Medicine, by its nature, is a multifaceted domain that requires the synthesis of information across various modalities. Medical generative vision-language models (VLMs) make a first step in this direction and promise many exciting clinical applications. However, existing models typically have to be fine-tuned on sizeable down-stream datasets, which poses a significant limitation as in many medical applications data is scarce, necessitating models that are capable of learning from few examples in real-time. Here we propose Med-Flamingo, a multimodal few-shot learner adapted to the medical domain. Based on OpenFlamingo-9B, we continue pre-training on paired and interleaved medical image-text data from publications and textbooks. Med-Flamingo unlocks few-shot generative medical visual question answering (VQA) abilities, which we evaluate on several datasets including a novel challenging open-ended VQA dataset of visual USMLE-style problems. Furthermore, we conduct the first human evaluation for generative medical VQA where physicians review the problems and blinded generations in an interactive app. Med-Flamingo improves performance in generative medical VQA by up to 20\% in clinician's rating and firstly enables multimodal medical few-shot adaptations, such as rationale generation. We release our model, code, and evaluation app under https://github.com/snap-stanford/med-flamingo.
Kathryn Wantlin, Chenwei Wu, Shih-Cheng Huang et al. · stanford
Medical data poses a daunting challenge for AI algorithms: it exists in many different modalities, experiences frequent distribution shifts, and suffers from a scarcity of examples and labels. Recent advances, including transformers and self-supervised learning, promise a more universal approach that can be applied flexibly across these diverse conditions. To measure and drive progress in this direction, we present BenchMD: a benchmark that tests how well unified, modality-agnostic methods, including architectures and training techniques (e.g. self-supervised learning, ImageNet pretraining),perform on a diverse array of clinically-relevant medical tasks. BenchMD combines 19 publicly available datasets for 7 medical modalities, including 1D sensor data, 2D images, and 3D volumetric scans. Our benchmark reflects real-world data constraints by evaluating methods across a range of dataset sizes, including challenging few-shot settings that incentivize the use of pretraining. Finally, we evaluate performance on out-of-distribution data collected at different hospitals than the training data, representing naturally-occurring distribution shifts that frequently degrade the performance of medical AI models. Our baseline results demonstrate that no unified learning technique achieves strong performance across all modalities, leaving ample room for improvement on the benchmark. Code is released at https://github.com/rajpurkarlab/BenchMD.
Automated generation of clinically accurate radiology reports can improve patient care. Previous report generation methods that rely on image captioning models often generate incoherent and incorrect text due to their lack of relevant domain knowledge, while retrieval-based attempts frequently retrieve reports that are irrelevant to the input image. In this work, we propose Contrastive X-Ray REport Match (X-REM), a novel retrieval-based radiology report generation module that uses an image-text matching score to measure the similarity of a chest X-ray image and radiology report for report retrieval. We observe that computing the image-text matching score with a language-image model can effectively capture the fine-grained interaction between image and text that is often lost when using cosine similarity. X-REM outperforms multiple prior radiology report generation modules in terms of both natural language and clinical metrics. Human evaluation of the generated reports suggests that X-REM increased the number of zero-error reports and decreased the average error severity compared to the baseline retrieval approach. Our code is available at: https://github.com/rajpurkarlab/X-REM
The recent success of general-domain large language models (LLMs) has significantly changed the natural language processing paradigm towards a unified foundation model across domains and applications. In this paper, we focus on assessing the performance of GPT-4, the most capable LLM so far, on the text-based applications for radiology reports, comparing against state-of-the-art (SOTA) radiology-specific models. Exploring various prompting strategies, we evaluated GPT-4 on a diverse range of common radiology tasks and we found GPT-4 either outperforms or is on par with current SOTA radiology models. With zero-shot prompting, GPT-4 already obtains substantial gains ($\approx$ 10% absolute improvement) over radiology models in temporal sentence similarity classification (accuracy) and natural language inference ($F_1$). For tasks that require learning dataset-specific style or schema (e.g. findings summarisation), GPT-4 improves with example-based prompting and matches supervised SOTA. Our extensive error analysis with a board-certified radiologist shows GPT-4 has a sufficient level of radiology knowledge with only occasional errors in complex context that require nuanced domain knowledge. For findings summarisation, GPT-4 outputs are found to be overall comparable with existing manually-written impressions.
Alexander Ke, Shih-Cheng Huang, Chloe P O'Connell et al.
Pretraining on large natural image classification datasets such as ImageNet has aided model development on data-scarce 2D medical tasks. 3D medical tasks often have much less data than 2D medical tasks, prompting practitioners to rely on pretrained 2D models to featurize slices. However, these 2D models have been surpassed by 3D models on 3D computer vision benchmarks since they do not natively leverage cross-sectional or temporal information. In this study, we explore whether natural video pretraining for 3D models can enable higher performance on smaller datasets for 3D medical tasks. We demonstrate video pretraining improves the average performance of seven 3D models on two chest CT datasets, regardless of finetuning dataset size, and that video pretraining allows 3D models to outperform 2D baselines. Lastly, we observe that pretraining on the large-scale out-of-domain Kinetics dataset improves performance more than pretraining on a typically-sized in-domain CT dataset. Our results show consistent benefits of video pretraining across a wide array of architectures, tasks, and training dataset sizes, supporting a shift from small-scale in-domain pretraining to large-scale out-of-domain pretraining for 3D medical tasks. Our code is available at: https://github.com/rajpurkarlab/chest-ct-pretraining
Breast Magnetic Resonance Imaging (MRI) demonstrates the highest sensitivity for breast cancer detection among imaging modalities and is standard practice for high-risk women. Interpreting the multi-sequence MRI is time-consuming and prone to subjective variation. We develop a large mixture-of-modality-experts model (MOME) that integrates multiparametric MRI information within a unified structure, leveraging breast MRI scans from 5,205 female patients in China for model development and validation. MOME matches four senior radiologists' performance in identifying breast cancer and outperforms a junior radiologist. The model is able to reduce unnecessary biopsies in Breast Imaging-Reporting and Data System (BI-RADS) 4 patients, classify triple-negative breast cancer, and predict pathological complete response to neoadjuvant chemotherapy. MOME further supports inference with missing modalities and provides decision explanations by highlighting lesions and measuring modality contributions. To summarize, MOME exemplifies an accurate and robust multimodal model for noninvasive, personalized management of breast cancer patients via multiparametric MRI. Code is available at https://github.com/LLYXC/MOME/tree/main.
Vignav Ramesh, Nathan Andrew Chi, Pranav Rajpurkar
Current deep learning models trained to generate radiology reports from chest radiographs are capable of producing clinically accurate, clear, and actionable text that can advance patient care. However, such systems all succumb to the same problem: making hallucinated references to non-existent prior reports. Such hallucinations occur because these models are trained on datasets of real-world patient reports that inherently refer to priors. To this end, we propose two methods to remove references to priors in radiology reports: (1) a GPT-3-based few-shot approach to rewrite medical reports without references to priors; and (2) a BioBERT-based token classification approach to directly remove words referring to priors. We use the aforementioned approaches to modify MIMIC-CXR, a publicly available dataset of chest X-rays and their associated free-text radiology reports; we then retrain CXR-RePaiR, a radiology report generation system, on the adapted MIMIC-CXR dataset. We find that our re-trained model--which we call CXR-ReDonE--outperforms previous report generation methods on clinical metrics, achieving an average BERTScore of 0.2351 (2.57% absolute improvement). We expect our approach to be broadly valuable in enabling current radiology report generation systems to be more directly integrated into clinical pipelines.
Vivek Shankar, Xiaoli Yang, Vrishab Krishna et al.
The accurate classification of lymphoma subtypes using hematoxylin and eosin (H&E)-stained tissue is complicated by the wide range of morphological features these cancers can exhibit. We present LymphoML - an interpretable machine learning method that identifies morphologic features that correlate with lymphoma subtypes. Our method applies steps to process H&E-stained tissue microarray cores, segment nuclei and cells, compute features encompassing morphology, texture, and architecture, and train gradient-boosted models to make diagnostic predictions. LymphoML's interpretable models, developed on a limited volume of H&E-stained tissue, achieve non-inferior diagnostic accuracy to pathologists using whole-slide images and outperform black box deep-learning on a dataset of 670 cases from Guatemala spanning 8 lymphoma subtypes. Using SHapley Additive exPlanation (SHAP) analysis, we assess the impact of each feature on model prediction and find that nuclear shape features are most discriminative for DLBCL (F1-score: 78.7%) and classical Hodgkin lymphoma (F1-score: 74.5%). Finally, we provide the first demonstration that a model combining features from H&E-stained tissue with features from a standardized panel of 6 immunostains results in a similar diagnostic accuracy (85.3%) to a 46-stain panel (86.1%).
Luke W. Sagers, James A. Diao, Matthew Groh et al.
Dermatological classification algorithms developed without sufficiently diverse training data may generalize poorly across populations. While intentional data collection and annotation offer the best means for improving representation, new computational approaches for generating training data may also aid in mitigating the effects of sampling bias. In this paper, we show that DALL$\cdot$E 2, a large-scale text-to-image diffusion model, can produce photorealistic images of skin disease across skin types. Using the Fitzpatrick 17k dataset as a benchmark, we demonstrate that augmenting training data with DALL$\cdot$E 2-generated synthetic images improves classification of skin disease overall and especially for underrepresented groups.
Luke W. Sagers, James A. Diao, Luke Melas-Kyriazi et al.
While hundreds of artificial intelligence (AI) algorithms are now approved or cleared by the US Food and Drugs Administration (FDA), many studies have shown inconsistent generalization or latent bias, particularly for underrepresented populations. Some have proposed that generative AI could reduce the need for real data, but its utility in model development remains unclear. Skin disease serves as a useful case study in synthetic image generation due to the diversity of disease appearance, particularly across the protected attribute of skin tone. Here we show that latent diffusion models can scalably generate images of skin disease and that augmenting model training with these data improves performance in data-limited settings. These performance gains saturate at synthetic-to-real image ratios above 10:1 and are substantially smaller than the gains obtained from adding real images. As part of our analysis, we generate and analyze a new dataset of 458,920 synthetic images produced using several generation strategies. Our results suggest that synthetic data could serve as a force-multiplier for model development, but the collection of diverse real-world data remains the most important step to improve medical AI algorithms.
We present RadGraph2, a novel dataset for extracting information from radiology reports that focuses on capturing changes in disease state and device placement over time. We introduce a hierarchical schema that organizes entities based on their relationships and show that using this hierarchy during training improves the performance of an information extraction model. Specifically, we propose a modification to the DyGIE++ framework, resulting in our model HGIE, which outperforms previous models in entity and relation extraction tasks. We demonstrate that RadGraph2 enables models to capture a wider variety of findings and perform better at relation extraction compared to those trained on the original RadGraph dataset. Our work provides the foundation for developing automated systems that can track disease progression over time and develop information extraction models that leverage the natural hierarchy of labels in the medical domain.
Automatically generated reports from medical images promise to improve the workflow of radiologists. Existing methods consider an image-to-report modeling task by directly generating a fully-fledged report from an image. However, this conflates the content of the report (e.g., findings and their attributes) with its style (e.g., format and choice of words), which can lead to clinically inaccurate reports. To address this, we propose a two-step approach for radiology report generation. First, we extract the content from an image; then, we verbalize the extracted content into a report that matches the style of a specific radiologist. For this, we leverage RadGraph -- a graph representation of reports -- together with large language models (LLMs). In our quantitative evaluations, we find that our approach leads to beneficial performance. Our human evaluation with clinical raters highlights that the AI-generated reports are indistinguishably tailored to the style of individual radiologist despite leveraging only a few examples as context.
Aakash Mishra, Rajat Mittal, Christy Jestin et al.
Recent advances in zero-shot learning have enabled the use of paired image-text data to replace structured labels, replacing the need for expert annotated datasets. Models such as CLIP-based CheXzero utilize these advancements in the domain of chest X-ray interpretation. We hypothesize that domain pre-trained models such as CXR-BERT, BlueBERT, and ClinicalBERT offer the potential to improve the performance of CLIP-like models with specific domain knowledge by replacing BERT weights at the cost of breaking the original model's alignment. We evaluate the performance of zero-shot classification models with domain-specific pre-training for detecting low-prevalence pathologies. Even though replacing the weights of the original CLIP-BERT degrades model performance on commonly found pathologies, we show that pre-trained text towers perform exceptionally better on low-prevalence diseases. This motivates future ensemble models with a combination of differently trained language models for maximal performance.
We propose the Multimodal Clinical Benchmark for Emergency Care (MC-BEC), a comprehensive benchmark for evaluating foundation models in Emergency Medicine using a dataset of 100K+ continuously monitored Emergency Department visits from 2020-2022. MC-BEC focuses on clinically relevant prediction tasks at timescales from minutes to days, including predicting patient decompensation, disposition, and emergency department (ED) revisit, and includes a standardized evaluation framework with train-test splits and evaluation metrics. The multimodal dataset includes a wide range of detailed clinical data, including triage information, prior diagnoses and medications, continuously measured vital signs, electrocardiogram and photoplethysmograph waveforms, orders placed and medications administered throughout the visit, free-text reports of imaging studies, and information on ED diagnosis, disposition, and subsequent revisits. We provide performance baselines for each prediction task to enable the evaluation of multimodal, multitask models. We believe that MC-BEC will encourage researchers to develop more effective, generalizable, and accessible foundation models for multimodal clinical data.
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.
Vishwanatha M. Rao, Serena Zhang, Julian N. Acosta et al.
Accurately interpreting medical images and writing radiology reports is a critical but challenging task in healthcare. Both human-written and AI-generated reports can contain errors, ranging from clinical inaccuracies to linguistic mistakes. To address this, we introduce ReXErr, a methodology that leverages Large Language Models to generate representative errors within chest X-ray reports. Working with board-certified radiologists, we developed error categories that capture common mistakes in both human and AI-generated reports. Our approach uses a novel sampling scheme to inject diverse errors while maintaining clinical plausibility. ReXErr demonstrates consistency across error categories and produces errors that closely mimic those found in real-world scenarios. This method has the potential to aid in the development and evaluation of report correction algorithms, potentially enhancing the quality and reliability of radiology reporting.
Given the rapidly expanding capabilities of generative AI models for radiology, there is a need for robust metrics that can accurately measure the quality of AI-generated radiology reports across diverse hospitals. We develop ReXamine-Global, a LLM-powered, multi-site framework that tests metrics across different writing styles and patient populations, exposing gaps in their generalization. First, our method tests whether a metric is undesirably sensitive to reporting style, providing different scores depending on whether AI-generated reports are stylistically similar to ground-truth reports or not. Second, our method measures whether a metric reliably agrees with experts, or whether metric and expert scores of AI-generated report quality diverge for some sites. Using 240 reports from 6 hospitals around the world, we apply ReXamine-Global to 7 established report evaluation metrics and uncover serious gaps in their generalizability. Developers can apply ReXamine-Global when designing new report evaluation metrics, ensuring their robustness across sites. Additionally, our analysis of existing metrics can guide users of those metrics towards evaluation procedures that work reliably at their sites of interest.
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.
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.
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
Hong-Yu Zhou, Julián Nicolás Acosta, Subathra Adithan et al.
Current medical AI systems are often limited to narrow applications, hindering widespread adoption. We present MedVersa, a generalist foundation model trained on tens of millions of compiled medical instances. MedVersa unlocks generalist learning from multimodal inputs and outputs, representing the first example of a generalist model reaching competitive performance with leading specialized solutions across a variety of medical imaging scenarios. MedVersa achieves state-of-the-art performance in nine tasks, sometimes outperforming counterparts by over 10%. Radiologist evaluation shows MedVersa-generated reports get superior performance in 95% of normal studies, while matching or exceeding human reports in 71% of cases overall. User studies showed notable reductions in report writing time and discrepancies with the use of MedVersa. Our findings underscore the value of flexible, multimodal AI systems in advancing medical image interpretation and supporting clinical expertise.
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.
The integration of Large Language Models (LLMs) into medical applications has sparked widespread interest across the healthcare industry, from drug discovery and development to clinical decision support, assisting telemedicine, medical devices, and healthcare insurance applications. This perspective paper aims to discuss the inner workings of building LLM-powered medical AI applications and introduces a comprehensive framework for their development. We review existing literature and outline the unique challenges of applying LLMs in specialized medical contexts. Additionally, we introduce a three-step framework to organize medical LLM research activities: 1) Modeling: breaking down complex medical workflows into manageable steps for developing medical-specific models; 2) Optimization: optimizing the model performance with crafted prompts and integrating external knowledge and tools, and 3) System engineering: decomposing complex tasks into subtasks and leveraging human expertise for building medical AI applications. Furthermore, we offer a detailed use case playbook that describes various LLM-powered medical AI applications, such as optimizing clinical trial design, enhancing clinical decision support, and advancing medical imaging analysis. Finally, we discuss various challenges and considerations for building medical AI applications with LLMs, such as handling hallucination issues, data ownership and compliance, privacy, intellectual property considerations, compute cost, sustainability issues, and responsible AI requirements.
AI-assisted imaging made substantial advances in tumor diagnosis and management. However, a major barrier to developing robust oncology foundation models is the scarcity of large-scale, high-quality annotated datasets, which are limited by privacy restrictions and the high cost of manual labeling. To address this gap, we present PASTA, a pan-tumor radiology foundation model built on PASTA-Gen, a synthetic data framework that generated 30,000 3D CT scans with pixel-level lesion masks and structured reports of tumors across ten organ systems. Leveraging this resource, PASTA achieves state-of-the-art performance on 45 of 46 oncology tasks, including non-contrast CT tumor screening, lesion segmentation, structured reporting, tumor staging, survival prediction, and MRI-modality transfer. To assess clinical applicability, we developed PASTA-AID, a clinical decision support system, and ran a retrospective simulated clinical trial across two scenarios. For pan-tumor screening on plain CT with fixed reading time, PASTA-AID increased radiologists' throughput by 11.1-25.1% and improved sensitivity by 17.0-31.4% and precision by 10.5-24.9%; additionally, in a diagnosis-aid workflow, it reduced segmentation time by up to 78.2% and reporting time by up to 36.5%. Beyond gains in accuracy and efficiency, PASTA-AID narrowed the expertise gap, enabling less-experienced radiologists to approach expert-level performance. Together, this work establishes an end-to-end, synthetic data-driven pipeline spanning data generation, model development, and clinical validation, thereby demonstrating substantial potential for pan-tumor research and clinical translation.
Serena Zhang, Sraavya Sambara, Oishi Banerjee et al.
Generating accurate radiology reports from medical images is a clinically important but challenging task. While current Vision Language Models (VLMs) show promise, they are prone to generating hallucinations, potentially compromising patient care. We introduce RadFlag, a black-box method to enhance the accuracy of radiology report generation. Our method uses a sampling-based flagging technique to find hallucinatory generations that should be removed. We first sample multiple reports at varying temperatures and then use a Large Language Model (LLM) to identify claims that are not consistently supported across samples, indicating that the model has low confidence in those claims. Using a calibrated threshold, we flag a fraction of these claims as likely hallucinations, which should undergo extra review or be automatically rejected. Our method achieves high precision when identifying both individual hallucinatory sentences and reports that contain hallucinations. As an easy-to-use, black-box system that only requires access to a model's temperature parameter, RadFlag is compatible with a wide range of radiology report generation models and has the potential to broadly improve the quality of automated radiology reporting.
Tumor is a leading cause of death worldwide, with an estimated 10 million deaths attributed to tumor-related diseases every year. AI-driven tumor recognition unlocks new possibilities for more precise and intelligent tumor screening and diagnosis. However, the progress is heavily hampered by the scarcity of annotated datasets, which demands extensive annotation efforts by radiologists. To tackle this challenge, we introduce FreeTumor, an innovative Generative AI (GAI) framework to enable large-scale tumor synthesis for mitigating data scarcity. Specifically, FreeTumor effectively leverages a combination of limited labeled data and large-scale unlabeled data for tumor synthesis training. Unleashing the power of large-scale data, FreeTumor is capable of synthesizing a large number of realistic tumors on images for augmenting training datasets. To this end, we create the largest training dataset for tumor synthesis and recognition by curating 161,310 publicly available Computed Tomography (CT) volumes from 33 sources, with only 2.3% containing annotated tumors. To validate the fidelity of synthetic tumors, we engaged 13 board-certified radiologists in a Visual Turing Test to discern between synthetic and real tumors. Rigorous clinician evaluation validates the high quality of our synthetic tumors, as they achieved only 51.1% sensitivity and 60.8% accuracy in distinguishing our synthetic tumors from real ones. Through high-quality tumor synthesis, FreeTumor scales up the recognition training datasets by over 40 times, showcasing a notable superiority over state-of-the-art AI methods including various synthesis methods and foundation models. These findings indicate promising prospects of FreeTumor in clinical applications, potentially advancing tumor treatments and improving the survival rates of patients.
Julián N. Acosta, Siddhant Dogra, Subathra Adithan et al.
Radiologists face increasing workload pressures amid growing imaging volumes, creating risks of burnout and delayed reporting times. While artificial intelligence (AI) based automated radiology report generation shows promise for reporting workflow optimization, evidence of its real-world impact on clinical accuracy and efficiency remains limited. This study evaluated the effect of draft reports on radiology reporting workflows by conducting a three reader multi-case study comparing standard versus AI-assisted reporting workflows. In both workflows, radiologists reviewed the cases and modified either a standard template (standard workflow) or an AI-generated draft report (AI-assisted workflow) to create the final report. For controlled evaluation, we used GPT-4 to generate simulated AI drafts and deliberately introduced 1-3 errors in half the cases to mimic real AI system performance. The AI-assisted workflow significantly reduced average reporting time from 573 to 435 seconds (p=0.003), without a statistically significant difference in clinically significant errors between workflows. These findings suggest that AI-generated drafts can meaningfully accelerate radiology reporting while maintaining diagnostic accuracy, offering a practical solution to address mounting workload challenges in clinical practice.
Sameer Khanna, Daniel Michael, Marinka Zitnik et al.
Medical image interpretation using deep learning has shown promise but often requires extensive expert-annotated datasets. To reduce this annotation burden, we develop an Image-Graph Contrastive Learning framework that pairs chest X-rays with structured report knowledge graphs automatically extracted from radiology notes. Our approach uniquely encodes the disconnected graph components via a relational graph convolution network and transformer attention. In experiments on the CheXpert dataset, this novel graph encoding strategy enabled the framework to outperform existing methods that use image-text contrastive learning in 1% linear evaluation and few-shot settings, while achieving comparable performance to radiologists. By exploiting unlabeled paired images and text, our framework demonstrates the potential of structured clinical insights to enhance contrastive learning for medical images. This work points toward reducing demands on medical experts for annotations, improving diagnostic precision, and advancing patient care through robust medical image understanding.
The integration of AI-assisted biomedical image analysis into clinical practice demands AI-generated findings that are not only accurate but also interpretable to clinicians. However, existing biomedical AI models generally lack the ability to simultaneously generate diagnostic findings and localize corresponding biomedical objects. This limitation makes it challenging for clinicians to correlate AI-generated findings with visual evidence (e.g., tiny lesions) in images and interpret the results of AI models. To address this challenge, we introduce UniBiomed, the first universal foundation model for grounded biomedical image interpretation, which is capable of generating accurate diagnostic findings and simultaneously segmenting the corresponding biomedical targets. UniBiomed is based on a novel integration of Multi-modal Large Language Model and Segment Anything Model, which can effectively unify diverse biomedical tasks in universal training for advancing grounded interpretation. To develop UniBiomed, we curate a large-scale dataset comprising over 27 million triplets of images, region annotations, and text descriptions across ten biomedical imaging modalities. Extensive validation on 70 internal and 14 external datasets demonstrated the state-of-the-art performance of UniBiomed in diverse biomedical tasks, including image segmentation, disease recognition, region-aware diagnosis, vision question answering, and report generation. In summary, UniBiomed is a powerful and versatile biomedical foundation model, unlocking the untapped grounded interpretation capability for optimizing AI-assisted biomedical image analysis.
Mohammed Baharoon, Siavash Raissi, John S. Jun et al.
We introduce RadGame, an AI-powered gamified platform for radiology education that targets two core skills: localizing findings and generating reports. Traditional radiology training is based on passive exposure to cases or active practice with real-time input from supervising radiologists, limiting opportunities for immediate and scalable feedback. RadGame addresses this gap by combining gamification with large-scale public datasets and automated, AI-driven feedback that provides clear, structured guidance to human learners. In RadGame Localize, players draw bounding boxes around abnormalities, which are automatically compared to radiologist-drawn annotations from public datasets, and visual explanations are generated by vision-language models for user missed findings. In RadGame Report, players compose findings given a chest X-ray, patient age and indication, and receive structured AI feedback based on radiology report generation metrics, highlighting errors and omissions compared to a radiologist's written ground truth report from public datasets, producing a final performance and style score. In a prospective evaluation, participants using RadGame achieved a 68% improvement in localization accuracy compared to 17% with traditional passive methods and a 31% improvement in report-writing accuracy compared to 4% with traditional methods after seeing the same cases. RadGame highlights the potential of AI-driven gamification to deliver scalable, feedback-rich radiology training and reimagines the application of medical AI resources in education.
Pranav Rajpurkar, Julian N. Acosta, Siddhant Dogra et al.
We present a comprehensive evaluation of a2z-1, an artificial intelligence (AI) model designed to analyze abdomen-pelvis CT scans for 21 time-sensitive and actionable findings. Our study focuses on rigorous assessment of the model's performance and generalizability. Large-scale retrospective analysis demonstrates an average AUC of 0.931 across 21 conditions. External validation across two distinct health systems confirms consistent performance (AUC 0.923), establishing generalizability to different evaluation scenarios, with notable performance in critical findings such as small bowel obstruction (AUC 0.958) and acute pancreatitis (AUC 0.961). Subgroup analysis shows consistent accuracy across patient sex, age groups, and varied imaging protocols, including different slice thicknesses and contrast administration types. Comparison of high-confidence model outputs to radiologist reports reveals instances where a2z-1 identified overlooked findings, suggesting potential for quality assurance applications.
Three-dimensional (3D) scene understanding in colonoscopy presents significant challenges that necessitate automated methods for accurate depth estimation. However, existing depth estimation models for endoscopy struggle with temporal consistency across video sequences, limiting their applicability for 3D reconstruction. We present ColonCrafter, a diffusion-based depth estimation model that generates temporally consistent depth maps from monocular colonoscopy videos. Our approach learns robust geometric priors from synthetic colonoscopy sequences to generate temporally consistent depth maps. We also introduce a style transfer technique that preserves geometric structure while adapting real clinical videos to match our synthetic training domain. ColonCrafter achieves state-of-the-art zero-shot performance on the C3VD dataset, outperforming both general-purpose and endoscopy-specific approaches. Although full trajectory 3D reconstruction remains a challenge, we demonstrate clinically relevant applications of ColonCrafter, including 3D point cloud generation and surface coverage assessment.
Arnold Caleb Asiimwe, Dídac Surís, Pranav Rajpurkar et al.
In medical reporting, the accuracy of radiological reports, whether generated by humans or machine learning algorithms, is critical. We tackle a new task in this paper: image-conditioned autocorrection of inaccuracies within these reports. Using the MIMIC-CXR dataset, we first intentionally introduce a diverse range of errors into reports. Subsequently, we propose a two-stage framework capable of pinpointing these errors and then making corrections, simulating an \textit{autocorrection} process. This method aims to address the shortcomings of existing automated medical reporting systems, like factual errors and incorrect conclusions, enhancing report reliability in vital healthcare applications. Importantly, our approach could serve as a guardrail, ensuring the accuracy and trustworthiness of automated report generation. Experiments on established datasets and state of the art report generation models validate this method's potential in correcting medical reporting errors.
Oishi Banerjee, Hong-Yu Zhou, Subathra Adithan et al.
Recent advances in generative vision-language models (VLMs) have exciting potential implications for AI in radiology, yet VLMs are also known to produce hallucinations, nonsensical text, and other unwanted behaviors that can waste clinicians' time and cause patient harm. Drawing on recent work on direct preference optimization (DPO), we propose a simple method for modifying the behavior of pretrained VLMs performing radiology report generation by suppressing unwanted types of generations. We apply our method to the prevention of hallucinations of prior exams, addressing a long-established problem behavior in models performing chest X-ray report generation. Across our experiments, we find that DPO fine-tuning achieves a 3.2-4.8x reduction in lines hallucinating prior exams while maintaining model performance on clinical accuracy metrics. Our work is, to the best of our knowledge, the first work to apply DPO to medical VLMs, providing a data- and compute- efficient way to suppress problem behaviors while maintaining overall clinical accuracy.
Jon Braatz, Pranav Rajpurkar, Stephanie Zhang et al.
In recent years, deep learning has successfully been applied to automate a wide variety of tasks in diagnostic histopathology. However, fast and reliable localization of small-scale regions-of-interest (ROI) has remained a key challenge, as discriminative morphologic features often occupy only a small fraction of a gigapixel-scale whole-slide image (WSI). In this paper, we propose a sparse WSI analysis method for the rapid identification of high-power ROI for WSI-level classification. We develop an evaluation framework inspired by the early classification literature, in order to quantify the tradeoff between diagnostic performance and inference time for sparse analytic approaches. We test our method on a common but time-consuming task in pathology - that of diagnosing gastric intestinal metaplasia (GIM) on hematoxylin and eosin (H&E)-stained slides from endoscopic biopsy specimens. GIM is a well-known precursor lesion along the pathway to development of gastric cancer. We performed a thorough evaluation of the performance and inference time of our approach on a test set of GIM-positive and GIM-negative WSI, finding that our method successfully detects GIM in all positive WSI, with a WSI-level classification area under the receiver operating characteristic curve (AUC) of 0.98 and an average precision (AP) of 0.95. Furthermore, we show that our method can attain these metrics in under one minute on a standard CPU. Our results are applicable toward the goal of developing neural networks that can easily be deployed in clinical settings to support pathologists in quickly localizing and diagnosing small-scale morphologic features in WSI.
Cécile Logé, Emily Ross, David Yaw Amoah Dadey et al.
Recent advances in Natural Language Processing (NLP), and specifically automated Question Answering (QA) systems, have demonstrated both impressive linguistic fluency and a pernicious tendency to reflect social biases. In this study, we introduce Q-Pain, a dataset for assessing bias in medical QA in the context of pain management, one of the most challenging forms of clinical decision-making. Along with the dataset, we propose a new, rigorous framework, including a sample experimental design, to measure the potential biases present when making treatment decisions. We demonstrate its use by assessing two reference Question-Answering systems, GPT-2 and GPT-3, and find statistically significant differences in treatment between intersectional race-gender subgroups, thus reaffirming the risks posed by AI in medical settings, and the need for datasets like ours to ensure safety before medical AI applications are deployed.
Saahil Jain, Ashwin Agrawal, Adriel Saporta et al.
Extracting structured clinical information from free-text radiology reports can enable the use of radiology report information for a variety of critical healthcare applications. In our work, we present RadGraph, a dataset of entities and relations in full-text chest X-ray radiology reports based on a novel information extraction schema we designed to structure radiology reports. We release a development dataset, which contains board-certified radiologist annotations for 500 radiology reports from the MIMIC-CXR dataset (14,579 entities and 10,889 relations), and a test dataset, which contains two independent sets of board-certified radiologist annotations for 100 radiology reports split equally across the MIMIC-CXR and CheXpert datasets. Using these datasets, we train and test a deep learning model, RadGraph Benchmark, that achieves a micro F1 of 0.82 and 0.73 on relation extraction on the MIMIC-CXR and CheXpert test sets respectively. Additionally, we release an inference dataset, which contains annotations automatically generated by RadGraph Benchmark across 220,763 MIMIC-CXR reports (around 6 million entities and 4 million relations) and 500 CheXpert reports (13,783 entities and 9,908 relations) with mappings to associated chest radiographs. Our freely available dataset can facilitate a wide range of research in medical natural language processing, as well as computer vision and multi-modal learning when linked to chest radiographs.
Christian Garbin, Pranav Rajpurkar, Jeremy Irvin et al.
Billions of X-ray images are taken worldwide each year. Machine learning, and deep learning in particular, has shown potential to help radiologists triage and diagnose images. However, deep learning requires large datasets with reliable labels. The CheXpert dataset was created with the participation of board-certified radiologists, resulting in the strong ground truth needed to train deep learning networks. Following the structured format of Datasheets for Datasets, this paper expands on the original CheXpert paper and other sources to show the critical role played by radiologists in the creation of reliable labels and to describe the different aspects of the dataset composition in detail. Such structured documentation intends to increase the awareness in the machine learning and medical communities of the strengths, applications, and evolution of CheXpert, thereby advancing the field of medical image analysis. Another objective of this paper is to put forward this dataset datasheet as an example to the community of how to create detailed and structured descriptions of datasets. We believe that clearly documenting the creation process, the contents, and applications of datasets accelerates the creation of useful and reliable models.
Bryan Gopal, Ryan W. Han, Gautham Raghupathi et al.
We propose 3KG, a physiologically-inspired contrastive learning approach that generates views using 3D augmentations of the 12-lead electrocardiogram. We evaluate representation quality by fine-tuning a linear layer for the downstream task of 23-class diagnosis on the PhysioNet 2020 challenge training data and find that 3KG achieves a $9.1\%$ increase in mean AUC over the best self-supervised baseline when trained on $1\%$ of labeled data. Our empirical analysis shows that combining spatial and temporal augmentations produces the strongest representations. In addition, we investigate the effect of this physiologically-inspired pretraining on downstream performance on different disease subgroups and find that 3KG makes the greatest gains for conduction and rhythm abnormalities. Our method allows for flexibility in incorporating other self-supervised strategies and highlights the potential for similar modality-specific augmentations for other biomedical signals.
Although deep learning models for chest X-ray interpretation are commonly trained on labels generated by automatic radiology report labelers, the impact of improvements in report labeling on the performance of chest X-ray classification models has not been systematically investigated. We first compare the CheXpert, CheXbert, and VisualCheXbert labelers on the task of extracting accurate chest X-ray image labels from radiology reports, reporting that the VisualCheXbert labeler outperforms the CheXpert and CheXbert labelers. Next, after training image classification models using labels generated from the different radiology report labelers on one of the largest datasets of chest X-rays, we show that an image classification model trained on labels from the VisualCheXbert labeler outperforms image classification models trained on labels from the CheXpert and CheXbert labelers. Our work suggests that recent improvements in radiology report labeling can translate to the development of higher performing chest X-ray classification models.
We propose a selective learning method using meta-learning and deep reinforcement learning for medical image interpretation in the setting of limited labeling resources. Our method, MedSelect, consists of a trainable deep learning selector that uses image embeddings obtained from contrastive pretraining for determining which images to label, and a non-parametric selector that uses cosine similarity to classify unseen images. We demonstrate that MedSelect learns an effective selection strategy outperforming baseline selection strategies across seen and unseen medical conditions for chest X-ray interpretation. We also perform an analysis of the selections performed by MedSelect comparing the distribution of latent embeddings and clinical features, and find significant differences compared to the strongest performing baseline. We believe that our method may be broadly applicable across medical imaging settings where labels are expensive to acquire.
A major obstacle to the integration of deep learning models for chest x-ray interpretation into clinical settings is the lack of understanding of their failure modes. In this work, we first investigate whether there are patient subgroups that chest x-ray models are likely to misclassify. We find that patient age and the radiographic finding of lung lesion, pneumothorax or support devices are statistically relevant features for predicting misclassification for some chest x-ray models. Second, we develop misclassification predictors on chest x-ray models using their outputs and clinical features. We find that our best performing misclassification identifier achieves an AUROC close to 0.9 for most diseases. Third, employing our misclassification identifiers, we develop a corrective algorithm to selectively flip model predictions that have high likelihood of misclassification at inference time. We observe F1 improvement on the prediction of Consolidation (0.008 [95% CI 0.005, 0.010]) and Edema (0.003, [95% CI 0.001, 0.006]). By carrying out our investigation on ten distinct and high-performing chest x-ray models, we are able to derive insights across model architectures and offer a generalizable framework applicable to other medical imaging tasks.
We systematically evaluate the performance of deep learning models in the presence of diseases not labeled for or present during training. First, we evaluate whether deep learning models trained on a subset of diseases (seen diseases) can detect the presence of any one of a larger set of diseases. We find that models tend to falsely classify diseases outside of the subset (unseen diseases) as "no disease". Second, we evaluate whether models trained on seen diseases can detect seen diseases when co-occurring with diseases outside the subset (unseen diseases). We find that models are still able to detect seen diseases even when co-occurring with unseen diseases. Third, we evaluate whether feature representations learned by models may be used to detect the presence of unseen diseases given a small labeled set of unseen diseases. We find that the penultimate layer of the deep neural network provides useful features for unseen disease detection. Our results can inform the safe clinical deployment of deep learning models trained on a non-exhaustive set of disease classes.
Automatic extraction of medical conditions from free-text radiology reports is critical for supervising computer vision models to interpret medical images. In this work, we show that radiologists labeling reports significantly disagree with radiologists labeling corresponding chest X-ray images, which reduces the quality of report labels as proxies for image labels. We develop and evaluate methods to produce labels from radiology reports that have better agreement with radiologists labeling images. Our best performing method, called VisualCheXbert, uses a biomedically-pretrained BERT model to directly map from a radiology report to the image labels, with a supervisory signal determined by a computer vision model trained to detect medical conditions from chest X-ray images. We find that VisualCheXbert outperforms an approach using an existing radiology report labeler by an average F1 score of 0.14 (95% CI 0.12, 0.17). We also find that VisualCheXbert better agrees with radiologists labeling chest X-ray images than do radiologists labeling the corresponding radiology reports by an average F1 score across several medical conditions of between 0.12 (95% CI 0.09, 0.15) and 0.21 (95% CI 0.18, 0.24).
Yen Nhi Truong Vu, Richard Wang, Niranjan Balachandar et al.
Self-supervised contrastive learning between pairs of multiple views of the same image has been shown to successfully leverage unlabeled data to produce meaningful visual representations for both natural and medical images. However, there has been limited work on determining how to select pairs for medical images, where availability of patient metadata can be leveraged to improve representations. In this work, we develop a method to select positive pairs coming from views of possibly different images through the use of patient metadata. We compare strategies for selecting positive pairs for chest X-ray interpretation including requiring them to be from the same patient, imaging study or laterality. We evaluate downstream task performance by fine-tuning the linear layer on 1% of the labeled dataset for pleural effusion classification. Our best performing positive pair selection strategy, which involves using images from the same patient from the same study across all lateralities, achieves a performance increase of 14.4% in mean AUC from the ImageNet pretrained baseline. Our controlled experiments show that the keys to improving downstream performance on disease classification are (1) using patient metadata to appropriately create positive pairs from different images with the same underlying pathologies, and (2) maximizing the number of different images used in query pairing. In addition, we explore leveraging patient metadata to select hard negative pairs for contrastive learning, but do not find improvement over baselines that do not use metadata. Our method is broadly applicable to medical image interpretation and allows flexibility for incorporating medical insights in choosing pairs for contrastive learning.
Medical image segmentation models are typically supervised by expert annotations at the pixel-level, which can be expensive to acquire. In this work, we propose a method that combines the high quality of pixel-level expert annotations with the scale of coarse DNN-generated saliency maps for training multi-label semantic segmentation models. We demonstrate the application of our semi-supervised method, which we call CheXseg, on multi-label chest X-ray interpretation. We find that CheXseg improves upon the performance (mIoU) of fully-supervised methods that use only pixel-level expert annotations by 9.7% and weakly-supervised methods that use only DNN-generated saliency maps by 73.1%. Our best method is able to match radiologist agreement on three out of ten pathologies and reduces the overall performance gap by 57.2% as compared to weakly-supervised methods.
Pranav Rajpurkar, Anirudh Joshi, Anuj Pareek et al.
Recent advances in training deep learning models have demonstrated the potential to provide accurate chest X-ray interpretation and increase access to radiology expertise. However, poor generalization due to data distribution shifts in clinical settings is a key barrier to implementation. In this study, we measured the diagnostic performance for 8 different chest X-ray models when applied to (1) smartphone photos of chest X-rays and (2) external datasets without any finetuning. All models were developed by different groups and submitted to the CheXpert challenge, and re-applied to test datasets without further tuning. We found that (1) on photos of chest X-rays, all 8 models experienced a statistically significant drop in task performance, but only 3 performed significantly worse than radiologists on average, and (2) on the external set, none of the models performed statistically significantly worse than radiologists, and five models performed statistically significantly better than radiologists. Our results demonstrate that some chest X-ray models, under clinically relevant distribution shifts, were comparable to radiologists while other models were not. Future work should investigate aspects of model training procedures and dataset collection that influence generalization in the presence of data distribution shifts.
Alexander Ke, William Ellsworth, Oishi Banerjee et al.
Deep learning methods for chest X-ray interpretation typically rely on pretrained models developed for ImageNet. This paradigm assumes that better ImageNet architectures perform better on chest X-ray tasks and that ImageNet-pretrained weights provide a performance boost over random initialization. In this work, we compare the transfer performance and parameter efficiency of 16 popular convolutional architectures on a large chest X-ray dataset (CheXpert) to investigate these assumptions. First, we find no relationship between ImageNet performance and CheXpert performance for both models without pretraining and models with pretraining. Second, we find that, for models without pretraining, the choice of model family influences performance more than size within a family for medical imaging tasks. Third, we observe that ImageNet pretraining yields a statistically significant boost in performance across architectures, with a higher boost for smaller architectures. Fourth, we examine whether ImageNet architectures are unnecessarily large for CheXpert by truncating final blocks from pretrained models, and find that we can make models 3.25x more parameter-efficient on average without a statistically significant drop in performance. Our work contributes new experimental evidence about the relation of ImageNet to chest x-ray interpretation performance.
Pranav Rajpurkar, Anirudh Joshi, Anuj Pareek et al.
The use of smartphones to take photographs of chest x-rays represents an appealing solution for scaled deployment of deep learning models for chest x-ray interpretation. However, the performance of chest x-ray algorithms on photos of chest x-rays has not been thoroughly investigated. In this study, we measured the diagnostic performance for 8 different chest x-ray models when applied to photos of chest x-rays. All models were developed by different groups and submitted to the CheXpert challenge, and re-applied to smartphone photos of x-rays in the CheXphoto dataset without further tuning. We found that several models had a drop in performance when applied to photos of chest x-rays, but even with this drop, some models still performed comparably to radiologists. Further investigation could be directed towards understanding how different model training procedures may affect model generalization to photos of chest x-rays.
Viswesh Krishna, Anirudh Joshi, Philip L. Bulterys et al.
The application of deep learning to pathology assumes the existence of digital whole slide images of pathology slides. However, slide digitization is bottlenecked by the high cost of precise motor stages in slide scanners that are needed for position information used for slide stitching. We propose GloFlow, a two-stage method for creating a whole slide image using optical flow-based image registration with global alignment using a computationally tractable graph-pruning approach. In the first stage, we train an optical flow predictor to predict pairwise translations between successive video frames to approximate a stitch. In the second stage, this approximate stitch is used to create a neighborhood graph to produce a corrected stitch. On a simulated dataset of video scans of WSIs, we find that our method outperforms known approaches to slide-stitching, and stitches WSIs resembling those produced by slide scanners.