MNet-SAt: A Multiscale Network with Spatial-enhanced Attention for Segmentation of Polyps in Colonoscopy

TL;DR

MNet-SAt introduces a novel encoder–decoder network for polyp segmentation that blends edge-guided refinement, multiscale feature aggregation, spatial-enhanced attention, and channel-aware ASPP. The Edge-Guided Feature Enrichment units preserve fine boundary details, while the Hybrid Multi-Scale Attention module learns robust spatial-global dependencies across scales. The Channel-Enhanced ASPP further recalibrates multiscale features, yielding state-of-the-art performance on Kvasir-SEG and CVC-ClinicDB with DSCs of 96.61% and 98.60%, respectively. Comprehensive ablations confirm the contributions of EGFE, MSFA, SEAt, and CE-ASPP, and cross-dataset tests demonstrate good generalization, underscoring potential clinical impact for early polyp detection and CRC mortality reduction.

Abstract

Objective: To develop a novel deep learning framework for the automated segmentation of colonic polyps in colonoscopy images, overcoming the limitations of current approaches in preserving precise polyp boundaries, incorporating multi-scale features, and modeling spatial dependencies that accurately reflect the intricate and diverse morphology of polyps. Methods: To address these limitations, we propose a novel Multiscale Network with Spatial-enhanced Attention (MNet-SAt) for polyp segmentation in colonoscopy images. This framework incorporates four key modules: Edge-Guided Feature Enrichment (EGFE) preserves edge information for improved boundary quality; Multi-Scale Feature Aggregator (MSFA) extracts and aggregates multi-scale features across channel spatial dimensions, focusing on salient regions; Spatial-Enhanced Attention (SEAt) captures spatial-aware global dependencies within the multi-scale aggregated features, emphasizing the region of interest; and Channel-Enhanced Atrous Spatial Pyramid Pooling (CE-ASPP) resamples and recalibrates attentive features across scales. Results: We evaluated MNet-SAt on the Kvasir-SEG and CVC-ClinicDB datasets, achieving Dice Similarity Coefficients of 96.61% and 98.60%, respectively. Conclusion: Both quantitative (DSC) and qualitative assessments highlight MNet-SAt's superior performance and generalization capabilities compared to existing methods. Significance: MNet-SAt's high accuracy in polyp segmentation holds promise for improving clinical workflows in early polyp detection and more effective treatment, contributing to reduced colorectal cancer mortality rates.

Figures (7)

• Figure 1: Polyp patient demographics, image quality issues, and size distribution
• Figure 2: MNet-SAt framework for polyp segmentation employs a U-shaped architecture with an encoder and decoder linked by skip connections. Edge-Guided Feature Enrichment (EGFE) units within each stage enable feature transfer by extracting features from polyp anatomy, enhancing the segmentation task. A Hybrid Multi-Scale Attention (HMAtt) module bridges the encoder and decoder, learning multi-scale features by prioritizing important regions.
• Figure 3: Edge maps of the Sobel operator in x-y directions
• Figure 4: Multi-Head Spatial-Enhanced Attention
• Figure 5: Channel-Enhanced Atrous Spatial Pyramid Pooling