Pillaries-UNIST Develop 'Virtual CGM' Blood Glucose Prediction AI Model

Predicting blood sugar with everyday data! Philise, in collaboration with UNIST, unveils innovative 'Virtual CGM' model. Philise, an AI-powered health management app, has developed an innovative blood sugar prediction model called 'Virtual CGM (Continuous Glucose Monitoring)' that utilizes everyday life data, in collaboration with Professor Min-hyuk Lim's team at Ulsan National Institute of Science and Technology (UNIST). The research findings were published in the international journal Scientific Reports, attracting significant attention. This model is an artificial intelligence technology that analyzes repetitive lifestyle pattern data, such as meals, sleep, and exercise, to predict blood sugar changes. Its core objective is to provide seamless continuity in blood sugar management, especially when there are gaps in existing continuous glucose monitors (CGMs) or when they are difficult to use. Philise, in collaboration with Professor Min-hyuk Lim's team at UNIST, applied advanced artificial intelligence neural network technology to precisely analyze the impact of time and behavioral factors on blood sugar, thereby implementing the prediction model. Based on data from 171 healthy adults, the research results showed excellent predictive performance, including an RMSE of 19.49 and a MAPE of 12.34%, proving that blood sugar changes can be predicted to a certain level even without a sensor. This advanced technology will be immediately applied to Philise's blood sugar management service, 'SugarCare'. By consistently recording their lifestyle data, such as diet, sleep, and exercise, users can train personalized blood sugar response patterns and receive early notifications about lifestyle habits that may pose a risk of blood sugar elevation. This feature activates when users have sufficiently accumulated CGM data and lifelog data, aiding in proactive health management. However, Philise emphasized that this 'Virtual CGM' is not intended to completely replace existing continuous glucose monitors, but rather serves as a complementary tool that addresses the limitations of sensor use and supports self-directed health management in daily life. Its focus is on preventive use, not medical diagnostic purposes. Amid growing interest in blood sugar management among healthy individuals, the high cost of existing CGMs and the difficulty in interpreting their data have been cited as limitations. This 'Virtual CGM' model lowers these barriers, presenting a new alternative that enables practical and accessible blood sugar monitoring using only lifestyle data. Professor Min-hyuk Lim of UNIST evaluated it as "the possibility of predictable healthcare technology through the analysis of daily patterns," while Shin In-sik, CEO of Philise, emphasized the synergy, stating that "when used in conjunction with CGM, prediction accuracy and user convenience can be maximized." With over 1 million users and accumulating nearly 70 million lifelog data points, Philise is expected to establish itself as a leader in AI-based health prediction, leveraging this technology. The company holds over 12 related patents, demonstrating its technological prowess.