Currently there is no decision support system for insulin dosing on the market that adapts itself based on real-time blood glucose data and physical activity. The PEPPER team have addressed this by creating a personalised decision support tool that gathers such data on a mobile platform. The system makes predictions based both on the real-time data and manually-entered information, such as carbohydrate consumption and alcohol intake. When users eat a meal, they consult the system to determine how much insulin they need. The artificial intelligence adapts the dose advice according to the previous history of that individual, which is updated dynamically. Users’ safety is guaranteed through two levels of supervision. A first level consisting of a safety system including glucose alerts/alarm and constraints on insulin delivery, and a second layer consisting of a secure cloud-based server allowing remote supervision by clinicians.
The project team has built a strong associated research community, with highlights including prize-winning publications and joint organisation of three workshops on Artificial Intelligence for Diabetes, each co-located with a major conference, and involving editorship of associated journal special issues.
We are excited about the project results which show that, despite limitations, the data is promising and suggestive that an adaptive bolus advisor, in combination with a safety system, may have potential to improve health outcomes for people with insulin-dependent diabetes. Thus, there is wide scope for integrating PEPPER into routine diabetes management. In addition, the adaptation feature of the PEPPER algorithm also has potential for use within artificial pancreas systems. These findings show that our work has potential benefit to society by improving health outcomes.