Simultaneous and correlated detection of endocardial and epicardial borders on short-axis MR images for the measurement of left ventricular mass.

摘要

A method was developed for fast, reproducible assessment of left ventricular mass with detection of the endocardial and epicardial borders. The method makes use of short-axis magnetic resonance (MR) imaging in conjunction with an integrated segmentation approach that includes region growing, edge detection, and adaptive thresholding. The mass of the left ventricle in nine Yucatan minipigs was measured ex vivo and compared with the mass as determined in vivo with both full automated segmentation and an interactive process involving two trained observers. The test sample consisted of 59 sections obtained during diastole. There was good correlation between actual mass and both automatically and manually assessed mass. Linear regression analysis showed that the difference between ex vivo measurement and automatic mass assessment (standard error of estimation [SEE] = 5 g) was similar to interobserver variability in interactive mass assessment (SEE = 4 g). Mass assessment with short-axis MR imaging and an integrated segmentation approach does not require location of the ventricular cavity and allows precise delineation of the endocardial and epicardial borders. The results of this experimental study are encouraging for the potential use of this technique in clinical cardiac functional analyses.