Human machine interfacing technique for diagnosis of ventricular arrhythmia using supervisory machine learning algorithms

• Published in: Concurrency and computation Vol. 33; no. 4
• Main Authors: Karnan, Hemalatha, Natarajan, Sivakumaran, Manivel, Rajajeyakumar
• Format: Journal Article
• Language: English
• Published: Hoboken Wiley Subscription Services, Inc 25.02.2021
• Subjects: Algorithms; Arrhythmia; Cardiac arrhythmia; Diagnosis; electrocardiogram; Electrocardiography; Estimates; Feature extraction; Iterative methods; Machine learning; mean grey wolf algorithm; Optimization; Parameter estimation; Parametric analysis; Particle swarm optimization; ranking score; Reducing agents; Signal classification; Support vector machines; ventricular arrhythmias
• ISSN: 1532-0626; 1532-0634
• DOI: 10.1002/cpe.5001

Summary The state of art to integrate bio‐signals with computer based diagnosis is taking dominance. The man‐machine interface is useful for early and immediate clinical interpretation. The electrocardiogram (ECG) signal plays a vital role in revealing the possible data towards categorizing normal and abnormal cardiac functioning. The fatal conditions exhibited by ventricular arrhythmias (VA) pose a remarkable change in the feature set of the ECG signals. In this work, a novel approach to segregate the superior feature toward the ventricular arrhythmias are extracted using feature ranking score algorithm (FRSA). The FRSA collects feature vectors in three different domains and ranks it to find out the more prevalent feature for diagnosis of VA. The Support Vector Machine (SVM) classifier is administered by supervisory machine learning optimization algorithm Mean Grey Wolf Optimization (MGWO). The performance estimates of SVM‐MGWO is compared for classification of VA signals with other optimization also like SVM‐Particle Swarm Optimization (SVM‐PSO) and SVM‐Grey Wolf Optimization (SVM‐GWO). The non‐parametric and parametric analysis evidently shows the improved performance of feature parameter estimates for classification. The accuracy of classification for SVM‐MGWO attains 100% for finding test data with VA at a minimal convergence iteration while comparing it with the other mentioned supervisory algorithms. The standard deviation during parametric analysis is negligible, which reveals the fact that reductant feature extracted and utilized for testing of ECG data is minimal. The performance estimates attained by the proposed algorithm shows the selection of optimal feature for the findings of VA through ECG. The man‐machine interface aides in the early diagnosis of ventricular arrhythmias using non‐invasive diagnosing tool, the ECG.