Developing a machine learning tool to improve prognostic and treatment response assessment on cardiac MRI data (360G-Wellcome-215799_Z_19_Z)

Cardiovascular diseases account for 26% of deaths in the UK. Current clinical imaging assessments rely on manual or semi-automated measurements. Emerging approaches focus on individual parts of the heart. We have developed the first tensorbased machine learning approach that holistically assesses the heart and surrounding structures on cardiovascular magnetic resonance imaging (CMRI) scans. We will develop this approach into a tool that can identify patients who respond to treatment or who will die early. Key advantages are rapid holistic assessment, minimal human error and full transparency with direct visualisation of features for the disease. We will assemble a large cohort of CMRI scans from 5, 000 patients with pulmonary hypertension, a severe condition affecting the heart, and assess the ability to predict treatment response and likelihood of early death. This tool will revolutionise disease assessment, and improve treatment delivery and patient care.