A Multi-Layer Acoustic Neural Network-Based Intelligent Early Diagnosis System for Rheumatic Heart Disease

Abstract

Rheumatic Heart Disease (RHD) is a disorder of heart caused by streptococcal throat infection followed by the organ damage, irreversible valve damage and heart failure. Acute Rheumatic Fever (ARF) is a precursor to the disease. Sometimes, RHD can occur without any signs or symptoms, and if there are any symptoms, they occur with the infection in the heart valves and fever. Due to these issues, respiratory problems occur with chest pain and tremors. Additionally, the symptoms include faint, heart murmurs, stroke and unexpected collapse. The techniques available try to detect the RHD as early as possible. Although the recent medical health care department uses crucial techniques, they are not accurate in terms of symptom classification, precision and prediction. On the scope, we are developing Multi-Layered Acoustic Neural (MLAN) Networks to detect the RHD symptoms using heart beat sound and Electrocardiogram (ECG) measurements. In this proposed MLAN system, the novel techniques such as multi-attribute acoustic data sampling model, heart sound sampling procedures, ECG data sampling model, RHD Recurrent Convolutional Network (RRCN) and Acoustic Support Vector Machine (ASVM) are used for increasing the accuracy. In the implementation section, the proposed model has been compared to the Long Short-Term Memory-based Cardio (LSTC) data analysis model, Cardio-Net and Video-Based Deep Learning (VBDL) techniques. In this comparison, the proposed system has 10%–17% higher accuracy in RHD detection than existing techniques.