BAN: A Boundary-Aware Network for Accurate Colorectal Polyp Segmentation

Abstract

Colonoscopy images exhibit multi-frequency features, with polyp boundaries residing in a mid-frequency range, which are critical for accurate polyp segmentation. However, current deep learning models tend to prioritize low-frequency features, leading to reduced segmentation performance. To address this challenge, we propose a novel boundary-aware network (BAN) that integrates trainable Gabor filters into the polyp segmentation process through a dedicated module called Gabor-driven feature extraction (GFE). By developing and using a trajectory-directed frequency learning approach, Gabor filters are trained along a damping sinusoidal path, dynamically optimizing their frequency parameters within a proper mid-frequency range. This enhances boundary feature representation and significantly improves polyp segmentation accuracy. Extensive experiments demonstrate that our BAN outperforms existing state-of-the-art methods.