Cross-level Feature Aggregation Network for Polyp Segmentation

Tao Zhou, Yi Zhou, Kelei He, Chen Gong, Jian Yang, Huazhu Fu, Dinggang Shen

Abstract: Highlights•We propose a novel Cross-level Feature Aggregated Network to simultaneously exploit boundary information and capture hierarchical semantic information for accurate segmenting polyps.•A Cross-level Feature Fusion module is proposed to fully utilize the features from adjacent layers, which conducts cross-level feature fusion at different scales to deal with scale variations.•We propose a Boundary Aggregated Module to capture the boundary context information and then incorporate them into the polyp segmentation network.•Extensive experimental results show the effectiveness of the proposed method.