Researchers at Imperial College London have identified six distinct motion phenogroups in the left ventricle that predict cardiovascular outcomes and genetic risk. This breakthrough utilizes four-dimensional motion analysis to detect subtle cardiac abnormalities often missed by conventional global volumetric measurements.
The team applied computer vision artificial intelligence to over 20,000 participant samples from the UK Biobank. Image data was converted into 4D point cloud models to capture full ventricular shape changes throughout the cardiac cycle. These models were clustered by disease prevalence, future cardiac events, and polygenic risk scores to define diverse motion patterns.
Analysis revealed two low-risk groups with minimal obesity or diabetes markers. A fourth cluster correlated strongly with cardiometabolic diseases like diabetic cardiomyopathy. The final two groups exhibited the highest heart disease prevalence and polygenic risk scores. One specific high-risk cluster was uniquely linked to incident myocardial infarction and cardiac arrest, suggesting motion traits can predict fatal outcomes.
While promising, the current model focuses exclusively on the left ventricle and relies on predominantly European ancestry data from older adults. Further validation across diverse populations is required. If confirmed in clinical settings, this technology could redefine personalized screening and management for heart disease.