The most crucial component of prenatal diagnosis is the assessment of the unborn cardiovascular structure and anatomy using the ultrasound procedures. It is inevitable that the conventional approaches pose problems owing to the abundance of speckles in ultrasound videos, the diminutive size of fetal cardiac structures, and its unfixed fetal postures. So, a deep learning model was generated to fully automate the analysis of 2-dimensional cross-sectional images from fetal echocardiography in order to address these problems. This paper provides a real-time fetal cardiac substructure detection using electrocardiography video with the U-Y net framework. It predicts cardiac substructure objects, boxes, and class probabilities. Fetal echocardiography footages were trained using the newly developed U-Y Net architecture based on YOLOV7 neural network and then improved to function optimally in order to obtain consistent results. In order to assist ultrasound professionals to identify the fetal heart standard segment in echocardiography, a powerful machine-learning recognition model is designed by integrating ultrasound images with AI technology. The samples of the fetal apical four-chamber heart, three-vessel catheter, three-vessel trachea, right ventricular outflow tract, and left ventricular outflow tract has been collected at 20–24 weeks during pregnancy from a various hospital. The results of the experiment indicate that this technique can successfully distinguish between the anatomical components of various fetal heart sections and evaluate the standard sections in accordance with these anatomical components. This emphasizes a strong foundation for diagnosing congenital heart disease of fetus using by ultrasound imaging procedures.