Development of a Graphical User Interface for ECG Signals Classification Using Statistical Features Analysis
Abstract
Cardiac diseases are the most common cause of mortality in the world. The detection
of cardiac arrhythmias is not a straightforward process, since minor variations in the
electrocardiogram (ECG) signals cannot be easily identified manually. Therefore,
automatic detection and classification of cardiac arrhythmia would shorten the diagnostic
time and accelerate medical intervention resulting in reducing the mortality rate. In this
thesis, I have developed a simple and low-cost computer-aided diagnostic system using
MATLAB-based Graphical User Interface (GUI) to facilitate fast operation and access to
the data along with the overall accuracy of the system.
The acquired ECG signals are processed by wavelet-based filtering and feature
extraction techniques using Daubechies (db) wavelets to determine a combination of 15
statistical features. The significant wavelet features were subsequently used as categorical
inputs to perform pattern recognition of the ECG signals using artificial neural network
(ANN), support vector machine (SVM), and random forest (RF) and classify the output
into normal or abnormal classes. The performance of the proposed model was evaluated
using Massachusetts Institute of Technology-Beth Israel Hospital arrhythmia database
(MIT-BIH AD) over 46 ECG records including normal and arrhythmias signals. The
overall system performance was achieved with 98.3%, 95.65%, and 100% overall accuracy
using ANN, SVM, and RF, respectively.