AUTHOR=Wu Lanping , Dong Bin , Liu Xiaoqing , Hong Wenjing , Chen Lijun , Gao Kunlun , Sheng Qiuyang , Yu Yizhou , Zhao Liebin , Zhang Yuqi 

TITLE=Standard Echocardiographic View Recognition in Diagnosis of Congenital Heart Defects in Children Using Deep Learning Based on Knowledge Distillation

JOURNAL=Frontiers in Pediatrics

VOLUME=Volume 9 - 2021

YEAR=2022

URL=https://www.frontiersin.org/journals/pediatrics/articles/10.3389/fped.2021.770182

DOI=10.3389/fped.2021.770182

ISSN=2296-2360

ABSTRACT=Objective: To evaluate the feasibility and accuracy of standard echocardiographic view recognition in children using convolutional neural networks (CNNs) via knowledge distillation. 
Methods: A new CNN method for automatic echocardiographic view recognition utilizing knowledge distillation was proposed. 
Results: A total of 350,654 echocardiographic image slices from 3,505 subjects were used to train and validate the proposed echocardiographic view recognition model where 23 standard echocardiographic views commonly used to diagnose congenital heart defects in children were identified. The F1 scores of a majority of views, including the subcostal coronal / sagittal view of the atrium septum, the apex four / five-chamber view, the low parasternal four-chamber view, the parasternal short-axis view at the level of the mitral valve, and so on, were all greater than 0.90. 
Conclusions: In this study, an effective deep learning-based neural network method was proposed to identify commonly used standard echocardiographic views in the diagnosis of congenital heart defects in children. A knowledge distillation method was applied to maximize the accuracy of the network, while the complexity of the network remained unchanged. This study provides a solid foundation for the subsequent use of artificial intelligence to identify congenital heart defects.