Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Radiation Oncology, Cancer Center Amsterdam, Amsterdam, the Netherlands.
Amsterdam UMC, Vrije Universiteit Amsterdam, Department of Radiation Oncology, Cancer Center Amsterdam, Amsterdam, the Netherlands.
Int J Radiat Oncol Biol Phys. 2019 Jul 1;104(3):677-684. doi: 10.1016/j.ijrobp.2019.02.040. Epub 2019 Mar 2.
Organ-at-risk (OAR) delineation is a key step in treatment planning but can be time consuming, resource intensive, subject to variability, and dependent on anatomical knowledge. We studied deep learning (DL) for automated delineation of multiple OARs; in addition to geometric evaluation, the dosimetric impact of using DL contours for treatment planning was investigated.
The following OARs were delineated with DL developed in-house: both submandibular and parotid glands, larynx, cricopharynx, pharyngeal constrictor muscle (PCM), upper esophageal sphincter, brain stem, oral cavity, and esophagus. DL contours were benchmarked against the manual delineation (MD) clinical contours using the Sørensen-Dice similarity coefficient. Automated knowledge-based treatment plans were used. The mean dose to the manually delineated OAR structures was reported for the MD and DL plans.
DL delineation of all OARs took <10 seconds per patient. For 7 of 11 OARs, the average Sørensen-Dice similarity coefficient was good (0.78-0.83). However, performance was lower for the esophagus (0.60), brainstem (0.64), PCM (0.68), and cricopharynx (0.73), often because of variations in MD. Although the average dose was statistically significantly higher in the DL plans for the inferior PCM (1.4 Gy) and esophagus (2.2 Gy), these average differences were not clinically significant. Dose to 28 of 209 (13.4%) and 7 of 209 (3.3%) OARs was >2 Gy higher and >2 Gy lower, respectively, in the DL plans.
DL-based segmentation for head and neck OARs is fast; for most organs and most patients, it performs sufficiently well for treatment-planning purposes. It has the potential to increase efficiency and facilitate online adaptive radiation therapy.
危及器官（OAR）勾画是治疗计划制定的关键步骤，但可能耗时、资源密集、存在变异性且依赖于解剖学知识。我们研究了深度学习（DL）在自动勾画多个 OAR 中的应用；除了几何评估外，还研究了使用 DL 轮廓进行治疗计划的剂量学影响。
使用内部开发的 DL 对以下 OAR 进行了勾画：双侧下颌腺和腮腺、喉、环咽肌、咽缩肌、食管上括约肌、脑干、口腔和食管。使用 Sørensen-Dice 相似系数将 DL 轮廓与手动勾画（MD）临床轮廓进行了基准测试。使用自动基于知识的治疗计划。报告了 MD 和 DL 计划中手动勾画 OAR 结构的平均剂量。
每位患者的 DL 勾画时间不到 10 秒。在 11 个 OAR 中有 7 个的平均 Sørensen-Dice 相似系数较好（0.78-0.83）。然而，对于食管（0.60）、脑干（0.64）、咽缩肌（0.68）和环咽肌（0.73），性能较低，这通常是由于 MD 的差异所致。尽管在 DL 计划中，下咽缩肌（1.4Gy）和食管（2.2Gy）的平均剂量统计上显著较高，但这些平均差异没有临床意义。在 DL 计划中，有 28 个（13.4%）和 7 个（3.3%）OAR 的剂量分别高 2Gy 以上和低 2Gy 以上。
基于 DL 的头颈部 OAR 分割速度快；对于大多数器官和大多数患者，其性能足以满足治疗计划的目的。它有可能提高效率并促进在线自适应放疗。
