Prediagnosis of Heart Failure (HF) Using Deep Learning and the Korotkoff Sound

Heart failure (HF) is a devastating condition that impairs people’s lives and health. Because of the high morbidity mortality associated with HF, early detection becoming increasingly critical. Many studies have focused on field heart disease diagnosis based sound (HS), demonstrating feasibility signals in diagnosis. In this paper, we propose non-invasive method for HF deep learning (DL) network Korotkoff (KS). The accuracy KS-based prediagnosis was investigated utilizing continuous wavelet transform (CWT) features, Mel frequency cepstrum coefficient (MFCC) signal segmentation. Fivefold cross-validation applied to four DL models: AlexNet, VGG19, ResNet50, Xception, performance each model evaluated using (Acc), specificity (Sp), sensitivity (Se), area under curve (AUC), time consumption (Tc). results reveal models MFCC datasets significantly improved when compared CWT datasets, performed considerably better non-segmented dataset than segmented dataset, indicating KS segmentation feature extraction had significant impact CHF performance. Our eventually achieves Acc (96.0%), Se (97.5%), Sp (93.8%) comparative study data set. research demonstrates proposed paper could accomplish accurate prediagnosis, which will offer new approaches more convenient way achieve prevention.