AI Teaches Surgery; AI Quality Registry; AI Matches Patients With Trials

Welcome to MedAI Roundup, highlighting the latest news and research in healthcare-related artificial intelligence each month.

Can AI help make sense of the world’s vast collection of scientific literature? (Nature)

A team at Johns Hopkins used videos of surgeries and the machine learning architecture behind ChatGPT to train a robot to do surgical tasks. (Axios)

As the FDA begins to develop new frameworks for regulating AI-enabled devices, health systems struggle with how to implement AI tools that are already available. (Axios)

The American College of Radiology announced that it launched an AI quality registry that will monitor the real-world performance of AI-powered imaging algorithms in clinical settings.

A large language model called TrialGPT was good at matching patients with clinical trials in early experiments, according to a study published in Nature Communications.

A case study in NEJM AI found that combining responses from several large language models (LLMs) could improve the accuracy of differential diagnoses compared with responses from one platform alone.

However, bias in AI models can negatively affect clinical outcomes and worsen healthcare disparities, according to a study in PLOS Digital Health.

Mount Sinai Health System in New York City announced that it launched a center for studying AI and human health, which will house about 40 principal investigators and 250 graduate students, postdoc fellows, computer scientists, and support staff.

Abridge announced a new partnership with a research lab at Beth Israel Deaconess Medical Center in Boston to evaluate the company’s AI-generated patient visit summaries.

Nvidia plans to increase the physical presence of AI in health systems by developing more AI-enabled robots and devices. (Business Insider)