A new artificial intelligence (AI) tool can now detect dangerously low blood sugar levels—also known as hypoglycemia—just by analyzing a person’s driving behavior and head and eye movements.
This breakthrough could one day help drivers with diabetes stay safe on the road, by alerting them to early signs of a drop in blood sugar before symptoms set in.
Hypoglycemia is one of the most serious complications of diabetes. It happens when blood sugar levels fall too low, which can lead to confusion, dizziness, poor coordination, and even fainting or seizures.
These symptoms are especially dangerous during tasks that demand quick thinking and steady motor skills, such as driving.
Until now, detecting hypoglycemia has required methods that are often too slow, invasive, expensive, or not widely available. But researchers from LMU (Ludwig Maximilian University of Munich), working with teams from the University Hospital of Bern, ETH Zurich, and the University of St. Gallen, have found a new way to spot these episodes in real time—without drawing blood or using traditional sensors.
Their study, published in NEJM AI, involved 30 people with diabetes driving a real car in two separate sessions: once with normal blood sugar and once during a hypoglycemic state. To safely bring about a hypoglycemic state, trained medical staff accompanied the drivers and monitored them closely.
During the drives, the researchers collected data on how each person operated the vehicle—like changes in speed—and tracked their head position and eye movement, including how fast their eyes moved.
Using this data, the team built a machine learning model—an AI system trained to detect patterns in behavior. The system was able to reliably tell whether someone was driving with low blood sugar based on their driving style and eye/head movement patterns alone.
For example, slowed or erratic eye movement and inconsistent driving speeds were signs that something was wrong.
Importantly, the AI model still worked well when it was fed only head and gaze data. This is especially useful for the future, where cars may become increasingly autonomous. In those cases, driving data like steering or acceleration may no longer exist—but facial motion data will still be trackable.
“This technology could act as an early warning system for drivers,” said Simon Schallmoser, a doctoral student at LMU and one of the researchers. “It could alert someone before their symptoms make driving dangerous.”
Professor Stefan Feuerriegel, who led the project at LMU’s Institute of AI in Management, emphasized that the findings have broader value. Not only can AI be used to improve the health and safety of individuals with diabetes, but it can also improve overall road safety by helping prevent accidents caused by sudden medical issues.
In summary, this study shows how AI, combined with everyday technology like in-car sensors or cameras, could soon help protect millions of drivers with diabetes. Instead of relying on blood tests or wearable devices, the car itself might one day warn drivers when it’s time to stop and treat low blood sugar—potentially saving lives on the road.
If you care about diabetes, please read studies about a cure for type 2 diabetes, and these vegetables could protect against kidney damage in diabetes.
For more information about diabetes, please see recent studies about bone drug that could lower risk of type 2 diabetes, and results showing eating more eggs linked to higher risk of type 2 diabetes.
The research findings can be found in NEJM AI.
Copyright © 2025 Knowridge Science Report. All rights reserved.
