Blood Glucose Level Prediction Challenge

Blood glucose level (BGL) prediction is a challenging task for AI researchers, with the potential to improve the health and wellbeing of people with diabetes. Knowing in advance when blood glucose is approaching unsafe levels provides time to pro-actively avoid hypo- and hyperglycemia and their concomitant complications. The drive to perfect an artificial pancreas has increased the interest in using machine learning (ML) approaches to improve prediction accuracy. Work in this area has been hindered, however, by a lack of real patient data; some researchers are only able to work on simulated patient data.

In the Machine Learning Blood Glucose Level Prediction (BGLP) Challenge, researchers will come together to compare the efficacy of different ML prediction approaches on a standard set of real patient data, the OhioT1DM Dataset. Participants will use the same data with their own diverse prediction approaches, and we will come together as a group to compare results and share experiences. We believe that this will foster continuing collaboration and accelerate the pace of progress in the field.

The OhioT1DM Dataset

The OhioT1DM Dataset contains 8 weeks worth of data for each of 6 people with type 1 diabetes. These people were all on insulin pump therapy with continuous glucose monitoring (CGM). They reported life-event data and provided physiological data from a Basis Peak fitness band. The dataset includes: a CGM blood glucose level every 5 minutes; blood glucose levels from periodic self-monitoring of blood glucose (finger sticks); insulin doses, both bolus and basal; self-reported meal times with carbohydrate estimates; self-reported times of sleep, work, and exercise; and 5-minute aggregations of heart rate, galvanic skin response (GSR), skin temperature, air temperature, and step count.

Post-Challenge Update: Researchers can still obtain the OhioT1DM Dataset now that the Challenge is over. Interested researchers can click here for additional information.

Participate in the Challenge!

To facilitate "apples to apples" comparisons, participants are asked to report results for a 30-minute prediction horizon, as specified in How to Report Results.

Submission requirements are detailed in the Call for Papers. Deadlines are posted with the Important Dates.

BGLP Challenge Organization

The BGLP Challenge Co-Chairs are Razvan Bunescu, Aili Guo, and Cindy Marling, from Ohio University (USA). Contacts are: {bunescu,marling}@ohio.edu.

The BGLP Challenge Program Committee Members are: Ali Cinar, Illinois Institute of Technology (USA); J. Manuel Colmenar, Universidad Rey Juan Carlos (Spain); Alexandra Constantin, Bigfoot Biomedical (USA); Ivan Contreras, University of Girona (Spain); Andrea Facchinetti, University of Padova (Italy); Pau Herrero, Imperial College London (UK); J. Ignacio Hidalgo, Universidad Complutense de Madrid (Spain); David Klonoff, University of California, San Francisco (USA); Alexander Schliep, Gothenburg University (Sweden); Giovanni Sparacino, University of Padova (Italy); Josep Vehi, University of Girona (Spain); and Qian Wang, Pennsylvania State University (USA).

The BGLP Challenge is supported by grant 1R21EB022356 from the National Institutes of Health (NIH).