Type 2 diabetes (T2D) is a complex metabolic disorder associated with an increased risk of cardiovascular disease (CVD) and neurological complications, including stroke and dementia. These complications share common underlying mechanisms, including insulin resistance and accelerated atherosclerosis. However, the specific pathways linking T2D to these outcomes are not fully understood. Furthermore, differences in genetic, lifestyle, and longitudinal factors between cross-ethnic populations, such as European (i.e., UK Biobank) and South Asian (i.e., Indian) populations, may influence disease trajectories and complication risks.
This study aims to implement an AI-driven, trajectory-based, multimodal framework using data from the UK Biobank (UKB). Multimodal data will include imaging biomarkers (e.g., carotid ultrasound), electronic health records (EHR), polygenic risk scores (PGS), metabolic markers, and other clinical and lifestyle factors. The proposed framework comprises two stages:
1. Prediction of T2D and generation of a T2D risk score.
2. Assessment of the T2D risk score, along with other covariates, as predictors of CVD and neurological outcomes (e.g., stroke and dementia).
To ensure generalizability, the developed multimodal sequential prediction framework will be validated in an independent cohort with comparable data, enabling assessment of cross-ethnic differences in sequential risk profiles. As T2D and Stroke are among the top 10 causes of death in India (reported by WHO), and recent studies have reported an increasing prevalence of dementia in India, plans are underway to initiate collaboration with an Indian or other independent cohort for future validation.
