Quantitative analysis of cardiac magnetic resonance in population imaging

According to the World Health Organisation, cardiovascular diseases are the most prevalent cause of death worldwide and taking nearly 18 million lives each year. Identifying individuals at risk of cardiovascular diseases and ensuring they receive appropriate treatment in time can prevent premature deaths. Early quantitative assessment of cardiac function, structure, and motion support preventive care and early cardiovascular treatment. Therefore, fully automated analysis and interpretation of large-scale population-based cardiovascular magnetic resonance imaging studies become of high importance. This analysis helps to identify patterns and trends across population groups, and accordingly, reveal insights into key risk factors before diseases fully develop. To date, few large-scale population-level cardiac imaging studies have been conducted. UK Biobank (UKB) is currently the world’s most extensive prospective population study, which in addition to various biological and physical measurements, contain cardiovascular magnetic resonance (CMR) images to establish cardiovascular imaging-derived phenotypes. CMR is an essential element of multi-organ multi-modality imaging visits for patients in multiple dedicated UK Biobank imaging centres that will acquire and store imaging data from 100,000 participants by 2023. This thesis introduces CMR image analysis methods that appropriately scales up and can provide a fully automatic 3D analysis of the UKB CMR studies. Without manual user interactions, our pipeline performs end-to-end image analytics from multi-view cine CMR images all the way to anatomical and functional quantification. Besides, our pipelines provide 3D anatomical models of cardiac structures, which enable the extraction of detailed information of the morphodynamics of the cardiac structures for subsequent associations to genetic, omics, lifestyle habits, exposure information, and other available information in population imaging studies. We present the quantification results from 40,000 subjects of the UK Biobank at 50 time-frames, i.e. two million image volumes.