A Semi-Automated Approach for the Quantification of the Left Ventricle Chamber Volumes From Cine Magnetic Resonance Images

Cardiac magnetic resonance imaging is the golden standard for the quantification of left ventricular function. However manual segmentation of the endocardium and epicardium boundaries has proven to be a time-consuming task, and the tools used in in clinical practice depend on the cardiologist's experience to assess the mitral valve plane, identifying papillary muscles and trabeculations. The goal of this work was to develop a new algorithm for the segmentation of the endocardium and epicardium of the left ventricle. To this end both a probabilistic and a Malladi-Sethian level-set models were used, followed by morphological operators and curvature flow to regularize the surfaces and exclude unconnected regions. The proposed algorithm and the ensuing workflow were applied to cine magnetic resonance data in 12 patients. Linear regression and Bland-Altman analysis for end-diastolic, end-systolic volumes and ejection fraction were performed, and clinical indexes derived from the surfaces were in good agreement with the ones obtained by manual tracings. This work provides the basis for faster and still accurate quantification of the cardiac function from cardiac magnetic resonance, and the basis for further processing aimed at the assessment of heart remodeling and diseases.