Semi-automated quantification of left ventricular volumes and mass from cardiac magnetic resonance images by level set models

Cardiac magnetic resonance imaging (CMRI) is the standard for estimates of LV volumes, ejection fraction and mass. These computations are based on extensive manual tracing of endocardial and epicardial borders and are subjective and time-consuming. We developed a new technique for semi-automated suface detection for the measurement of LV end-systolic and end-diastolic volumes, ejection fraction and mass from CMRI data. This procedure is performed in the 3D domain, and does not rely on geometrical approximations. Twenty-two consecutive patients referred for CMRI were studied. Measurements were compared with the values derived by manual tracing using linear regression and Bland-Altman analyses. For both volumes, ejection fraction and mass, the analysis resulted in high correlation coeflcienrs and depicted no signijkant biases and narrow limits of agreement. The proposed technique is fast and objective and provides accurate measurements of LV volumes, ejection fraction and mass.