Dissecting the aetiology of sarcomere-negative hypertrophic cardiomyopathy using convolutional neural network analysis

Building on a previous BHF CRE infrastructure award which created a unique big data digital ECG resource from the international Hypertrophic Cardiomyopathy Registry (HCMR), we have developed an artificial intelligence (AI) test that learns patterns from these heart tracings. The AI-ECG tool is the first to group patients according to an underlying disease mechanism in the common, polygenic form of hypertrophic cardiomyopathy (HCM), where no faulty gene is found and aetiology is unknown. This current project aims to further develop an “explainable” AI model by exploring the feature basis of the model output. If recognisable ECG features can explain subtypes of disease, this could have immediate clinical utility for personalised treatments in a significant number of HCM patients and families.

Image credit: Dr Sunil Manohar