Where does AI belong in clinical workflows? A randomized trial from CDH Affiliate Faculty Jonathan Chen, Kevin Schulman, & co-authors, found that both 1st & 2nd-opinion AI improved clinicians’ accuracy, but who goes first can lead to different errors. www.nature.com/articles/s41...

We're excited to introduce the 2026 Stanford Center for Digital Health Grant Awardees, a cohort of scientists working on projects that can transform health. Join us in celebrating this exceptional group of researchers from across @stanforduniversity.bsky.social cdh.stanford.edu/research-por...

A generalizable deep learning system for cardiac MRI - Nature Biomedical Engineering A transformer-based vision system for cardiac MRI offers a generalizable and data-efficient system for diverse tasks in cardiology, offering clear advantages for contextualizing human cardiovascular d...

A study led by Rohan Shad & Will Hiesinger in collaboration with CDH's @euanashley.bsky.social & Curtis Langlotz & others introduces a foundational AI system for cardiac MRI, achieving clinical-grade diagnostic accuracy with a fraction of the data typically required. www.nature.com/articles/s41...

Designed to take on the vast landscape of modern biological data, CellVoyager is an AI agent that autonomously runs complex scRNA-seq analyses at scale.

Learn more about the work from CDH-funded researcher James Zou and team: www.nature.com/articles/s41...

What if all it took to start making better choices was one little nudge? A recent study found that digital “Microsteps" can shift expectations toward healthier behaviors for adults on GLP-1 RAs.

Learn more: jamanetwork.com/journals/jam... @jama.com

A population-based, regression discontinuity analysis examined the effects of nationwide alerting for acute kidney injury on health care and patient outcomes Previous randomized trials and real-world observational studies of electronic alerts for acute kidney injury (AKI) have yielded conflicting results. The applicability of trial findings to routine clin...

Hospitals often use e-alerts to flag acute kidney injury (AKI), but do they actually work? In a study of 3.1 million adults in Wales, CDH-funded researcher Pascal Geldsetzer & colleagues found no measurable impact on patient outcomes.

Learn more: www.kidney-international.org/article/S008...

With its ability to identify and flag potentially harmful gene variants, Evo 2 has clear clinical implications in the field of biological complexity — which is why CDH-funded researcher Brian Hie and his team have made it fully open source.

Learn more: www.nature.com/articles/s41...

Walk With Me: Fatima Rodriguez, cardiologist studying heart disease prevention YouTube video by Stanford Medicine

Walk with Me: Fatima Rodriguez studies heart disease prevention and population health to improve outcomes by reaching patients earlier. Hear what inspired her path—and how she starts each day. www.youtube.com/watch?v=0U_X...

@stanforddeptmed.bsky.social @stanfordcdh.bsky.social

Rapid directed evolution guided by protein language models and epistatic interactions Protein engineering is limited by the inefficient search through a high-dimensional sequence space to find combinations of synergistic mutations. Traditional approaches use stepwise mutation stacking,...

CDH-funded researcher @brianhie.bsky.social & co-authors recently introduced MULTI-evolve, a rapid evolution framework that has the potential to advance protein design, a big step forward for biological research, biotechnology, & therapeutic development.

Learn more: www.science.org/doi/10.1126/...

(2/2) CDH-funded researcher Akshay Chaudhari, Affiliated Faculty members Curtis Langlotz and Nigam Shah, and co-authors developed Merlin, a 3D model that outperforms current radiology AI systems. Learn more in @nature.com : www.nature.com/articles/s41...

(1/2) As the demand for abdominal CT scans increases and radiologists remain in short supply, the need for automated image analysis is growing.

Genome modelling and design across all domains of life with Evo 2 - Nature Evo 2 is an artificial intelligence-based biological foundation model trained on 9 trillion DNA base pairs spanning all domains of life that predicts functional properties from genomic sequences and p...

(2/2) Evo 2 is fully open source thanks to CDH-funded researcher Brian Hie and his team.

Learn more: www.nature.com/articles/s41...

(1/2) Evo 2 is a new AI model that could transform how we predict and prevent disease. Trained on 9 trillion DNA base pairs, it flags harmful genetic variants and generates realistic DNA sequences.

Learn more: www.nature.com/articles/s41...

Could habits on GLP-1 medications improve with a small digital nudge? CDH leadership & team found participants responded positively to simple digital reminders inspiring researchers to hone in on more microstep health tools. med.stanford.edu/news/insight...

@stanfordmedicine.bsky.social

What if a single night of sleep could flag your future health risks? SleepFM, developed by CDH-funded researcher James Zou & colleagues, was trained on 600K+ hours of sleep data from 65K+ individuals using AI to forecast risk for 100+ diseases. sciencedaily.com/releases/2026/01/260109023114.htm

Digital health tools in pathology have advanced rapidly, yet many have not been integrated effectively. SpEMO is a new computational framework that integrates multiple modalities creating a powerful tool for discovery & clinical applications. Learn more: www.nature.com/articles/s41...

Genome modelling and design across all domains of life with Evo 2 - Nature Evo 2 is an artificial intelligence-based biological foundation model trained on 9 trillion DNA base pairs spanning all domains of life that predicts functional properties from genomic sequences and p...

(2/2) Evo 2 is fully open source, thanks to CDH-funded researcher Brian Hie and the rest of the team. Learn more in the full publication: www.nature.com/articles/s41...

(1/2) Evo 2 is a new AI model trained on 9 trillion DNA base pairs that can predict the impact of disease-causing gene variants and generate realistic DNA sequences across a wide range of organisms.

We’re still gathering voices for the AI Healthcare Worker Perspectives survey, and we’d love to hear yours. Take the ten-minute survey here: stanforduniversity.qualtrics.com/jfe/form/SV_...

Evaluating Large Language Model–Generated Clinical Summaries Through a Dual-Perspective Framework: Retrospective Observational Study Large language models (LLMs) are increasingly used by patients and families to interpret complex medical documentation, yet most evaluations focus only on clinician-judged accuracy. In this study, 50 ...

In this study, 50 sets of notes interpreted by GPT-4o mini were rated by both clinicians & parents based on helpfulness, accuracy, & readability. Learn more: ai.jmir.org/2026/1/e85221

(1/2) Most research on language models (LLMs) in digital health asks whether they meet clinical standards of accuracy. But what about patients and families using this technology to understand their own care?

The AI Healthcare Worker Perspectives survey is still live, and if you work in healthcare, we’d love to hear from you. Take the ten-minute survey here: stanforduniversity.qualtrics.com/jfe/form/SV_...

We are pleased to share that the @stanfordmedicine.bsky.social Canary Center will host their Symposium on April 28th. The program will focus on minimally invasive strategies to detect diseases, including cancer, at earlier, more curable stages. Details here: canarycenter.stanford.edu/canary-cente...

Can animation help democratize global health messages? A new paper from CDH Associate Director Maya Adam & co-author Till Bärnighausen explores how animated storytelling could help spread the right messages about health around the world. Learn more: rdcu.be/e4sGn

Ambient AI scribes are already easing clinician burden & burnout, but still remain expensive for healthcare systems. How we define & measure ROI will shape not only adoption today, but also how health systems adapt to the broader AI transformation in care. Learn more: jamanetwork.com/journals/jam...

(2/2) Stanford CDH Director, Eleni Linos, spoke on a panel at the World Governments Summit alongside Daniel Getts of CREATE Medicines about how AI is helping improve access to healthcare for everyone regardless of where they live, and shift the focus of medicine from treatment to prevention.

(1/2) Wellbeing has traditionally been designed to treat illness after it appears. Today however, technology is shifting the focus toward prevention, optimization, and continuous improvement of human performance.

Holistic evaluation of large language models for medical tasks with MedHELM - Nature Medicine MedHELM, an extensible evaluation framework including a new taxonomy for classifying medical tasks and a benchmark of many datasets across these categories, enables the evaluation of large language mo...

MedHELM, developed by Stanford CDH-funded and affiliated researchers in partnership with Microsoft, evaluates LLMs on their ability to handle real clinical tasks. Learn more: www.nature.com/articles/s41... (2/2)

While large language models (LLMs) can achieve near-perfect scores on medical licensing exams, these tests do not capture the complexity of real-world clinical care. (1/2)

How is AI changing your work in healthcare? If you work in this space, we'd love 10 minutes of your time to share your experience in our study: stanforduniversity.qualtrics.com/jfe/form/SV_...