Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gland Segmentation Via Dual Encoders and Boundary-Enhanced Attention

Huadeng Wang, Jiejiang Yu, Bingbing Li, Xipeng Pan, Zhenbing Liu, Rushi Lan, Xiaonan Luo

TL;DR

This work tackles automated gland segmentation in histopathology by addressing deformable malignant glands and tightly adherent glands. It introduces DEA-Net, a dual-encoder architecture with a Local Semantic Guided Encoder, a main backbone encoder, a Feature Fusion Module, a Deep Feature Decoder, and a Boundary-Enhanced Attention mechanism to recover gland boundaries. Across the GlaS and CRAG datasets, the method achieves state-of-the-art F1 and Dice scores with lower Hausdorff distances, notably improving segmentation of irregular and adherent glands. The approach demonstrates that targeted boundary emphasis and multi-scale feature fusion can substantially enhance gland segmentation accuracy, with potential to aid computer-aided diagnosis in colorectal cancer.

Abstract

Accurate and automated gland segmentation on pathological images can assist pathologists in diagnosing the malignancy of colorectal adenocarcinoma. However, due to various gland shapes, severe deformation of malignant glands, and overlapping adhesions between glands. Gland segmentation has always been very challenging. To address these problems, we propose a DEA model. This model consists of two branches: the backbone encoding and decoding network and the local semantic extraction network. The backbone encoding and decoding network extracts advanced Semantic features, uses the proposed feature decoder to restore feature space information, and then enhances the boundary features of the gland through boundary enhancement attention. The local semantic extraction network uses the pre-trained DeepLabv3+ as a Local semantic-guided encoder to realize the extraction of edge features. Experimental results on two public datasets, GlaS and CRAG, confirm that the performance of our method is better than other gland segmentation methods.

Gland Segmentation Via Dual Encoders and Boundary-Enhanced Attention

TL;DR

This work tackles automated gland segmentation in histopathology by addressing deformable malignant glands and tightly adherent glands. It introduces DEA-Net, a dual-encoder architecture with a Local Semantic Guided Encoder, a main backbone encoder, a Feature Fusion Module, a Deep Feature Decoder, and a Boundary-Enhanced Attention mechanism to recover gland boundaries. Across the GlaS and CRAG datasets, the method achieves state-of-the-art F1 and Dice scores with lower Hausdorff distances, notably improving segmentation of irregular and adherent glands. The approach demonstrates that targeted boundary emphasis and multi-scale feature fusion can substantially enhance gland segmentation accuracy, with potential to aid computer-aided diagnosis in colorectal cancer.

Abstract

Accurate and automated gland segmentation on pathological images can assist pathologists in diagnosing the malignancy of colorectal adenocarcinoma. However, due to various gland shapes, severe deformation of malignant glands, and overlapping adhesions between glands. Gland segmentation has always been very challenging. To address these problems, we propose a DEA model. This model consists of two branches: the backbone encoding and decoding network and the local semantic extraction network. The backbone encoding and decoding network extracts advanced Semantic features, uses the proposed feature decoder to restore feature space information, and then enhances the boundary features of the gland through boundary enhancement attention. The local semantic extraction network uses the pre-trained DeepLabv3+ as a Local semantic-guided encoder to realize the extraction of edge features. Experimental results on two public datasets, GlaS and CRAG, confirm that the performance of our method is better than other gland segmentation methods.
Paper Structure (11 sections, 12 equations, 3 figures, 2 tables)

This paper contains 11 sections, 12 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. proposed method
  3. Feature Fusion Method
  4. Deep Feature Decoder Module
  5. Boundary Enhanced Attention Mechanism
  6. Experiments and analysis
  7. Datasets
  8. Implementation Details and Evaluation Criteriat
  9. Ablation Experiment
  10. Performance Comparison Of Different Methods
  11. conclusion

Figures (3)

  • Figure 1: Morphological structure of glands. (a) The green framed part is the cohesive gland, (b) The orange framed part is the severely deformed malignant gland, (c) The red framed part is the gland with different shapes.
  • Figure 2: DEA-Net Architecture diagram
  • Figure 3: (a) Backbone segmentation result, (b) Backbone+LD segmentation result, (c) Backbone+LD+FFM segmentation result, (d) Backbone+LD+FFM+DFB segmentation result, (e) DEA-Net segmentation result