ESTRO 2023 - Abstract Book

S1366

Digital Posters

ESTRO 2023

in the medical product “AccuLearning AI model training platform” (Manteia Medical Technologies). The data set were divided into the training set, validation set, and testing set at a ratio of 8:1:1. The Dice Similarity Coefficient (DSC) and Hausdorff Distance (HD) were calculated for quantitative evaluation. Results High correlations were observed between automatic and manual contours. Based on the results of the independent test group, the mean DSC and 95th HD for the models with best performance were 0.918/8.102 mm for BCS_Left-side (Segresnet), 0.917/13.999 mm for BCS_Right-side (Unet), 0.835/7.986 mm for BRM_Left-side (Flexnet), 0.841/9.993 mm for BRM_Right-side (Unet), 0.881/6.892 mm for cervical_pre (Segresnet), 0.847/9.236 mm for cervical_post (Flexnet), and 0.924/9.206 mm for rectal carcinomas (Segresnet).

Conclusion