Matthew P Lungren

Debadutta Dash, Rahul Thapa, Juan M. Banda et al.

Despite growing interest in using large language models (LLMs) in healthcare, current explorations do not assess the real-world utility and safety of LLMs in clinical settings. Our objective was to determine whether two LLMs can serve information needs submitted by physicians as questions to an informatics consultation service in a safe and concordant manner. Sixty six questions from an informatics consult service were submitted to GPT-3.5 and GPT-4 via simple prompts. 12 physicians assessed the LLM responses' possibility of patient harm and concordance with existing reports from an informatics consultation service. Physician assessments were summarized based on majority vote. For no questions did a majority of physicians deem either LLM response as harmful. For GPT-3.5, responses to 8 questions were concordant with the informatics consult report, 20 discordant, and 9 were unable to be assessed. There were 29 responses with no majority on "Agree", "Disagree", and "Unable to assess". For GPT-4, responses to 13 questions were concordant, 15 discordant, and 3 were unable to be assessed. There were 35 responses with no majority. Responses from both LLMs were largely devoid of overt harm, but less than 20% of the responses agreed with an answer from an informatics consultation service, responses contained hallucinated references, and physicians were divided on what constitutes harm. These results suggest that while general purpose LLMs are able to provide safe and credible responses, they often do not meet the specific information need of a given question. A definitive evaluation of the usefulness of LLMs in healthcare settings will likely require additional research on prompt engineering, calibration, and custom-tailoring of general purpose models.

Arjun Soin, Jameson Merkow, Jin Long et al.

Clinical Artificial lntelligence (AI) applications are rapidly expanding worldwide, and have the potential to impact to all areas of medical practice. Medical imaging applications constitute a vast majority of approved clinical AI applications. Though healthcare systems are eager to adopt AI solutions a fundamental question remains: \textit{what happens after the AI model goes into production?} We use the CheXpert and PadChest public datasets to build and test a medical imaging AI drift monitoring workflow to track data and model drift without contemporaneous ground truth. We simulate drift in multiple experiments to compare model performance with our novel multi-modal drift metric, which uses DICOM metadata, image appearance representation from a variational autoencoder (VAE), and model output probabilities as input. Through experimentation, we demonstrate a strong proxy for ground truth performance using unsupervised distributional shifts in relevant metadata, predicted probabilities, and VAE latent representation. Our key contributions include (1) proof-of-concept for medical imaging drift detection that includes the use of VAE and domain specific statistical methods, (2) a multi-modal methodology to measure and unify drift metrics, (3) new insights into the challenges and solutions to observe deployed medical imaging AI, and (4) creation of open-source tools that enable others to easily run their own workflows and scenarios. This work has important implications. It addresses the concerning translation gap found in continuous medical imaging AI model monitoring common in dynamic healthcare environments.

Joseph Paul Cohen, Joseph D. Viviano, Paul Bertin et al.

TorchXRayVision is an open source software library for working with chest X-ray datasets and deep learning models. It provides a common interface and common pre-processing chain for a wide set of publicly available chest X-ray datasets. In addition, a number of classification and representation learning models with different architectures, trained on different data combinations, are available through the library to serve as baselines or feature extractors.

Harsha Nori, Mayank Daswani, Christopher Kelly et al.

Artificial intelligence holds great promise for expanding access to expert medical knowledge and reasoning. However, most evaluations of language models rely on static vignettes and multiple-choice questions that fail to reflect the complexity and nuance of evidence-based medicine in real-world settings. In clinical practice, physicians iteratively formulate and revise diagnostic hypotheses, adapting each subsequent question and test to what they've just learned, and weigh the evolving evidence before committing to a final diagnosis. To emulate this iterative process, we introduce the Sequential Diagnosis Benchmark, which transforms 304 diagnostically challenging New England Journal of Medicine clinicopathological conference (NEJM-CPC) cases into stepwise diagnostic encounters. A physician or AI begins with a short case abstract and must iteratively request additional details from a gatekeeper model that reveals findings only when explicitly queried. Performance is assessed not just by diagnostic accuracy but also by the cost of physician visits and tests performed. We also present the MAI Diagnostic Orchestrator (MAI-DxO), a model-agnostic orchestrator that simulates a panel of physicians, proposes likely differential diagnoses and strategically selects high-value, cost-effective tests. When paired with OpenAI's o3 model, MAI-DxO achieves 80% diagnostic accuracy--four times higher than the 20% average of generalist physicians. MAI-DxO also reduces diagnostic costs by 20% compared to physicians, and 70% compared to off-the-shelf o3. When configured for maximum accuracy, MAI-DxO achieves 85.5% accuracy. These performance gains with MAI-DxO generalize across models from the OpenAI, Gemini, Claude, Grok, DeepSeek, and Llama families. We highlight how AI systems, when guided to think iteratively and act judiciously, can advance diagnostic precision and cost-effectiveness in clinical care.

Yu Gu, Jingjing Fu, Xiaodong Liu et al.

Large frontier models like GPT-5 now achieve top scores on medical benchmarks. But our stress tests tell a different story. Leading systems often guess correctly even when key inputs like images are removed, flip answers under trivial prompt changes, and fabricate convincing yet flawed reasoning. These aren't glitches; they expose how today's benchmarks reward test-taking tricks over medical understanding. We evaluate six flagship models across six widely used benchmarks and find that high leaderboard scores hide brittleness and shortcut learning. Through clinician-guided rubric evaluation, we show that benchmarks vary widely in what they truly measure yet are treated interchangeably, masking failure modes. We caution that medical benchmark scores do not directly reflect real-world readiness. If we want AI to earn trust in healthcare, we must demand more than leaderboard wins and must hold systems accountable for robustness, sound reasoning, and alignment with real medical demands.

Anirudh Joshi, Sabri Eyuboglu, Shih-Cheng Huang et al.

FDG PET/CT imaging is a resource intensive examination critical for managing malignant disease and is particularly important for longitudinal assessment during therapy. Approaches to automate longtudinal analysis present many challenges including lack of available longitudinal datasets, managing complex large multimodal imaging examinations, and need for detailed annotations for traditional supervised machine learning. In this work we develop OncoNet, novel machine learning algorithm that assesses treatment response from a 1,954 pairs of sequential FDG PET/CT exams through weak supervision using the standard uptake values (SUVmax) in associated radiology reports. OncoNet demonstrates an AUROC of 0.86 and 0.84 on internal and external institution test sets respectively for determination of change between scans while also showing strong agreement to clinical scoring systems with a kappa score of 0.8. We also curated a dataset of 1,954 paired FDG PET/CT exams designed for response assessment for the broader machine learning in healthcare research community. Automated assessment of radiographic response from FDG PET/CT with OncoNet could provide clinicians with a valuable tool to rapidly and consistently interpret change over time in longitudinal multi-modal imaging exams.

Imon Banerjee, Ananth Reddy Bhimireddy, John L. Burns et al.

Background: In medical imaging, prior studies have demonstrated disparate AI performance by race, yet there is no known correlation for race on medical imaging that would be obvious to the human expert interpreting the images. Methods: Using private and public datasets we evaluate: A) performance quantification of deep learning models to detect race from medical images, including the ability of these models to generalize to external environments and across multiple imaging modalities, B) assessment of possible confounding anatomic and phenotype population features, such as disease distribution and body habitus as predictors of race, and C) investigation into the underlying mechanism by which AI models can recognize race. Findings: Standard deep learning models can be trained to predict race from medical images with high performance across multiple imaging modalities. Our findings hold under external validation conditions, as well as when models are optimized to perform clinically motivated tasks. We demonstrate this detection is not due to trivial proxies or imaging-related surrogate covariates for race, such as underlying disease distribution. Finally, we show that performance persists over all anatomical regions and frequency spectrum of the images suggesting that mitigation efforts will be challenging and demand further study. Interpretation: We emphasize that model ability to predict self-reported race is itself not the issue of importance. However, our findings that AI can trivially predict self-reported race -- even from corrupted, cropped, and noised medical images -- in a setting where clinical experts cannot, creates an enormous risk for all model deployments in medical imaging: if an AI model secretly used its knowledge of self-reported race to misclassify all Black patients, radiologists would not be able to tell using the same data the model has access to.

Arjun Soin, Pratik Bhatu, Rohit Takhar et al.

Adoption of artificial intelligence medical imaging applications is often impeded by barriers between healthcare systems and algorithm developers given that access to both private patient data and commercial model IP is important to perform pre-deployment evaluation. This work investigates a framework for secure, privacy-preserving and AI-enabled medical imaging inference using CrypTFlow2, a state-of-the-art end-to-end compiler allowing cryptographically secure 2-party Computation (2PC) protocols between the machine learning model vendor and target patient data owner. A common DenseNet-121 chest x-ray diagnosis model was evaluated on multi-institutional chest radiographic imaging datasets both with and without CrypTFlow2 on two test sets spanning seven sites across the US and India, and comprising 1,149 chest x-ray images. We measure comparative AUROC performance between secure and insecure inference in multiple pathology classification tasks, and explore model output distributional shifts and resource constraints introduced by secure model inference. Secure inference with CrypTFlow2 demonstrated no significant difference in AUROC for all diagnoses, and model outputs from secure and insecure inference methods were distributionally equivalent. The use of CrypTFlow2 may allow off-the-shelf secure 2PC between healthcare systems and AI model vendors for medical imaging, without changes in performance, and can facilitate scalable pre-deployment infrastructure for real-world secure model evaluation without exposure to patient data or model IP.