Clustering of signal components within most likely ECG episodes to analyze the ECG-waves

• Published in: Pattern recognition and image analysis Vol. 19; no. 1; pp. 30 - 34
• Main Authors: Lahiri, T., Sarkar, S., Sanyal, S., Morozov, A. A., Obukhov, Yu. V.
• Format: Journal Article
• Language: English
• Published: Dordrecht SP MAIK Nauka/Interperiodica 01.03.2009; Springer Nature B.V
• Subjects: Analysis; Clustering; Computer Science; Fourier transforms; Heart attacks; Image Processing and Computer Vision; Image processing systems; Medical imaging; Pattern Recognition; Processing; Representation; Signal processing; Studies; Understanding of Images; Velocity; Diagnostic Efficiency; Root Mean Square Distance; Base Correction; Discrete Fourier Transform; Radial Artery Pressure
• ISSN: 1054-6618; 1555-6212
• DOI: 10.1134/S1054661809010040

A computational cardiology based approach to find the number of ECG-waves responsible for generation and presentation of an ECG episode is proposed. The methodology adopted here is based on the comparison and study of diagnostic efficiency calculated from 12 lead ECG data. The diagnostic-efficiency is measured by using NN classifier whereas clustered points of most likely ECG-episode constitute the diagnostic feature. The result is also cross-validated using episodes transformed by existing signal transform technique, “Discrete Fourier Transform” and another transformation. “Velocity-Wave number Transform.” The result indicates the possible number of episode-clusters giving maximum diagnostic efficiency and also gives a new insight on ECG waves.