Spider-Net: High-Resolution Multi-Scale Attention Network with Full-Attention Decoder for Tumor Segmentation in Kidney, Liver and Pancreas

• DOI: 10.2139/ssrn.4555982
• Publication Date: 2023-09-07T19:36:16Z
• Authors: Yanjun Peng; Xiqing Hu; Xiaobo Hao; Pengcheng Liu; Yanhui Deng; Zhengyu Li

In this paper, the abdominal tumor is a general term for tumors in kidney, liver and pancreas. Accurate segmentation of essential their treatment. However, varying shapes sizes organs result significant differences regions. Existing convolution neural networks (CNNs) can only accurately segment individual tumors, lacking sufficient generalizability. We aim to design network that achieve good results different tumors. To end, we present Spider-net which consists high-resolution multi-scale attention encoder full-attention decoder. Additionally, scale integrates channel spatial designed generating output. have also classification branch distinguish whether segmented region real area or another benign lesion.We train evaluate on three organs: pancreas, liver. achieves state-of-the-art compared methods use CNNs transformers. The proposed provides deep fusion Transformers improve effectiveness generalization network.