Tumor
Effects of deep learning- versus atlas-based automatic contouring methods on the contouring of organs-at-risk in rectal cancer
Yucheng Li, Cheng Wang, Yongshi Jia, Jianming Tang, Wenming Zhan, Qiang Li, Lingyun Qiu, Weijun Chen
Published 2021-10-15
Cite as Chin J Prim Med Pharm, 2021, 28(10): 1490-1495. DOI: 10.3760/cma.issn1008-6706.2021.10.011
Abstract
ObjectiveTo investigate the effects of deep learning-based AiContour®©versus atlas-based Raystation®© automatic contouring methods on the contouring of organs-at-risk on CT images of patients with rectal cancer who undergo radiotherapy, providing evidence for clinical application.
MethodsFifty patients with rectal cancer who received treatment during January to June 2020 in Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College) were included in this study. The CT images from 20 patients with rectal cancer that had been contoured by experienced radiotherapist were selected as target images and automatically contoured using the data template library of AiContour®© and Raystation®© automatic contouring methods. Hausdorff distance, mean distance to agreement, dice similarity coefficient, Jaccard coefficient were used to quantitatively evaluate the accuracy of the volume of contour of organs-at-risk automatically sketched by the two methods.
ResultsThere was no significant difference in Hausdorff distance in left femoral head [(6.81 ± 2.66) vs. (7.24 ± 2.10)], right femoral head [(7.38 ± 3.91) vs. (8.14 ± 3.71)], pelvis [(24.00 ± 9.01) vs. (24.66 ± 9.67)] between AiContour®© and Raystation®© automatic contouring methods (tleft femoral head = -0.831, tright femoral head = -0.821, tpelvis = -0.357, all P > 0.05). Significant differences were observed in mean distance to agreement, dice similarity coefficient and Jaccard coefficient of organs-at-risk (all P < 0.05). The mean values of dice similarity coefficient automatically sketched by AiContour®© method were > 0.7. The DSC of left kidney, right kidney, rectum and bladder automatically sketched by Raystation®© method were < 0.7, and the dice similarity coefficient values of other organs-at-risk automatically sketched by Raystation®© method were > 0.7. In addition, Hausdorff distance, mean distance to agreement and Jaccard coefficient values of organs-at-risk automatically sketched by AiContour®© method were superior to those automatically sketched by Raystation®©.
ConclusionAfter slight modification, the organs-at-risk automatically sketched by AiContour®© and Raystation®© methods can meet clinical requirement. The contouring effects provided byAiContour®© method were superior to those provided by Raystation®© method.
Key words:
Rectal neoplasms; Radiotherapy; Organs at risk; Deep learning; Automatic sketching; Hausdorff index; Mean distance index; Jaccard index; Dice Similarity Coefficient
Contributor Information
Yucheng Li
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China
Cheng Wang
School of Nuclear Science and Technology, University of South China, Hengyang 421000, Hunan Province, China
Yongshi Jia
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China
Jianming Tang
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China
Wenming Zhan
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China
Qiang Li
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China
Lingyun Qiu
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China
Weijun Chen
Radiotherapy Center, Zhejiang Provincial People's Hospital (Affiliated Hospital of Hangzhou Medical College), Hangzhou 310014, Zhejiang Province, China