Development of a neuro-symbolic AI approach to characterize diabetes distress profiles in people with type-1 diabetes

Dulce Canha

PhD student at Luxembourg Institute of Health (LIH)

Abstract: Type-1 diabetes (T1D) is an autoimmune disorder representing 5-10% of global diabetes cases, with a predicted patient growth from 8.4 million in 2021 to potentially 17.4 million by 2040. This chronic disease requires complex daily management, making people with T1D more prone to psychological problems than those without diabetes. Notably, the psychological comorbidity in T1D affects glycemic outcomes, self-management behaviors, and overall quality of life. Given the lack of comprehensive studies on associated psychosocial issues, this research project aims to delve deeper into the determinants and patterns of psychological stress in T1D patients. Key methods include deep digital phenotyping, deep immuno-phenotyping, and the integration of neuro-symbolic AI techniques to unravel patterns in both digital and immuno-phenotyping. The research further employs unsupervised machine learning techniques and aims to establish a correlation between psychological stress and various contributing factors using logic-based modeling. The research utilizes data from the French National Cohort of People with Type 1 Diabetes (SFDT1) study. Our project aims to derive clinically relevant clusters that are interpretable to both clinicians and patients and contribute to the development of personalized health strategies. Expected outcomes include enhanced understanding of T1D’s psychological implications and AI-assisted decision-making processes that prioritize transparency and trust. With AI’s growing importance in the medical field, the emphasis on explainability and trust is paramount.

Keywords: type-1 diabetes, diabetes distress, psychological disorders, digital phenotyping, immuno-phenotyping, neuro-symbolic AI
Scientific area: Healthcare, Artificial Intelligence
Visiting period: 25/03/2024 to 21/09/2024 (first 2 weeks on site)
Visiting Lab: German Research Center for Artificial Intelligence (DFKI)

Bio: My name is Dulce Canha and I am a PhD student in the Deep Digital Phenotyping Research Unit (DDP) at the Department of Precision Health of the Luxembourg Institute of Health (LIH). My background is in Biomedical Engineering, having completed my master in November 2022 at Instituto Superior Técnico, University of Lisbon, Portugal. The topic of my research is deep digital and immune-phenotyping of people with type 1 diabetes (T1D) considering metabolic and psychological stress levels for precision prevention of diabetes-related complications. Mental burden, diabetes distress, fatigue, and depressive symptoms are key risk factors for T1D-related complications, and my project aims to better understand their immunological impact, as little is known about this aspect. My research is based on SFDT1 cohort, which presents a unique opportunity to increase medical knowledge about T1D and potentially suggest new intervention strategies to improve the quality of life of people living with T1D and decrease the burden of the disease. This collaboration will allow me to get trained on neuro-symbolic AI methods and improve the interpretability and readability of my results.