Abstract:
Due to the problem that the accuracy of the existing algorithms was not good enough, a novel method of the classification of congenital heart disease (CHD) heart sound based on deep learning models with the variable logic theory model was put forward in this paper. Firstly, the heart sound signal was pre-processed to remove non-pathological noise and extract the heart sound envelope. Secondly, the variable logic theory was applied to the feature extraction, labeled the envelope data of the heart sound signal and converted into Pseudo-DNA sequence and visual analysis measurement data. Finally, the deep learning models, such as Inception_Resnet_v2 was used to perform multi-classification for some common CHD heart sounds and compared with other existing algorithms. There were 1000 cases of heart sound used in this study. The average accuracies of 0.931 for multi-classification of heart sound were obtained on test set by using the novel method. The experimental results showed that the algorithm performs better than the several existing algorithms. It is expected to be used for machine-assisted diagnosis of the initial diagnosis of congenital heart disease.