Automated algorithm for computing left ventricle volume changes from cine-MR images

To determine the left ventricle (LV) chamber volume from cine-MR images, it is necessary to locate the aortic / mitral valve plane with guidance from the long-axis images to isolate (or trim) the LV. However, manual identification of the trimming planes for all frames in the cine-MR scan is tedious and does not guarantee that the LV myocardium volume is kept constant. This work aims to develop an automatic algorithm to compute the LV volumetric changes during the cardiac cycle while ensuring constant myocardium volume. We reconstruct 4D (i.e., spatial + time) LV mesh models from border-delineated cine-MR images. The trimming value at end-diastole (ED) is determined manually and the corresponding ED myocardium volume is computed and used as a reference. For the remaining frames, an iterative Newton-Raphson-like procedure is used to find the trimming plane such that the resulting myocardium volume is conserved. The end-systole (ES) frame is determined as the frame with the smallest LV chamber volume. ECG-gated cine-MR scans of 8 normal subjects were acquired and the LV chamber volumes over the cardiac cycle were computed based on our proposed algorithm. The results showed a mean ejection fraction (EF) of 62.4 ± 4.4%, which agrees with the clinical EF of 67.1 ± 6.2%.