Cardiac motion analysis in MRI for classification

Abstract

Although several techniques exist for the analysis of cardiac tagged MR images, a rapid screening tool do not yet exits. Our proposed technique tries to perform rapid classification to diagnose the abnormalities in human left ventricle and the final aim of this study is to identify the investigated myocardium in the analyzed tagged MR images as pathological and non.In this thesis, images are first analyzed using harmonic phase (HARP) analysis and synthetic tags are computed over the myocardium. The data is normalized to perform a comparison between different myocardiums having various tag lines and time frames. The aim of the normalization is to eliminate the shift, scale and rotation variance. Cubic curves are fitted to the normalized tags and curve parameters are compared at various regions of the myocardium. In this initial study, the curve parameters are examined with probability density function between normal and diseased hearts, such as left ventricles with dilated cardiomyopathy (DCM) and infarcted regions. Finally, the confusion matrix is evaluated to examine the correctness of the segmentation algorithm. This method could be a very fast and automatic screening tool for identifying diseased locations in tagged MRI.