Fast algorithm for estimating the regional mechanical function of the left ventricle from 4D cardiac CT data

Abstract

Cardiac pathologies are generally associated with regional ventricular dysfunction. Methods for estimating the regional myocardial motion from cardiac CT image data generally ignore the rotational velocities. Reasons for this include the challenges of sparse image deformation clues, low SNR and the low temporal resolution. In the current study we propose a fast algorithm for evaluating the mechanical function of the left ventricle from cardiac CT data. A compact parametric motion model is used to describe the regional 3D contraction and twist. The algorithm is based on regularized multi-homography image registration. The rotational velocities are estimated and compared to their respective values in the literature, with good agreement. Good performance in classifying the segments as normal or abnormal with respect to expert's visual scores is obtained.