Development of cardioid based graph ECG heart abnormalities classification technique

• Main Authors: Mohd Azam, Siti Nurfarah Ain, Zainal, Nur Izzati, Sidek, Khairul Azami
• Format: Article
• Language: English
• Published: Asian Research Publishing Network (ARPN) 2015
• Subjects: TK7885 Computer engineering
• Online Access: http://irep.iium.edu.my/46305/1/jeas_1115_2979.pdf; http://irep.iium.edu.my/46305/; http://www.arpnjournals.org/jeas/research_papers/rp_2015/jeas_1115_2979.pdf
• Institution: Universiti Islam Antarabangsa Malaysia

In this study, the development of Cardioid based graph electrocardiogram heart abnormalities classification technique is presented. ECG signals in this work were acquired from a public online database UCD Sleep Apnea database (UCDB) with sampling rate of 250 Hz. Each recording has 60 seconds of electrocardiogram signals. Unique features were extracted using the Pan Tompkins algorithm, later Cardioid based graph was formed as the result of the differentiation process. The various shapes of closed-loop created were then observed. From the Cardioid loop, we evaluated the area and standard deviation to differentiate between normal and abnormal heartbeats. As a result, the area and standard deviation values of abnormal heartbeat were twice the value of a normal heartbeat thus indicating the differences between two types of heart morphologies. In order to justify the results, the signal is then classified by using Bayes Network classifier. Classification outcomes suggests that the proposed technique gives heart abnormality identification with a classification accuracy of as low as 12.5% when normal and abnormal heartbeat are matched (two different conditions). Thus, the output of the study suggests the proof-of-concept of our proposed mechanisms to detect heart abnormalities and has the potential to act as an alternative to the current techniques.