One of the most serious effects of diabetes is low blood sugar or hypoglycemia. This increases the likelihood of problems requiring complex motor skills and cognitive demands, like driving. Current techniques’ minimal availability, high cost, invasiveness, and diagnostic latency restrict their utility in detecting hypoglycemia.
A recent study published in the journal NEJM AI presents a novel method for detecting hypoglycemia while driving. Scientists at LMU researched in cooperation with colleagues from the University of St. Gallen, ETH Zurich, and the University Hospital of Bern (Inselspital).
In their study, the researchers collected data from 30 diabetics while driving a real car. Data was recorded for each subject twice: once when their blood sugar was normal and again when they were hypoglycemic. To achieve this objective, medical personnel in the automobile purposefully caused each patient to become hypoglycemic. The gathered data comprised driving signals, such as vehicle speed, and head/gaze motion data, such as eye movement speed.
The researchers then created a brand-new machine learning (ML) model that can identify hypoglycemia episodes automatically. And with high accuracy based only on regularly gathered driving and head/gaze motion data.
Simon Schallmoser, a doctoral candidate at the Institute of AI in Management at LMU and one of the contributing researchers, says
This technology could serve as an early warning system in cars and enable drivers to take necessary precautions before hypoglycemic symptoms impair their ability to drive safely.
Crucially for future self-driving cars, the newly created ML model also scored well when using head/gaze motion data. Head of the Institute of AI in Management and project partner, Professor Stefan Feuerriegel says,
“This study not only showcases the potential for AI to improve individual health outcomes but also its role in improving safety on public roads.”
