AG-CUResNeSt: A Novel Method for Colon Polyp Segmentation

TL;DR

This paper introduces AG-CUResNeSt, a new neural network architecture that improves colon polyp segmentation accuracy by effectively combining multi-level features using ResNeSt backbone and attention gates, outperforming existing methods.

Contribution

The paper presents a novel neural network architecture, AG-CUResNeSt, that enhances polyp segmentation by integrating ResNeSt backbone with attention mechanisms for better feature fusion.

Findings

Achieves state-of-the-art accuracy on five benchmark datasets.

Effectively handles variations in polyp size, shape, texture, and color.

Outperforms existing segmentation methods.

Abstract

Colorectal cancer is among the most common malignancies and can develop from high-risk colon polyps. Colonoscopy is an effective screening tool to detect and remove polyps, especially in the case of precancerous lesions. However, the missing rate in clinical practice is relatively high due to many factors. The procedure could benefit greatly from using AI models for automatic polyp segmentation, which provide valuable insights for improving colon polyp detection. However, precise segmentation is still challenging due to variations of polyps in size, shape, texture, and color. This paper proposes a novel neural network architecture called AG-CUResNeSt, which enhances Coupled UNets using the robust ResNeSt backbone and attention gates. The network is capable of effectively combining multi-level features to yield accurate polyp segmentation. Experimental results on five popular benchmark datasets show that our proposed method achieves state-of-the-art accuracy compared to existing methods.