摘要
目的 评估 3种深度学习(DL)算法在子宫内膜癌(EC)术后患者高剂量率近距离放射治疗(high-dose-rate bra-chytherapy,HDR BT)中,自动分割临床靶区(CTV)的应用结果.方法 数据集由 306 名子宫内膜癌术后患者的计算机断层扫描(CT)图像组成,按比例分为训练集(246例)、验证集(30例)和测试集(30例).比较 3种深度卷积神经网络模型(3D U-Net、3D Res U-Net和V-Net)在CTV分割上的性能.采用定量指标分别为戴斯相似性系数(DSC)、豪斯多夫距离(HD)、豪斯多夫距离第 95 百分位数(HD95%)和交并比(IoU).结果 在测试阶段中,3D U-Net、3D Res U-Net和V-Net分割CTV得到的DSC平均值分别为 0.90±0.07、0.95±0.06 和 0.95±0.06;HD平均值(mm)分别为2.51±1.70、0.96±1.01和 0.98±0.95;HD95%平均值(mm)分别为 1.33±1.02、0.65±0.91和 0.40±0.72,IoU平均值分别为 0.85±0.11、0.91±0.09 和 0.92±0.09.其中,V-Net分割结果与高级临床医生勾画结果更接近,CTV的分割时间<3.2 s,节省了临床医生的工作时间.结论 V-Net在CTV分割方面表现最佳,定量指标和临床评估均优于其他模型.该方法与基准真实值高度一致,有效减少医生间差异,缩短治疗时间.
Abstract
Objective To evaluate the application of three deep learning algorithms in automatic segmentation of clinical target volumes(CTVs)in high-dose-rate brachytherapy after surgery for endometrial carcinoma.Methods A dataset com-prising computed tomography scans from 306 post-surgery patients with endometrial carcinoma was divided into three sub-sets:246 cases for training,30 cases for validation,and 30 cases for testing.Three deep convolutional neural network mod-els,3D U-Net,3D Res U-Net,and V-Net,were compared for CTV segmentation.Several commonly used quantitative met-rics were employed,i.e.,Dice similarity coefficient,Hausdorff distance,95th percentile of Hausdorff distance,and Intersec-tion over Union.Results During the testing phase,CTV segmentation with 3D U-Net,3D Res U-Net,and V-Net showed a mean Dice similarity coefficient of 0.90±0.07,0.95±0.06,and 0.95±0.06,a mean Hausdorff distance of 2.51±1.70,0.96±1.01,and 0.98±0.95 mm,a mean 95th percentile of Hausdorff distance of 1.33±1.02,0.65±0.91,and 0.40±0.72 mm,and a mean Intersection over Union of 0.85±0.11,0.91±0.09,and 0.92±0.09,respectively.Segmentation based on V-Net was similarly to that performed by experienced radiation oncologists.The CTV segmentation time was<3.2 s,which could save the work time of clinicians.Conclusion V-Net is better than other models in CTV segmentation as indicated by quantitative metrics and clinician assessment.Additionally,the method is highly consistent with the ground truth,reducing inter-doctor variability and treatment time.
NETL
NSTL
万方数据