Cardiac hemodynamics play a crucial role in diagnosing and managing heart failure. Current invasive methods are resource-intensive and pose risks, while non-invasive alternatives often lack reliability. We recently demonstrated that AI-driven analysis of cardiac MRI images could accurately predict LVEDP, offering a non-invasive alternative to catheter-based measurements (Lehmann et al., 2024). By integrating cardiac MRI and genetic data from the UK Biobank, this project seeks to advance the understanding of hemodynamic phenotypes and develop AI-driven diagnostic tools. This project aims to leverage the UK Biobank’s unique dataset with paired cardiac MRI and genotype data to uncover genetic determinants of cardiac hemodynamic parameters and enhance diagnostic prediction. The research focuses on:
– What genetic variants are associated with hemodynamic parameters such as LVEDP, PAWP and SVR?
– What insights can be gained about the interplay between phenotypic and genotypic determinants of cardiac function?
To solve these questions we aim to conduct GWA studies to identify genetic variants linked to hemodynamic traits using UK Biobank’s genotype data. We will begin by applying our AI-LVEDP model, previously developed and validated, to predict LVEDP using cardiac MRI data from the UK Biobank. Following the predictions, we will conduct GWAS to identify genetic variants associated with AI-predicted LVEDP as well as hemodynamic parameters currently in the pipeline. This will enable us to explore genotype-phenotype associations specifically linked to the predicted hemodynamic variables. The same approach will be extended to other hemodynamic parameters once the models are available and validated. Finally, we aim to investigate the clinical and epidemiological relevance of AI-predicted hemodynamic parameters. Specifically, we will assess their prognostic ability for cardiovascular outcomes (cardiovascular death, all-cause mortality, HF hospitalizations).
