Application of Signal Processing and Deep Hybrid Learning in Phonocardiogram and Electrocardiogram Signals to Detect Early Stage Heart Diseases

Abstract

This thesis describes a variety of projects on analyzing two common biomedical signals known as Phonocardiogram (PCG) and Electrocardiogram (ECG/EKG) to detect early-stage heart diseases. The projects include the design of prototypes to compress, denoise, segment, and classify PCG and ECG signals accurately. PCG signal is the graphical representation of heart sound which represents the mechanical activities of the human heart. PCG signal contains useful information about the functionality and the condition of the heart. ECG signal represents the electrical activities of the human heart. ECG signal has been widely used in hospitals and clinics to diagnose cardiac diseases. Analysis of PCG and ECG signals is critical in diagnosis of different cardiac diseases as they can provide early indication of potential cardiac abnormalities. Extracting cardiac information from PCG and ECG signals to diagnose heart diseases in the initial stage can play a vital role in remote patient monitoring. In this thesis, we have combined different signal processing techniques, Machine Learning (ML), and Deep Learning (DL) methods to compress, denoise, segment, and classify PCG and ECG signals effectively and accurately. First, PCG signals are compressed and denoised by using a multi-resolution analysis technique based on the Discrete Wavelet Transform (DWT). Then, a segmentation algorithm, based on the Shannon energy envelope and zero crossing is applied to segment the PCG signal into four major parts: the first heart sound (S1), the systole interval, the second heart sound (S2), and the diastole interval. Finally, Mel-scaled power spectrogram and Mel-frequency cepstral coefficients (MFCC) are employed to extract informative features from PCG signals, which are then fed into a classifier to classify each PCG signal into a normal or an abnormal signal. We have combined traditional ML and DL approaches to develop Deep Hybrid Learning (DHL) models. A Convolutional Neural Network (CNN) is used along with seven traditional ML methods including Logistic Regression (LR), Random Forest (RF), K-Nearest Neighbors (KNN), Decision Tree (DT), Naive Bayes (NB), Support Vector Machine (SVM), and AdaBoost (AB) to build hybrid PCG classification models. Our experimental results have shown that significant improvements in the classification accuracy can be achieved by using DHL models compared to traditional ML and DL models. We have also applied the same methods to analyze ECG signals and got promising results. Besides providing valuable information regarding heart condition, our proposed signal processing and DHL approaches can help cardiologists to take appropriate and reliable steps toward diagnosis if any cardiovascular disorder is found in the initial stage.