Identification of systolic and diastolic heart failure progression with Krawtchouk moment feature-aided Harris hawks optimized support vector machine

• Published in: Signal, image and video processing Vol. 16; no. 1; pp. 127 - 135
• Main Authors: Muthunayagam, Muthulakshmi, Ganesan, Kavitha
• Format: Journal Article
• Language: English
• Published: London Springer London 01.02.2022; Springer Nature B.V
• Subjects: Algorithms; Computer Imaging; Computer Science; Feature extraction; Heart failure; Image Processing and Computer Vision; Image segmentation; Inhomogeneity; Magnetic resonance imaging; Multimedia Information Systems; Myocardium; Noise intensity; Optimization; Original Paper; Pattern Recognition and Graphics; Shape; Signal,Image and Speech Processing; Support vector machines; Vision; Heart failure; Support vector machine; Cardiovascular magnetic resonance; Krawtchouk moment; Harris hawks optimization; Corr-entropy level set
• ISSN: 1863-1703; 1863-1711
• DOI: 10.1007/s11760-021-01978-3

The systolic and diastolic heart failure (HF) subjects are typically categorized based on clinical indices only. The relationship between different stages of systolic and diastolic heart failure and left ventricle (LV) myocardial tissue variations is presented in this work. The corr-entropy and optimized edge criterion has been incorporated into the level set (CEOELS) for effective segmentation of myocardium in cardiovascular magnetic resonance images to handle noise, intensity inhomogeneity and contour initialization. In order to learn shape and local variations in segmented myocardium, Krawtchouk moment features are computed for ten different moment orders. The relevant extracted features are obtained through Harris hawks optimization algorithm. The optimized features are fed to support vector machine (SVM) that uses fivefold cross-validation approach for classification. Experimental results show that CEOELS has provided better segmentation of LV blood cavity and myocardium with a similarity measure of 0.93 and 0.92, respectively. It is also observed that individual Krawtchouk moment orders greater than 30 have provided better HF prediction performance. Consequently, optimized Krawtchouk moment features produced an increased overall accuracy (80.8%) than individual feature sets. Significant improvement has also been achieved in distinction of hyperdynamic patients from normal and systolic dysfunction subjects that is less explored.