Automatic detection of myocardial perfusion defects using object-based myocardium segmentation

Determining the relevance of coronary artery pathologies is a major task in diagnosis and therapy planning for coronary heart disease. Magnetic resonance (MR) perfusion imaging provides non-invasive means to assess the influence of artery stenosis on the myocardial perfusion. The overall goal of the presented approach is to enable a fully automatic data analysis that supports both the conventional AHA model perfusion quantification and a voxel-based segmentation of suspicious regions in the heart muscle. To this end, an automatic pipeline for detecting and segmenting perfusion defects was developed and evaluated. The myocardium is segmented using an object-based image analysis approach, which then forms the basis for the perfusion parameter calculation and detection of underperfused regions. The approach has been applied to six datasets of patients with known multivessel coronary heart disease. Results show a good agreement with findings from MR delayed enhancement examination and conventional coronary angiography.