A deep residual networks classification algorithm of fetal heart CT images

• Published in: 2018 IEEE International Conference on Imaging Systems and Techniques (IST) pp. 1 - 4
• Main Authors: Lei, Li, Zhu, Haogang, Gong, Yuxin, Cheng, Qian
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.10.2018
• Subjects: CT images classification; deep residual network; fetal congenital heart disease; Instrumentation and measurement
• DOI: 10.1109/IST.2018.8577179

This paper proposes a deep residual networks classification algorithm of fetal heart CT images. It is difficult to diagnose Fetal Congenital Heart Disease (FCHD) due to medical CT images of fetal heart has much noisy than general natural scenes images and fetal body position is not fixed. These are great difficulties for medical experts so they cannot give every subject correct diagnosis. The algorithm in this paper exploits deep residual networks to classify the FCHD CT images and may give higher accuracy and precision than medical experts. The residual networks we proposed are based on ResNet34 [1] and a fully connected (FC) layer is added to the last layer of ResNet34 due to the binary classification for negative or positive of FCHD. This residual networks mechanism achieves superior performance than other baseline deep network for binary classification of FCHD.