This prognostic study describes the development and validation of machine learning models trained on both structured data and manually abstracted variables from unstructured clinical notes to predict major complications after postmastectomy breast reconstruction.

This randomized clinical trial evaluates the efficacy of computer-aided detection–assisted vs standard colonoscopy in routine practice among patients with a high risk for colorectal cancer.

This randomized clinical trial assesses the efficacy of a conversational artificial intelligence (AI)–based platform for anxiety, depression, posttraumatic stress disorder, well-being, and life satisfaction among university students in Israel reporting psychological distress.

This cohort study evaluates the discrimination and calibration of the Predicting Risk of Cardiovascular Disease Events (PREVENT) equations using 2 cohorts of data collected from electronic health records (EHRs), comparing the discrimination and calibration with derivation and validation cohorts.

This cross-sectional study evaluates the end-to-end clinical reasoning ability of off-the-shelf large language models using standardized clinical vignettes and introduces a multidimensional comprehensive benchmark for clinical-grade artificial intelligence.

This Viewpoint explores the entry of AI into clinical care, the role of the US Food and Drug Administration, and the associated legal, public health, and medical implications.

This cohort studies evaluates whether photography-based and spirometry-based age estimates are associated with overall and 2-year survival among adults aged 60 years and older with non–small cell lung cancer (NSCLC) who received definitive stereotactic body radiotherapy.

This Viewpoint discusses workflow blocking, constraints limiting how interoperable tools function, and how addressing the gap between the availability of these tools and their integration into clinical practice could improve care quality and clinical experience.

This diagnostic study assesses whether the incorporation of vascular severity, either via artificial intelligence or clinician assessment, is associated with improved short-term prediction of treatment-requiring retinopathy of prematurity and reduced examination burden.

This article discusses the increasing prevalence of individuals seeking mental health help from artificial intelligence (AI) and recommends strategies to help patients navigate AI use.

This US longitudinal cohort study assesses the association of artificial intelligence scribe adoption with changes in electronic health record time expenditure and visit volume and how associations vary by clinician characteristics.

This Viewpoint assesses the consequences of not striving for equipoise between electronic health record developers and third-party AI tools trying to compete with them.

This cross-sectional study evaluates whether large language model scoring of research domain criteria can be successfully applied to law enforcement and coroner or medical examiner death narratives in the US National Violent Death Reporting System.

Nature Machine Intelligence, Published online: 09 April 2026; doi:10.1038/s42256-026-01205-z

Neural networks may be overconfident before they see real data. By briefly training on random noise, models can learn to be uncertain, leading to better calibration, improved identification of out-of-distribution inputs and thus more reliable predictions.

Nature Machine Intelligence, Published online: 09 April 2026; doi:10.1038/s42256-026-01215-x

Cheon and Paik show that overconfidence in deep neural networks arises from standard initialization practices, and that brief warm-up training with random noise improves uncertainty calibration and meta-cognitive recognition of unknown inputs.

Nature Machine Intelligence, Published online: 01 April 2026; doi:10.1038/s42256-026-01210-2

Wang et al. introduce TDFold, which reformulates 3D protein structure prediction as a 2D image-like diffusion task. Its geometric template diffusion framework offers greater accuracy, speed and efficiency than leading models.

Nature Machine Intelligence, Published online: 01 April 2026; doi:10.1038/s42256-026-01206-y

Marwitz et al. demonstrate the use of large language models to build semantic concept graphs from materials science abstracts and train a machine learning model to predict emerging topic combinations from historical data. They show that the model enables experts to find suggestions that can inspire new research.

Nature Machine Intelligence, Published online: 25 March 2026; doi:10.1038/s42256-026-01219-7

A few years ago, we introduced an article format called Reusability Reports to highlight good practices in code sharing and reporting. A renewed focus on reproducibility and transparency in code reporting seems warranted, as research output has accelerated with the widespread adoption of large language models.

Nature Machine Intelligence, Published online: 25 March 2026; doi:10.1038/s42256-026-01195-y

Frank et al. introduce Euclidean fast attention, a linear-scaling framework for 3D data. By leveraging Euclidean rotary encodings, the method overcomes the quadratic cost of standard attention to accurately capture long-range effects in physical systems.

Nature Machine Intelligence, Published online: 25 March 2026; doi:10.1038/s42256-026-01204-0

Muzellec and Kar use reverse predictivity to show that only a subset of artificial neural network (ANN) units align with primate brain responses. This reveals a substantial misalignment between ANNs and brains compared with the strong bidirectional alignment observed between two primate brains.

Nature Machine Intelligence, Published online: 23 March 2026; doi:10.1038/s42256-026-01177-0

We introduce a framework to analyse interpretability in deep learning, by drawing on a formal notion of model semantics from the philosophy of science. We argue that interpretability is only one aspect of a model’s semantics and illustrate our framework with examples from biomedicine.

npj Digital Medicine, Published online: 25 April 2026; doi:10.1038/s41746-026-02661-y

Longitudinal assessment of chorea in Huntington’s disease using digital passive monitoring

npj Digital Medicine, Published online: 24 April 2026; doi:10.1038/s41746-026-02641-2

Angiography-free diagnosis of retinal diseases via interpretable multi-modal learning

npj Digital Medicine, Published online: 24 April 2026; doi:10.1038/s41746-026-02634-1

Exploring the limits of localization: federated model stacking improves hospital-level prediction in a national research network

npj Digital Medicine, Published online: 24 April 2026; doi:10.1038/s41746-026-02648-9

CT-based AI system for quantitative and integrated management of acute respiratory distress syndrome in critical care

npj Digital Medicine, Published online: 24 April 2026; doi:10.1038/s41746-026-02653-y

A review of the application of digital phenotyping in predicting peripartum depressive symptoms

npj Digital Medicine, Published online: 23 April 2026; doi:10.1038/s41746-026-02672-9

Trial emulation for validating the clinical efficacy of a foundational AI model in embryo selection

npj Digital Medicine, Published online: 23 April 2026; doi:10.1038/s41746-026-02640-3

Comprehensive analysis of predictive models for disease manifestations and case fatality in systemic lupus erythematosus

npj Digital Medicine, Published online: 23 April 2026; doi:10.1038/s41746-026-02667-6

Virtual interviews have now become standard practice in health professions training admissions following the COVID-19 pandemic, but the advent of generative AI technologies has raised concerns about the fairness and integrity of such practices. Admissions programs are responding with AI detection software, increased proctoring, and outright bans, all of which are difficult to enforce. A recent randomized controlled trial by Eva and colleagues examining Generative AI tool use among applicants to a medical school during virtual Multiple Mini‑Interviews (MMIs) suggests a different solution: good interview structure may be more resistant to AI advantage, and minor modifications can limit AI use without compromising reliability,authenticity or acceptability. Reframing generative AI as a design problem rather than a detection problem may also help align integrity, equity, and learning values.

This prospective real-world quality improvement study evaluated an artificial intelligence–powered digital front door symptom assessment deployed within Portugal’s largest private health care network. The intervention was associated with reduced patient uncertainty, meaningful shifts in care-seeking behavior toward more appropriate care pathways, and a substantial improvement in the appropriateness of health care utilization.

This study introduces ChexGen, a generative vision-language foundation model that provides a unified framework for text-, mask-, and bounding box–guided synthesis of chest radiographs, and demonstrates its applications in training data augmentation, data-efficient learning, and bias detection and mitigation.

This article introduces EchoNext-Mini, an open dataset of 100,000 electrocardiograms with curated structural heart disease labels and an accompanying convolutional neural network model for detecting structural heart disease from electrocardiogram data. The authors describe the dataset’s development and demonstrate that the EchoNext-Mini model achieves strong performance, providing an open resource to support further research.

This Perspective explores the rapid integration of artificial intelligence into health care, highlighting how safety-net hospitals and underserved patients risk being left behind. It argues that deliberate policies, equitable access, and clinician-led governance are essential to ensure AI strengthens care without deepening existing disparities.

Health systems govern AI by reviewing specific AI tools for defined use cases. But over two thirds of physicians use general-purpose AI daily in their practice, often through consumer platforms outside institutional oversight. Viewing these tools as risky because of the use of generative AI, health systems may disallow their use, paradoxically increasing the cybersecurity risk of patient data entering these platforms without agreed-upon institutional protections.

This Perspective examines the misalignment between health needs, health care resources, and artificial intelligence implementation capacity, demonstrating that Hart’s inverse care law persists in the AI era. Rural areas with the greatest health needs have the lowest AI capacity, risking amplification of existing disparities.

This Policy Corner proposes a structured human factors framework for artificial intelligence–enabled medical devices, distinguishing technical model performance from the safety and effectiveness of human interaction with AI outputs in real-world clinical workflows. It maps key AI-specific cognitive and system integration risks — such as automation bias, trust miscalibration, deskilling, and workflow disruption — to seven regulator-aligned design and evaluation recommendations spanning premarket usability engineering and postmarket monitoring.

CardioNets is a crossmodal deep learning framework that translates standard 12-lead electrocardiography signals into cardiac magnetic resonance–aligned functional parameters and synthetic CMR images, enabling scalable cardiac assessment without direct imaging. Across large, multicohort datasets, the model achieved superior performance to ECG-only baselines and diagnostic accuracy comparable to CMR-based models, supporting its potential to expand access to advanced cardiovascular assessment.

This article introduces MedVersa, a multimodal generalist foundation model trained on tens of millions of medical imaging instances that can accept heterogeneous inputs and generate diverse outputs across imaging workflows. The authors show that MedVersa matches or outperforms task-specific systems on multiple imaging tasks while producing clinically equivalent radiology reports and reducing reporting time and discrepancies in real-world evaluations.

There are tremendous benefits for healthcare companies that strategically leverage large language models, machine learning diagnostics tools, and other Artificial Intelligence capabilities. The competitive advantages that these companies stand to gain are beyond just operational efficiency. With their prediction capabilities and streamlining documentation companies that leverage these AI tools can dramatically improve the quality of care.   However, with every new technology adoption comes a risk and healthcare ...;

In the high-stakes world of drug development, time isn’t just money, it’s measured in human lives. Yet, for decades, the process of finding the right location to test a new life-saving therapy has remained stuck in a pre-digital purgatory. While we can find a five-star hotel in a remote corner of the world in seconds, ...;

The Future of Health Communication   For those of us who are leading marketing in healthcare the conversation around AI isn’t abstract or theoretical, it’s here and how you adopt it will indicate your growth. It’s how you stay competitive, especially when coping with the relentless and complex volumes of content teams are expected to deliver.  Right now,  AI is one of the most useful tools for our ...;

Most healthcare AI is judged as if medicine is a search problem: feed in symptoms, retrieve the right answer, measure accuracy. Real medicine does not work that way. Doctors make a chain of decisions under uncertainty, time pressure, and imperfect signals, and outcomes often depend on the path they take, not just the final label. ...;

Medical science has made monumental leaps forward over the last century, yet women’s healthcare remains plagued by systemic delays and inefficiencies. According to recent data by the Royal College of Obstetricians and Gynaecologists (February 2026), more than 570,000 women in England are currently trapped on gynaecology waiting lists. That covers only those who have already been ...;

With AYURA, he is creating a global care companion that compares options, reduces cost fear, and helps people stay out of dead ends. Desidi Narsimha Reddy founded AYURA with certain realities in mind. Healthcare can break people without ever touching their bodies. It can drain savings, create debt, and trap families in decisions they do ...;

Healthcare payer automation has traditionally centered on control. Enrollment and billing workflows operate through policy checks, audit gates, and compliance rules that protect both regulators and organizations. However, when real-world ambiguity enters the process, that same structure begins to strain. Deterministic engines execute predefined rules efficiently. Enrollment intake, however, rarely arrives in a clean format. ...;

Every surgery generates a flood of information. From measurements to movements, thousands of decisions are captured at every step of the procedure. Roughly 40 million gigabytes of video data is recorded annually through minimally invasive procedures, creating a wealth of real-time insights and feedback. Yet for many years, much of that data vanished as soon as the operation ended. Today, with the help of artificial intelligence, it’s become ...;

The United States spent an estimated $5.6 trillion on healthcare in 2025, more than any other country on earth. That figure dwarfs peer nations on a per-capita basis. Yet Americans have shorter life expectancies, higher rates of preventable disease, and worse chronic condition outcomes than citizens in most comparable countries.   The gap between spending and results ...;

The winners of the MedGemma Impact Challenge demonstrated the potential of Google’s open medical models for solving diverse healthcare challenges.

Google and DocMorris have announced a partnership to create a more intuitive and supportive digital health experience.

<p data-block-key="1tp3e">At Google’s annual health event, The Check Up, we shared how our products, research and partnerships are making the most of AI to help everyone live healthier lives.</p>

At The Check Up, Google announced a $10M investment in clinician AI training and how AI is upgrading Search and Fitbit for better health data.

An overview of how Google Earth AI is supporting the global health community’s work to predict outbreaks and deliver proactive care.

A new Google AI initiative aims to improve heart health outcomes for people living in remote Australian communities.

New research shows how Google AI helps radiologists detect breast cancer earlier and more accurately, while giving radiologists more time for patient care.

Working with Google, Taiwan uses 20 years of health data and Gemini to bring predictive diabetes care to millions in its population-wide health system.

If you were not able to attend this week's The Winning Edge webinar live, you can watch the full video...

Nurse leaders should work towards sustainable succession planning and invest in nurse education, says this CNO.

The latest webinar in HealthLeaders' The Winning Edge series focused on the best practices for managing AI tools in clinical...

Adoption and implementation of AI tools in clinical care requires assessing clinical and financial ROI, garnering clinician support, and effective...

Ambient listening scalability depends on further integration, sustainable collaboration, and focused strategies, say these nurse leaders.

Ambient listening technology helps bolster existing nurse workflows. Here's how.

Get valuable insights from a pair of experts on crucial considerations such as determining the clinical and financial ROI of...

With a $21 billion savings opportunity on the table, revenue cycle leaders must prioritize workflow standardization and vendor partnership ahead...

Mercy is reengineering its revenue cycle strategy to reduce preventable denials, counter increasingly aggressive payer algorithms, and elevate the role...

Presbyterian Healthcare Services offers a model for how health systems can scale AI safely and effectively.

These are the areas seeing the biggest jump in implementation and producing the highest return on investment.