Detection of frequency sub-bands on Heart Rate Variability in Supra-ventricular Tachyarrhythmia patients using artificial neural networks

Supra-ventricular Tachyarrhythmia (SVTA) called as disturbances of heart around atria and Atrioventricular (AV) node is one of the most common heart arrhythmias. Heart Rate Variability (HRV) is a pointer for classification of autonomic nervous system (ANS) and heart arrhythmias. Wavelet Packet Transform (WPT) is an efficient tool for HRV like non stationary signals. This study presents critical frequency intervals for HRV analysis in SVTA patients and the effectiveness of these frequency intervals on base-bands. In the study, sub-frequency regions on HRV in MIT-BIH SVTA database obtained from half-hour ECG recordings of 78 patients are calculated and analyzed. Each data is decomposed in sub-frequency region using WPT with 8 levels and their domination effects of each sub-frequency region on its base-band are evaluated using Multi Layer Perceptron Neural Networks (MLPNN). While 0.0546875 - 0.078125 Hz on Low Frequency band (LF) points the highest accuracy value, subfrequency bands including 0.1171875 - 0.15625 Hz frequency interval have the lowest accuracy values. However, accuracy values of sub-frequency regions on High Frequency (HF) band points near values each other and the detection of dominant sub-bands got harder. In this band, while sub-frequency bands including 0.15625 - 0.2734375 Hz frequency interval point high accuracy value, sub-frequency bands including 0.25 - 0.328125 Hz frequency interval have lower accuracy values.