Classification Probability Analysis for Arrhythmia and Ischemia Using Frequency Domain Features of QRS Complex

Abstract

Two of the most common cardiovascular diseases are myocardial ischemia and cardiac arrhythmias. Using the frequency domain features of QRS complex (i.e., frequency of the maximum peak in power spectrum and total average power) the proposed approach analyzes classification probability for these diseases by implementing Linear Discriminant Analysis (LDA) and Decision Tree. Moreover the classification probability is visualized using Naive Bayes classification algorithm. The methodology includes the QRS complex detection technique which is mainly comprises of three stages: Stage-1 - baseline drifts and noise cancellation using Moving Average Filter (MAF) and Stationary Wavelet Transform (SWT); Stage-2 - R-peaks localization using threshold based windowed filter: Stage-3 - Q and S inflection points detection using search interval method. To perform uniform classification probability analysis, the proposed methodology is evaluated with 108 selected episodes which show 100% accuracy in QRS complex detection. The 108 episodes includes 36 lengthy ECG recordings from FANTASIA database (healthy subjects), MIT-BIH Arrhythmia database (arrhythmic subjects) and Long-Term ST database (ischemic subjects) respectively. Moreover, the energy surface distribution of segmented QRS complex is analysed with Short-Term Fourier Transform (STFT) which transforms time domain information of the complex into time-frequency domain.

Keywords

• Ischemia
• Arrhythmia
• LDA
• Decision tree
• Naive Bayes classification