Hypoglycemia is a challenging aspect of health management for people with diabetes. Physiological reponses to hypoglycemia can be detected by wearable devices and could be leveraged to noninvasively predict its occurrence. Commercially available consumer-grade wearables such as smartwatches are an attractive, yet underexplored, candidate for such application. To explore the potential of smartwatches in this regard, data from an Apple Watch and a continuous glucose monitoring system (CGMS) were collected from adults with type 1 diabetes, alongside manually documented data. In this repository, one can find:
- the code necessary for data extraction from the Apple Watch and CGMS
- the code required for generating features required for modeling
- the code used for model building
Please note that data from manually-documented diaries, sleeping times and unique codes for days/nights have been incorporated manually in the datasets.