Fetal Heart Disease Detection Via Deep Reg Network Based on Ultrasound Images

Congenital heart disease (CHD) is the most prevalent congenital ailment. One in every four newborns born with serious coronary artery disease will require surgery or other early therapy. Early identification of CHD in the fetal heart, on the other hand, is more critical for diagnosis. Extracting information from ultrasound (US) images is a difficult and time-consuming job. Deep learning (Dl) CNNs have been frequently utilized in fetal echocardiography for CAD identification to overcome this difficulty. In this work, a DL based neural network is proposed for classifying the normal and abnormal fetal heart based on US images. A total of 363 pregnant women between the ages of 18 and 34 weeks who had coronary artery disease or fetal good hearts were included. These US images are pre-processed using SCRAB (scalable range based adaptive bilateral filter) for eliminating the noise artifacts. The relevant features are extracted from the US images and classify them into normal and CHD by using the deep Reg net network. The proposed model integrates the Reg net -module with the CNN architecture to diminish the computational complexity and, simultaneously, attains an effectual classification accuracy. The proposed network attains higher accuracy of 98.4% for the normal and 97.2% for CHD. To confirm the efficiency of the proposed Reg net is compared to the various deep learning networks