AI for Enhanced Screening and Management of Familial Hypercholesterolemia

Familial hypercholesterolemia (FH) is a genetic condition that predisposes individuals to early onset cardiovascular disease.

Systematic identification of probands remains challenging, leading to underdiagnosis and delayed intervention. Improved access to genetic testing through the proposed Genomic Assessment Centre (GAC) offers great potential. But current strategies (primarily lipid level-based referrals) to refer individuals in the general population for genetic testing have low diagnostic yield.

This project seeks to develop an AI/ML approach to improve the diagnostic yield of current FH screening strategies and inform potential genetic screening decisions.

We will integrate multimodal real-world health records data with genetic testing results for a representative population and their family members and develop a machine learning model to enable more targeted genetic testing for FH. The modelling approach will include FH detection based on multimodal real-world electronic health records (her) and lifestyle data and generate natural language explanations for decision support.

We will design a prototypical decision support system to inform downstream genetic screening and management decisions at point-of-care. We will conduct a validation study within a clinical sandbox environment at the National University Hospital (NUH), with quantitative and qualitative evaluations, to assess performance and relevance for guiding genetic testing referrals.