Diabetes Mellitus is the world's most common chronic disease, affecting approximately 415 million people, including approximately 7 million in Germany. For many patients, diabetes means taking daily blood glucose readings, estimating the carbohydrates they consume and injecting insulin. In addition, there is the regular and complete documentation of the values. The GlycoRec system co-developed at L3S helps diabetes patients in their daily management of their disease.
Esysta, a wireless blood glucose meter with insulin pen from project partner Emperra, is already replacing the handwritten diabetes diary. It automatically measures blood sugar and insulin and leads to more complete data. The GlycoRec app also captures physical activity, provides nutritional functions, predicts blood glucose development, and gives advice and warnings. The researchers at L3S have developed a prediction model that is also sufficiently precise for the selective measurement of blood sugar, for example only before meals. For this purpose, the model learns to forecast a prediction reliability depending on the prediction situation.
An important basis is the reliable recording of patient nutrition. In order to determine the carbohydrate content of a dish, the researchers use recipes from the Internet, for example on kochbar.de. Many of the recipes contain information on nutritional values. Uncertainties about the nutritional values of individual recipes, which are predominantly assigned by laypersons, can be largely offset by aggregating many inputs - a principle that is common in crowdsourcing. In order to assess the quality of the recipes provided by users, the scientists have isolated suitable indicators: the adjustment of the assessments of friendship effects already led to a significant improvement in the quality of the data with regard to the reliability of the carbohydrate data of the aggregated "crowd average courts".
Markus Rokicki is a research associate at L3S. His research interests include human computation, game based crowdsourcing motivational mechanisms, and the analysis of user behavior in online food sharing communities.