AUTHOR=Li Yang , Yao Qianqian , Yu Haitao , Xie Xiaofeng , Shi Zeren , Li Shanshan , Qiu Hui , Li Changqin , Qin Jian 

TITLE=Automated segmentation of vertebral cortex with 3D U-Net-based deep convolutional neural network

JOURNAL=Frontiers in Bioengineering and Biotechnology

VOLUME=Volume 10 - 2022

YEAR=2022

URL=https://www.frontiersin.org/journals/bioengineering-and-biotechnology/articles/10.3389/fbioe.2022.996723

DOI=10.3389/fbioe.2022.996723

ISSN=2296-4185

ABSTRACT=Abstract
Objectives: We developed a 3D U-Net-based deep convolutional neural network for the automatic segmentation of vertebral cortex. The purpose of this study was to evaluate the accuracy of the 3D U-Net deep learning model.
Methods: 418 3D images were divided into training, validation and test sets according to 7:2:1,and combined with CT scan images using deep learning.Ten fold cross-validation was utilized in the training and validation sets. Through experiments comparing 3D U-Net, Res U-Net, Ki U-Net and Seg Net, the 3D U-Net deep network was finally selected as the segmentation method. The network included an encoding part and a decoding part. The encoding part was used to analyze the entire image and perform feature extraction and analysis. The decoding part generated a segmented block image. The segmentation performance and inference speed of 3D U-Net and the other three deep learning algorithms were evaluated using DSC and FPS.
Results: The DSC of 3D U-Net is better than the other three strategies, reaching 0.7139 on the validation set and 0.7038 on the test set, indicating promising automated segmentation results.
Conclusion: Cortical bone can be effectively segmented based on 3D U-net.