Utility of artificial intelligence in diabetes self-management, a narrative literature review

Abstract

This thesis looks at the utility of Artificial Intelligence (AI) for diabetes self-management and in particular, for diet and physical activity management. The research method used is a narrative literature review through which eight studies (n=8) were identified. Each of the study introduces a different AI-powered eHealth system for diet and/or physical activity self-management by diabetics. Search queries were performed on several databases: Google Scholar, Pubmed, ScienceDirect and EBSCOhost / Cinahl complete. The articles selected for the narrative literature review covered three apps, one virtual assistant, one serious game and three robots. The systems target different user groups, answer different use cases and resort to a variety of AI techniques, including machine learning techniques. The utility of AI in these systems is first that it allows to automate low-value tasks such as food nutritive and calorific content calculation and activity monitoring. Second, AI offers the possibility to create new formats of care delivery to reach specific user groups. Third, AI enables the personalization of eHealth tools to individuals for maximum efficacy and adoption. For seven of the systems, the efficacy on diabetes self-management has not been demonstrated yet.