RT-Mind自动勾画技术应用于鼻咽癌放射治疗可行性研究

目的:探讨RT-Mind软件在鼻咽癌放疗临床靶区（CTV）、危及器官（OARs）自动勾画的可行性，为临床应用提供依据。方法:回顾性选取28例应用调强技术治疗的鼻咽癌患者，将放疗医师手动勾画CTV、OARs（脑干、脊髓、左右晶体、视交叉、左右视神经、左右腮腺、左右颞叶、左右颞颌关节、下颌骨）做为参考标准，再使用RT-Mind软件自动勾画CTV及OARs。对比手动与自动勾画在CTV和OARs区域的Dice相似性系数（DSC）、Jaccard系数（JAC）、敏感性指数（SI）、包容性系数（lncI）、质心偏差（DC）、Hausdorff距离（HD）等参数，从而评估自动勾画效果。结果:CTV的DSC、JAC、SI、lncl、DC、HD分别为:0.78±0.04、0.70±0.05、0.85±0.08、0.87±0.04、（7.76±5.03） mm、（12.3±1.16） mm，OARs中DC、HD值均在1 cm之内。结论:RT-Mind软件能够基本满足临床要求，能够较为准确地实现鼻咽癌患者CTV和OARs的自动勾画。由于病人既往病史的个体差异，放射治疗医师必须根据临床需要，对自动勾画的CTV和OARs进行修改后，才能用于治疗。若依据临床需求进一步完善自动勾画的个性化定制，相信RT-Mind软件在解剖结构复杂的鼻咽癌放疗中能够辅助放疗医生提高工作效率，更好地为患者服务。

Feasibility of RT-Mind auto-segmentation technique in radiotherapy for nasopharyngeal carcinoma

Abstract: Objective To explore the feasibility of RT-Mind software in the automatic segmentations of clinical target volume (CTV) and organs-at-risk (OAR) for radiotherapy for nasopharyngeal carcinoma, so as to provide a basis for its clinical application. Methods Twenty-eight cases of nasopharyngeal carcinoma patients treated with radiotherapy were enrolled retrospectively. Taking manual segementations of CTV and OAR (brain stem, spinal cord, lens, optic chiasm, optic nerves, parotid glands, temporal lobes, temporomandibular joints, and mandible) by radiotherapists as reference standards, RT-Mind software was used to carry out the auto-segmentations of CTV and OAR. The Dice similarity coefficient (DSC), Jaccard index (JAC), sensitivity index (SI), inclusiveness index (Incl), deviation of centroid (DC), and Hausdorff distance (HD) between manual and automatic segmentations of CTV and OAR were calculated for evaluating the performance of automatic segmentation. Results The DSC, JAC, SI, Incl, DC, and HD of CTV were 0.78±0.04, 0.70±0.05, 0.85±0.08, 0.87±0.04, (7.76±5.03) mm, and (12.3±1.16) mm, respectively. The DC and HD of OAR were all less than 1 cm. Conclusion RT-Mind software can basically meet the clinical requirements, and can more accurately realize the automatic segementations of OAR and CTV in patients with nasopharyngeal carcinoma. Due to individual differences in the patients medical history, radiotherapy physicians need to make certain modifications to the auto-segmentation results of OAR and CTV according to clinical needs. If the personalized customization of auto-segmentation can be further improved according to clinical requirements, RT-Mind software can assist doctors in improving their work efficiency in the radiotherapy of nasopharyngeal carcinoma with complex anatomical structures, and serving patients better.