Background : This study was designed to build a automated segmentation of biliary structure based on magnetic resonance cholangiopancreatography using deep learning model.

Methods : Living liver donors with magnetic resonance cholangiopancreatography gradient and spine echo technique and underwent three-dimensional modeling were eligible for this study. The three-dimensional residual U-Net model was implemented for deep learning process. Training set and test set were allocated in 9:1 ratio. For evaluation of the performance, dice similarity coefficient score was evaluated between the ground truth and inference of the auto-segmentation model.

Results : A total of 250 cases were included to the study. There was no difference in the baseline characteristics between the train set (n=225) and test set. (n=25) The overall mean dice similarity coefficient was 0.80 ± 0.20 between the ground truth and inference result. The qualitative assessment of the model showed relatively high accuracy especially for common bile duct (88%), common hepatic duct (92%), hilum (96%), right hepatic duct (100%) and left hepatic duct (96%), while the third order branch of right hepatic duct (18.2%) showed low accuracy.

Conclusions : The auto-segmentation model of biliary structure based on magnetic resonance cholangiopancreatography using deep learning method showed high performance and shows promising results for future development of automation.