This editorial examines the use of generative AI to simulate patient encounters for students preparing for the Objective Structured Clinical Examination, highlighting both its pedagogical promise and potential risks. We argue for thoughtful integration that supports diagnostic re… #NEJMAI #Editorial
This study evaluates the AI-Standardized Clinical Examination framework, which leverages text-based simulations with virtual patients and AI-driven assessment. Through a single-blind randomized trial, it assesses the impact of ASCE training on objective structured clinical … #NEJMAI #OriginalArticle
This case study evaluates the ability of six frontier large language models — including GPT-4o, Gemini 1.5 Pro, and Llama 3.1 — to incorporate newly updated medical knowledge through commercial fine-tuning application programming interfaces. Despite modest gains, most models stru… #NEJMAI #CaseStudy
This letter to the editor highlights the need for further research into the cost-effectiveness, real-world impact on patient outcomes, and broader public health value of computer-aided diagnosis tools for tuberculosis. #NEJMAI #Perspective
This editorial examines the use of generative AI to simulate patient encounters for students preparing for the Objective Structured Clinical Examination, highlighting both its pedagogical promise and potential risks. We argue for thoughtful integration that supports diagnostic re… #NEJMAI #Editorial
This study evaluates the AI-Standardized Clinical Examination framework, which leverages text-based simulations with virtual patients and AI-driven assessment. Through a single-blind randomized trial, it assesses the impact of ASCE training on objective structured clinical … #NEJMAI #OriginalArticle
PadChest Grounded Reporting is a novel dataset derived from PadChest and designed to train and evaluate grounded report generation models from chest x-ray images. PadChest-GR includes comprehensive sentence-level bounding-box annotations for all clinically … #NEJMAI #Datasets,Benchmarks,andProtocols
This perspective provides data on the landscape of medical AI in China, covering research focus areas, key players from academia and industry, and the main drivers of medical AI development, as well as insights into these trends relative to those seen in the United States. #NEJMAI #Perspective
A perspective on applying technology to alleviate the current mental health workforce shortage by focusing on the initial stages of mental health treatment for both primary care providers — for whom over 70% of visits involve a mental health component — and mental health profes… #NEJMAI #Perspective
This commentary examines the shortcomings of Canada’s proposed Artificial Intelligence and Data Act, in particular its failure to adequately provide regulatory guidance for health care AI. It argues for a more targeted, sector-specific approach to ensure patient safety, transp… #NEJMAI #PolicyCorner
This perspective explores the capabilities of five large language models (ChatGPT-4o mini, Claude 3.5 Sonnet, Copilot for Microsoft 365, Meta AI Llama 3, and Gemini 1.5 Flash) to respond to ethics scenarios that may emerge when AI is used in health care, and finds that though A… #NEJMAI #Perspective
This review analyzes 950 U.S. Food and Drug Administration–regulated artificial intelligence medical devices, revealing stark contrasts between public and private manufacturers in production scale, transparency, and recall rates. It highlights how commercialization strategies… #NEJMAI #ReviewArticle
The Development, Evaluation, and Assessment of Large Language Models (DEAL) checklist offers two pathways — DEAL-A for advanced model development and DEAL-B for applied research — to ensure comprehensive and consistent reporting of LLM studies, fostering reliable scientific com… #NEJMAI #Perspective
The UC San Diego Health system examines the ethical and practical implications of using generative artificial intelligence (AI) to assist in drafting messages to patients through an approach that prioritizes transparency by disclosing AI involvement in clinical communication an… #NEJMAI #Perspective
This perspective explores 10 strategies for how artificial intelligence can be thoughtfully integrated into HIV programs amid tightening global health resources. It offers actionable guidance to ensure these tools reinforce — rather than disrupt — community priorities, ethical … #NEJMAI #Perspective
This article presents a comprehensive analysis of how artificial intelligence–generated medical responses are perceived and evaluated by nonexperts. The results show that the accuracy of AI-generated responses, on average, was perceived as similar to or even better than the… #NEJMAI #OriginalArticle
This article presents AI-CAC, a deep learning algorithm developed to automatically quantify coronary artery calcium (CAC) from nongated, noncontrast chest computed tomography scans across the U.S. Veterans Affairs health care system. Validated against clinical electrocardio… #NEJMAI #OriginalArticle
This article introduces the Melanoma Research Alliance Multimodal Image Dataset for Artificial Intelligence–Based Skin Cancer (MIDAS), the largest publicly available dataset of biopsy-confirmed skin lesions with paired clinical and dermoscopic images. Using… #NEJMAI #Datasets,Benchmarks,andProtocols
This study evaluates the accuracy of the National Institutes of Health Integrated Data Analysis Platform Text Extraction Program, a Generative Pretrained Transformer 4–powered tool for extracting data from unstructured electronic health records. Compared with a manually curated g… #NEJMAI #CaseStudy
This case study investigates how handwritten labels and watermarks on histopathological images can unintentionally act as prompt injections, misleading state-of-the-art vision–language models such as GPT-4o and Claude. The findings reveal that these models often treat such visual… #NEJMAI #CaseStudy
In this issue of NEJM AI, Azar et al. demonstrate that AI can accurately extract and enter data from prostatectomy pathology reports. This suggests that AI tools may be accurate enough to be deployed for research, minimizing the burden of manual data extraction and entry. #NEJMAI #Editorial
This article examines how the use of artificial intelligence (AI) in radiology influences public perceptions of liability when a radiologist misses a pathology. Participants were more likely to hold the radiologist accountable when AI detected the abnormality, but providing… #NEJMAI #OriginalArticle
This policy commentary explores the transformative role of artificial intelligence (AI) in virtual front doors, which serve as the initial consumer interface in health systems. It highlights how AI is enabling more personalized, equitable, and efficient care triage in Australi… #NEJMAI #PolicyCorner
How should we measure AI in health care? NEJM asks: “Compared with what?”-reminding us to benchmark AI not just against ideals, but against the real-world care patients receive today. The true test: can AI raise the standard we actually have? #AIinHealthcare #NEJMAI
Our perspective published today reflects the input of global AI & HIV experts convened at Banbury Center last year, offering a framework of 10 principles to introduce AI with humility, rigor, and accountability, to keep it honest.
#AI4HIV #AIFrameworks #NEJMAI
ai4hiv.com/10-principle...
Increased state involvement — especially in the absence of strong federal legislation — could either hinder AI development through regulatory fragmentation or encourage self-regulation to meet high standards set by states and international bodies. #NEJMAI #PolicyCorner
This review examines the evolving role of artificial intelligence (AI) in oncology, particularly in radiology and pathology, and treatment selection using large language models. It highlights key challenges — such as data quality, model validation, and clinical integration — … #NEJMAI #ReviewArticle
This study evaluates the CathEF deep-learning algorithm, which estimates left ventricular ejection fraction directly from routine left coronary angiograms in acute coronary syndrome patients, eliminating the need for additional contrast or catheterization. The study found that Ca… #NEJMAI #CaseStudy
Unpacking an important application of model calibration for an out-of-the-box artificial intelligence algorithm used to flag patients who are at high risk for hypertrophic cardiomyopathy, this article explains the use of calibrated scores for patient-level decision-making. #NEJMAI #Editorial
This study finds that an out-of-the-box risk classification by a U.S. Food and Drug Administration–cleared artificial intelligence–enabled algorithm for identifying hypertrophic cardiomyopathy from electrocardiograms can have a low predictive value when used for diagnosis i… #NEJMAI #OriginalArticle
