OC01.06: To evaluate the value of AI quality control in nine standard sections of the fetal heart

Abstract

To explore the application value of artificial intelligence (AI) deep learning method in standard section quality control of fetal heart. The images of fetal heart ultrasound examination of 8856 pregnant women undergoing pregnancy examination in Sichuan Provincial Maternity and Child Health Care Hospital from January to December 2020 were selected. According to the International Ultrasound Guidelines of Obstetrics and Gynecology, Nine fetal heart sections recommended to be retained were quality control on standard images of fetal heart by using artificial intelligence deep learning model and a team of doctors. A total of 92,000 ultrasound images of 8856 cases were evaluated. The average AP value of the AI method is 0.885, which can accurately identify the fetal echocardiography required to retain all the anatomical structures in the section. At the same time, for the evaluation of 92,000 images, the AI method judged that the average length of fetal heart ultrasound images is about 0.028 seconds per piece, while our doctor team, the average length is about 3.77 seconds per piece, used the AI-based method to evaluate the quality of fetal heart ultrasound images 134.6 times faster than the doctor team. The use of artificial intelligence integrated learning method to quality control the standard section of the fetal heart can reach the level of the expert team, which takes far less time than the manual evaluation of the doctor team.

Ähnliche Arbeiten