Abstract

Info

Title:: Advances in automatic segmentation of medical images in radiotherapy

Author(s):: ZHANG Fuli1; WANG Yadi2; WANG Qiusheng3; 1. the Seventh Medical Center of Chinese PLA General Hospital, Beijing 100700, China 2. the Fifth Medical Center of Chinese PLA General Hospital，Beijing 100071, China 3. School of Automation and Electrical Engineering, Beihang University, Beijing 100191, China

Keywords:: radiotherapy medical image automatic segmentation algorithm review

Abstract:: The clinical application and development of traditional (non-deep learning) automatic segmentation algorithms and the newly emerging deep learning (convolutional neural networks and full convolutional networks) algorithms in radiotherapy are reviewed. The first-generation automatic segmentation technology represented by intensity threshold algorithm, the second-generation represented by clustering algorithm, and the third-generation represented by atlas-based segmentation algorithm are briefly summarized. Although the fourth-generation automatic segmentation technology based on deep learning has achieved great improvements in robustness, consistency, efficiency, there are still some limitations such as deviations with unknown reasons, inconsistent image acquisition protocols, lack of high-quality data sets, etc., and these deficiencies need to be continuously improved in future work. Finally, how to conduct clinical commissioning and quality assurance for automatic segmentation software is discussed.