Multi-modal cardiac function signals classification algorithm based on improved D-S evidence theory. (January 2022)

• Record Type: Journal Article
• Authors: Li, Jinghui; Ke, Li; Du, Qiang; Chen, Xiangmin; Ding, Xiaodi
• Is Part Of: Biomedical signal processing and control. Volume 71(2022)Part A
• Publication Date: 2022-01
• Subjects: Multi-modal cardiac function signal -- Wavelet scattering transform -- Support vector machine -- Basic probability assignment (BPA) -- D-S evidence theory; Signal processing -- Periodicals; Biomedical engineering -- Periodicals; Signal Processing, Computer-Assisted -- Periodicals; Image Processing, Computer-Assisted -- Periodicals; Biomedical Engineering -- Periodicals; 610.28
• DOI: 10.1016/j.bspc.2021.103078
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: Multi-modal cardiac function signal studied is innovative to judge heart disease. Classification performance of multi-modal signal is better than single-modal signal. Wavelet scattering transform is used to extract rich feature information. It is a new attempt to apply D-S theory to fuse multi-modal cardiac function signal. The D-S theory is improved to obtain the better classification effect. Abstract: According to the different generation mechanisms of multi-modal cardiac function signals such as phonocardiogram (PCG) signal and electrocardiogram (ECG) signal which reflect the different aspects of heart health, the classification algorithm of multi-modal cardiac function signals based on improved D-S evidence theory is proposed. The implement process of this algorithm is: firstly, the multi-modal cardiac function signals are acquired from database, which includes PCG signals and ECG signals collected synchronously. The wavelet scattering transform is selected to extract the characteristics. Then support vector machine (SVM) is used to classify the multi-modal cardiac function signals. The posterior probability is calculated based on the classification results, which is used to construct the basic probability assignment (BPA) function in the D-S evidence theory. The confusion matrices from the SVM classification results are used to estimate the local credibility. The weighted correction is carried out by the local credibility when constructing the BPA. Finally, the classification results are given according to the decision rules. The proposed method obtains the best performance results with 86.42%, 84.96%, 93.10%, 98.26% and 91.13% in terms of Accuracy, Sensitivity, Specificity, Precision and F1 Score. The experimental results show that the classification effect of the proposed method is better than the single-modal cardiac function signal classification method. It is also superior to tradition D-S evidence theory. This proposed algorithm not only improves the classification accuracy, but also lays the theoretical foundation for all-round and multi-angle diagnosis of heart disease.