Automatic Segmentation of Abdominal Organs with 3D-UNet

Wenhui Lei

Abstract: Automatic segmentation of abdominal organs are of great importance for clinical use. Current deep learning based fully-supervised segmentation methods have achieved promising results on abdominal organs. In this paper, we investigate the performance of a fully-supervised segmentation model with limited labeled data on abdominal organs. The mean DSC, NSD, running time of our method on the FLARE2022 validation set is 0.041, 0.041, 42s, respectively.