Deep learning approach to integrate continuous glucose monitoring in cardiovascular risk assessment for people with diabetes
Background: Diabetes Mellitus is a growing public health challenge globally, and is as-sociated with a significant burden of complications, including cardiovascular diseases (CVD). An association is established between glycemic variability and cardiovascular health. However, risk assessment tools to predict CVD risk in people with diabetes do not accommodate the rise in wearable technologies such as continuous glucose moni-toring (CGM) devices. New analytical methods are needed to exploit the full potential of data. Deep learning plays a crucial role in this, but transferring knowledge from epi-demiological cohorts to clinical cohort studies has not been exploited to date.
Aim: This project aims to unravel clinically relevant associations between glycemic control and cardiovascular (CV) risk and to translate this knowledge into risk assess-ment tools for people with diabetes.
Methods: We will investigate current glucose prediction models and transfer learning, i.e. the reuse of models across different tasks, from the perspective of algorithmic fair-ness. Novel patterns in CGM data associated with CV risks will be studied, by repur-posing models and using explainable artificial intelligence (AI) methods. We will ana-lyse longitudinal predictors and to predict CVD incidence in people with type 1 diabe-tes. Furthermore, we will predict blood glucose spikes in order to tailor diet recom-mendations and assist risk assessments of CVD in people with type 2 diabetes.
Perspectives: The proposed project will contribute with valuable insight in how CGM data can be used in CVD risk assessments and show the potential in knowledge trans-fer between studies. The diversity in study populations and the strong interdisciplinary collaborations on international levels (NL, GER, DK) gives a strong foundation for a double PhD-project conducted at Aarhus University and Maastricht University.
Cross-academy scholarship with co-funding from the Danish Cardiovascular Academy
Adam Hulman, Steno Diabetes Center Aarhus
EAN: 5798 0022 30642
Reference: 1025 0006
CVR: 29 19 09 09