Detection of heart valve disorders from PCG signals using TQWT, FA-MVEMD, Shannon energy envelope and deterministic learning
Wei Zeng,Zixiang Lin,Chengzhi Yuan,Qinghui Wang,Fenglin Liu,Ying Wang
DOI: https://doi.org/10.1007/s10462-021-09969-z
IF: 9.588
2021-02-18
Artificial Intelligence Review
Abstract:Heart valve disorders (HVDs) are the major causes of cardiovascular diseases (CVD), which may be detected at the early stage using routine auscultation examination. The phonocardiogram (PCG) is a graphical representation of the physiological condition of the heart, which differs with respect to heart diseases. It is closely related to valve functionality which provides vital information for the diagnosis of CVD. However, visual inspection of PCG is tedious and error-prone, which makes it necessary and urgent to develop an automated system for the detection of HVDs with PCG recordings. In the present study we propose a novel method for the identification and classification of normal and abnormal non-segmented PCG recordings based on hybrid signal processing tools and deterministic learning theory. First, PCG signal and its first derivative are decomposed into a set of frequency subbands with a number of decomposition levels by using the tunable Q-factor wavelet transform method. Second, fast and adaptive multivariate empirical mode decomposition decomposes the subbands of the PCG signal and its derivative into scale-aligned intrinsic mode components (IMFs). The first two IMFs are extracted, which contain most of the energy of the PCG signal and its derivative and are considered to be the predominant IMFs. Third, Shannon energy is used to extract the characteristic envelope of predominant IMFs. The properties associated with the nonlinear PCG system dynamics are preserved. They are utilized to derive features, which demonstrate significant difference in PCG system dynamics between normal versus abnormal individual heartbeats. Fourth, neural networks are then used to model, identify and classify PCG system dynamics between normal and abnormal PCG signals based on deterministic learning theory. Finally, experiments have been carried out on a publicly available PCG database to verify the effectiveness of the proposed method, which include two types of classification, one for binary classification (normal vs. abnormal) and the other for multi-class classification (normal vs. aortic stenosis vs. mitral regurgitation vs. mitral stenosis vs. mitral valve prolapse). The overall average accuracy for binary, four-class and five-class classification are reported to be 97.75, 98.69 and 98.48%, respectively. The proposed method has obtained the highest overall accuracy in comparison to other state-of-the-art approaches using the same database, which can serve as an assistant diagnostic tool for the automated detection of HVDs in clinical applications.
computer science, artificial intelligence