An AI-Based “Do-It-Yourself” Module for Interstitial Glucose Forecasting for People with Type 1 Diabetes

Diabetes mellitus (DM) is a chronic condition defined by increased blood glucose, which is suffered by more than 500 million adults. Type 1 diabetes (T1D), predominantly onset in childhood, needs to be treated with insulin. Keeping glucose within the desired range is a challenge. Despite the advances in the mHealth field, the successful appearance of the so-called do-it-yourself (DIY) tools, and the progress in glucose level prediction based on deep learning (DL), these tools fail to engage the users in the long-term. This limits the potential positive impact that they could have on the daily self-management of the condition, specifically by providing an accurate prediction of their short-term glucose level. Hence, this work proposed a DL-based DIY framework for interstitial glucose prediction using continuous glucose monitoring (CGM) data to generate one fully personalized DL model per user (i.e., without using additional data from other people). The DIY module has been designed to read the CGM raw data (as it would be uploaded by the potential users of this tool), and automatically prepare them to train and validate a DL model for his or her glucose prediction up to one hour ahead. For the DL models training and validation, one year of CGM data from 29 subjects (independently processed) with T1D collected at Complejo Hospitalario Insular-Materno Infantil de Las Palmas de Gran Canaria were used. Results show that this fully personalized approach presents state-of-the-art prediction performance, using only CGM data. Besides, to the best of our knowledge, this work is the first one in providing a DIY approach, since the code of this framework is open source and prepared for the DIY use standalone or in a broader application. Finally, this module has been tested in two different operating systems, demonstrating its easy integration as a standalone tool to help people to effectively aid with T1D self-management.