexible, effective image retrieval conditioned on specific anatomical structures described in text, achieving state-of-the-art results on 77 out of 78 metrics.

Jinghao Feng, Qiaoyu Zheng, Chaoyi Wu et al.

Agentic AI systems have gained significant attention for their ability to autonomously perform complex tasks. However, their reliance on well-prepared tools limits their applicability in the medical domain, which requires to train specialized models. In this paper, we make three contributions: (i) We present M3Builder, a novel multi-agent system designed to automate machine learning (ML) in medical imaging. At its core, M3Builder employs four specialized agents that collaborate to tackle complex, multi-step medical ML workflows, from automated data processing and environment configuration to self-contained auto debugging and model training. These agents operate within a medical imaging ML workspace, a structured environment designed to provide agents with free-text descriptions of datasets, training codes, and interaction tools, enabling seamless communication and task execution. (ii) To evaluate progress in automated medical imaging ML, we propose M3Bench, a benchmark comprising four general tasks on 14 training datasets, across five anatomies and three imaging modalities, covering both 2D and 3D data. (iii) We experiment with seven state-of-the-art large language models serving as agent cores for our system, such as Claude series, GPT-4o, and DeepSeek-V3. Compared to existing ML agentic designs, M3Builder shows superior performance on completing ML tasks in medical imaging, achieving a 94.29% success rate using Claude-3.7-Sonnet as the agent core, showing huge potential towards fully automated machine learning in medical imaging.

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.