Abnormal Heart Rhythm Detection Based on Spectrogram of Heart Sound using Convolutional Neural Network

• Published in: 2018 6th International Conference on Cyber and IT Service Management (CITSM) pp. 1 - 4
• Main Authors: Wibawa, Made Satria, Maysanjaya, I Md. Dendi, Novianti, Ni Kadek Dwi Pradnyani, Crisnapati, Padma Nyoman
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.08.2018
• Subjects: convolutional neural network; Convolutional neural networks; Feature extraction; Heart beat; heartbeat sound; Spectrogram; Testing; Training
• DOI: 10.1109/CITSM.2018.8674341

Based on WHO's data in 2012, cardiovascular diseases (CVD) is the most leading contributor to mortality, estimated at 17.5 million deaths. There are two effective way to examined heart condition, i.e. electrocardiogram (ECG) and echocardiogram. Although those two methods are effective, both methods are relatively expensive. The other method that can be used to examined heart condition is auscultation method. Auscultation requires substantial clinical experience. This process very subjective because it depends on the experience and the ability of paramedic hearing. Numerous research on developing an automatic system for detecting abnormal heart sound already conducted, but these studies still have some weaknesses making the accuracy in the classification less optimal. So, this paper aims to developed a scheme to classify the heart condition based on spectrogram of heartbeat sounds. To help increase the accuracy of the diagnosis, through this study, a CAD-based scheme was developed by applying a Convolutional Neural Network (CNN) algorithm. The achieved classification accuracy is 82.75%. This indicates that the proposed method has a potential to be implemented in the development of a computer aided diagnosis for CVD.