Convolutional neural network for semantic segmentation of fetal echocardiography based on four-chamber view

The acute shortage of trained and experienced sonographers causes the detection congenital heart defects (CHDs) extremely difficult. In order to minimize this difficulty, an accurate fetal segmentation early location such structural abnormalities prior delivery is essential. However, process not easy task due small size structure. Moreover, manual for identifying standard cardiac planes, primarily based on a four-chamber view, requires well-trained clinician experience. paper, CNN method using U-Net architecture was proposed automate planes from ultrasound images. A total 519 images obtained three videos. All data divided into training testing data. consist 106 slices tasks, i.e. atrial septal defect (ASD), ventricular (VSD), normal. post-processing needed enhanced result. combination technique with Otsu thresholding gives best performances 99.48%-pixel accuracy, 96.73% mean 94.92% intersection over union, 0.21% error. future, implementation Deep Learning in study CHDs holds significant potential novel heterogeneous hearts.