Abnormal heart sound detection using temporal quasi-periodic features and long short-term memory without segmentation. (August 2019)

• Record Type: Journal Article
• Authors: Zhang, Wenjie; Han, Jiqing; Deng, Shiwen
• Is Part Of: Biomedical signal processing and control. Volume 53(2019)
• Publication Date: 2019-08
• Subjects: Heart sound -- Average magnitude difference function -- Long short-term memory; 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.2019.101560
• Languages: English
• ISSNs: 1746-8094

Abstract

Highlights: The temporal quasi-periodic features are obtained by AMDF. In-depth analysis of the dependency relation within the features is given. The dependency relation is extracted by long short-term memory. More discriminative features are obtained. The weighted loss function and the threshold tuning strategy are combined. Abstract: Abnormal heart sound detection is an effective and convenient method for the preliminary diagnosis of heart diseases. In this study, we propose a novel method for abnormal heart sound detection using temporal quasi-periodic features and long short-term memory without segmentation. In the proposed method, the spectrogram of the heart sound signal is extracted using the short-time Fourier transform in the first step. Subsequently, the temporal quasi-periodic features of the heart sound signal are calculated by the average magnitude difference function from the spectrogram in different frequency bands. Moreover, to extract the dependency relation within the temporal quasi-periodic features, the method of long short-term memory is applied. Thus, more discriminative features are obtained. Finally, the performance of the proposed method is evaluated on the public dataset offered by the 2016 PhysioNet/Computing in Cardiology Challenge, and the results indicate that our proposed method is competitive compared with the state-of-the-art abnormal heart sound detection methods.