Automatic Detection and Segmentation of Liver Tumors in Multi- Phase CT Images by Phase Attention Mask R-CNN

Ryo Hasegawa,Yutaro Iwamoto,Xianhua Han,Lanfen Lin,Hongjie Hu,Xiujun Cai,Yen-Wei Chen

DOI: https://doi.org/10.1109/icce50685.2021.9427760

2020-01-01

Abstract:In computer-aided diagnosis of liver tumors, tumor detection and segmentation are essential preprocessing steps. In this study, we proposed a method based on mask R-CNN to detect and segment the liver tumor in a multi-phase CT image simultaneously. Triple phase images are used as three different channels just like a color RGB image. The segmentation accuracy (Dice value) is about 0.60 ~ 0.66 for a single phase CT image, while the accuracy can be improved to about 0.73 by using the multi-phase CT image.