HEADWIND: Design and Evaluation of a Vehicle Hypoglycemia Warning System in Diabetes - a proof of principle study

Background/Introduction: Despite ongoing developments in the treatment of diabetes, hypoglycaemia remains one of the most relevant acute complications associated with this disease. During hypoglycaemia cognitive, executive and psychomotor abilities significantly deteriorate. Accordingly, hypoglycaemia has consistently been shown to be associated with an increased risk of driving accidents and is, therefore, regarded as one of the relevant factors in traffic safety. Today’s cars continuously gather a broad spectrum of real-time information on various driving parameters. This may allow for an alternative approach to the problem of hypoglycaemia during driving. Using artificial intelligence constantly analyzing driving behavior it may be possible to timely detect changes in driving pattern characteristic for driving in hypoglycaemia. Based on these alterations in driving variables we aim at establishing algorithms capable of discriminating eu- and hypoglycemic driving patterns using artificial intelligence. Methods: In a proof of principle study we compared data regrading driving behavior of 5 individuals (3 non- diabetic and 2 with type 1 diabetes) tracking measurements in eu- and hypoglycemic condition while driving on a predefined route using a professional driving simulator (Carnetsoft BV). Over 60 driving parameters were assessed at a sampling rate of 30 Hz. Time series of car-based sensor data was then sliced into 5 minute windows and random forest machine learning classifier as well as deep neural networks were applied to build a system detecting hypoglycemia within 5 minute frames. Results: Car-based data provided 73'970 measurements in hypoglycemic condition (