Read about research on catching early signs of cognitive decline with AI, a tool for evaluating clinical LLM performance, a Senate bill proposing Medicare reimbursement for AI-enabled medical devices, an algorithm to help prescribe nutrition for premature infants, and more.

Adeli’s work is part of a growing field called ambient intelligence, in which sensors are embedded in everyday environments and AI is used to interpret the data. Symptoms that may presage dementia are often invisible to health care providers. Catching them with the help of AI could facilitate early interventions and support.

Read an article about the research.

Using data from electronic health records, the researchers trained an algorithm on nearly 80,000 prescriptions for intravenous nutrition linked to preemie outcomes. The tool has the potential to reduce medical errors, save time and money, and make it easier to care for preemies in low-resource settings, the researchers said.

The next step is to conduct a clinical trial in which some patients are prescribed nutrition regimens based on the traditional approach and others receive AI-recommended prescriptions.

Read the study and an article about it.

MedArena.ai is specifically designed for clinical medicine. The tool is available to clinicians and crowdsources their insights into which large language models are the best for medical queries. Today, the most common approaches for assessing the medical capabilities of LLMs rely on static, multiple-choice question formats. Zou and graduate students Eric Wu and Kevin Wu designed MedArena to be more dynamic, capable of reflecting the most current medical questions and adapting to the iterative nature of clinical questioning.

Read an article about the platform.

The team recorded more than 900 minutes of brain activity from eight mice watching movie clips. ​​Cameras monitored the animals’ eye movements and behavior. The aggregated data was used to train a core AI model, which could be customized into a digital twin of any individual mouse with a bit of additional training. The digital twins were able to closely simulate the neural activity of their biological counterparts in response to a variety of new visual stimuli, including videos and static images.

Such digital twins could make studying the inner workings of the brain easier and more efficient.

Read the study and an article about it.

A Senate bill that would provide a reimbursement pathway for AI-enabled medical devices cleared by the Food and Drug Administration has been introduced by Sens. Mike Rounds, R-S.D., and Martin Heinrich, D-N.M.

Through the Health Tech Investment Act, the senators seek to amend title XVIII of the Social Security Act to “ensure appropriate payment of certain algorithm-based healthcare services under the Medicare program.” Eligible devices would be assigned a new technology ambulatory payment classification for at least five years. “Medicare patients deserve access to the life-changing care that artificial intelligence-enabled devices can offer,” Rounds said in a statement.

Read the bill.

Illustration by Emily Moskal

An AI agent is software that uses artificial intelligence to work toward specific goals on its own — by interpreting information, making decisions, and taking actions in a sequence, without needing step-by-step human instruction. Unlike traditional AI models that are trained for narrow, single tasks (like analyzing images or predicting risk scores), agents are designed to handle more open-ended problems that require planning, adaptation, and coordination across different types of data and tasks.

In medicine, AI agents are still in early development. Researchers are exploring how they might help automate complex processes — such as pulling together patient history, lab results, and current symptoms to suggest personalized treatments. These tools aren’t yet part of routine care, but they point to how AI could evolve from supporting individual tasks to managing broader goals in clinical and research settings.