Abstract：Objective To evaluate the dose accuracy of using an automatic delineation framework of normal tissues in head and neck
named Ua-Net in treatment planning, and test the feasibility and safety of its clinical application. Methods Thirty patients with
nasopharyngeal carcinoma were selected retrospectively. 28 organs at risk (OARs) for each patient were delineated manually and
automatically by Ua-Net respectively. Dice similarity coefficient (DSC) was used to assess the accuracy of automatic delineation.
Automatic delineated OARs were used to make treatment plan for each patient, and then the plan was mapped to manual delineated
normal tissues. The dose differences with the same dose distribution between automatic delineation and manual delineation of OARs
was evaluated. Results The average DSC in 17 out of 28 OARs was greater than 0.8, meanwhile the dose differences in 28 out of the
840 OARs were more than 5 Gy. Conclusion In general, the results of automatic delineation by using Ua-Net in treatment planning
is accurate, but it still needs to be manually reviewed for OARs with small size such as hypophysis, optic chiasm, and optic nerve.
陈旭明，刘勇. 基于深度学习的正常组织自动勾画在计划设计中的剂量准确度评估[J]. 中国医疗设备, 2021, 36(10): 169-172.
CHEN Xuming, LIU Yong. Evaluation of the Dose Accuracy in Application of Deep Learning-Based Automatic
Delineation of Normal Tissues in Treatment Planning. China Medical Devices, 2021, 36(10): 169-172.
