Developing and investigating computing, machine learning and physiological modelling for understanding each individual heart towards personalised medicine

Deciphering the underlying cause of left ventricular hypertrophy (LVH) is a highly relevant clinical challenge given the need for a highly different therapeutic approach depending on the aetiology. This requires a combination of an extensive clinical as well as familial assessment, including multimodality imaging and genetics. While cardiac magnetic resonance imaging and genetics contribute to the assessment, they are also expensive and time-consuming. Finding approaches that provide more pathophysiological insight based on clinical combined with echocardiographic and electrocardiographic information would provide a way forward for individualised temporal diagnosis and follow-up.

Machine learning (ML), by analysing and integrating the full spectrum of accessible data sources (electrocardiogram and echography), together with clinical parameters, might provide a helpful tool. Currently, most ML research relies on deep learning (DL) techniques, with good results, but hard to interpret clinically. Representation learning (RL) allows to find representative and simplified patterns in the complex data available, by organizing individuals by their similarity in the original data. The aim of the current project is to evaluate and validate the clinical value of RL to the whole-cardiac cycle echocardiographic data, compare this approach to DL classification, and explore the added value of the combinatorial use of clinical, ECG and echocardiographic data through ML to identify the cause of unexplained LVH in daily cardiological practice. his will include a multicentre clinical study.

By exploring the phenotypic presentation of patients with hypertrophic cardiomyopathy, we will create a framework to better categorize patients, stratify their clinical risk, and allow earlier recognition of the disease in genotypic positive patients. In this project, we also aim to better discriminate hypertrophic aetiologies such as (familial) hypertrophic cardiomyopathy and amyloidosis patients from other non-hereditary LVH aetiologies, as well as improve the clinical description of patients presenting with coexisting conditions (like hypertension or aortic valve stenosis).

Experience in clinical cardiac imaging. Basic knowledge of machine learning. Clinical specialist or resident in Cardiology.