Publication

Abstract : Assessment of contractile function of heart plays a major role in characterization of cardiac abnormality. Worldwide, cardiovascular diseases (CVD) are of primary concern in healthcare. Hence, aim of this study is to evaluate the diagnostic accuracy of myocardial geometric features extracted from cardiovascular magnetic resonance (CMR) images and optimized classifier in monitoring of the CVD progression. This work investigates CMR slices of 100 subjects (Normal - 24, Mild - 23, Moderate - 21, Severe - 14, Hypertrophic - 18) from Kaggle challenge database. The segmentation of myocardium is carried out in all the slices at end-diastole (ED) and end-systole (ES) using LabVIEW. Further, geometric features have been extracted to capture the anatomical variations among the different classes. The appropriate choice of kernel parameters and feature set has significant impact on the performance of the classifier. Harris Hawks optimization (HHO) algorithm is used in integration with the geometric feature set to improve the prediction accuracy. The intensity based segmentation is able to effectively capture the myocardial boundaries at ED and ES. The entire feature set has provided an accuracy of 70 % prior to optimization. As the geometric features are able to capture the structural changes in myocardium under various cardiac abnormalities, the HHOSVM yields improved classification accuracy of 76.7 % than without optimizer. Thus, this framework would aid disease prognosis and planning of effective treatment strategies. REFERENCES