Dynamic cardiac analysis using cine magnetic resonance images

An efficient semautomated technique has been developed to determine both endocardial and epicardial left ventricular borders using cine MR images. These borders can be used to quantitate cardiac functions such as stroke volume (SVX ejection fraction (EF), and left ventricular mass (LVM). This thesis describes the development of a semiautomated dynamic cardiac analysis that requires minimal operator interaction and allows efficient contour correction. A study group of five volunteers underwent cine MR scans for evaluation. Two sets of contours were generated from the study group: 1) A complete set for all images using the semiautomated technique (1260 contours), 2) A set of manually traced contours from the selected end-systolic and end-diastolic images (160 contours). The mean-pixel-deviation between the computer-determined and hand-traced contours using all 160 contours was found to be 1.64, which corresponded to 6.7^ volume difference. The time for generating and correcting 12 6 endocardial and 126 epicardial contours was approximately 50 minutes. The experimental results strongly indicate that the semiautomated processing followed by correction can produce clinically acceptable contours.