ABSTRACT
This chapter presents a review of the recent advancements of the deep learning (DL)-based medical image multi-organ segmentation methods. The latest network architecture designs for medical image segmentation and applications of DL-based segmentation methods for auto-delineation of organs-at-risk for thoracic radiation treatment planning in radiation oncology are summarized. These methods are classified into six categories according to their network architectures, supervision, and popularity. A detailed review of each category is presented, highlighting important contributions, and identifying specific challenges. A short discussion is presented following the detailed review of each category to discuss its achievements and future potential. A comparison among DL-based methods for thoracic multi-organ segmentation using a standard benchmark dataset, the 2017 AAPM Thoracic Auto-segmentation Challenge, is presented.
