Evaluation of Auto-segmentation for External Radiotherapy Planning Structures:A Part of Deep Learning Based Workflow in Cervical Cancer

Abstract Objective:Deep learning (DL) based approach aims to construct a full workflow solution for cervical cancer with external radiotherapy (EBRT) and brachytherapy (BT).The purpose of this study was to evaluate the accuracy of EBRT planning structures derived from DL based auto-segmentation which would be a part implemented in the workflow of cervical cancer.Methods:Clinical target volumes (CTVs) and organs at risk (OARs) of 75 cervical cancer patients were manually delineated by senior radiation oncologists and auto-segmented by DL based method.The accuracy of DL based auto-segmented contours were evaluated using geometric and dosimetric metrics including dice similarity coefficient (DSC),95%hausdorf distance (95%HD),jaccard coefficient (JC) and dose-volume index (DVI).The correlation between geometric metrics and dosimetric difference was performed by Spearman’s correlation analysis. Results:The DL based auto-segmentation generated similar geometric performance in right kidney,left kidney,bladder,right femoral head and left femoral head with mean DSC of 0.88-0.93,95%HD of 1.03-2.96mm and JC of 0.78-0.88. Wilcoxon’s signed-rank test indicated significant dosimetric differences between manual and DL based auto-segmentation in CTV,spinal cord and pelvic bone (P<0.001). A strong correlation between the mean dose of pelvic bone and its 95%HD (R=0.843, P=0.000) was found in Spearman’s correlation analysis, and the remaining structures showed weak link between dosimetric difference and all of geometric metrics.Conclusion:Auto-segmentation achieved a satisfied agreement for most EBRT planning structures,although the dosimetric consistency of CTV was a concern. DL based auto-segmentation was an essential component in cervical cancer workflow which would generate the accurate contouring.