2:15 p.m. CT Monday, June 15
Any diabetes management system must face the diversity of metabolic situations during a patient’s daily life, behavioral habits, and compliance with the system’s indications for use. Robustness and risk mitigation in relation to scenarios such as abnormal metabolic states, inappropriate habits, omissions, and erroneous interventions by the patient, etc., must be an essential property of such systems. In this talk, several AI-based tools are presented to assess and predict the risks associated with glycemic control. Machine learning is used to predict hypoglycemic events, both after meals and at night. The models obtained are combined with pattern detection and behavior classification tools to monitor the patient’s condition and develop mitigation tools. Finally, the implementation of such systems both in the context of decision support systems and the artificial pancreas is discussed. As a conclusion, therapies combining smart pen with CGM will benefit the most of these tools. The resulting system provides the patient with additional safety in the management of diabetes, improves glycemic control, and reduces the burden of diabetes management. Risk prediction and condition assessment algorithms will also help to reduce manual intervention in artificial pancreas systems, increasing the time in auto mode and, thus, leading to better glycemic control. Josep Vehí, PhD, is Professor of Control and Biomedical Engineering at the University of Girona and associate researcher at the Institute of Biomedical Research of Girona. Since 2018, his Research Group belongs to the center of excellence in diabetes CIBERDEM: Networked Center for Biomedical Research in Diabetes and Associated Metabolic Diseases. His research interests include artificial pancreas, machine learning and its applications to biomedicine, modeling and control of biomedical systems, and artificial intelligence for digital health. Professor Vehi has led and participated in a series of national and European research projects on diabetes technologies since 2004. Main research results include tools and modules for diabetes management and risk assessment and a prototype of artificial pancreas.