Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Multi-Layer Dense Attention Decoder for Polyp Segmentation

Krushi Patel, Fengjun Li, Guanghui Wang

TL;DR

This work proposes a novel decoder architecture aimed at hierarchically aggregating locally enhanced multi-level dense features and introduces a novel module named Dense Attention Gate (DAG), which adaptively fuses all previous layers’ features to establish local feature relations among all layers.

Abstract

Detecting and segmenting polyps is crucial for expediting the diagnosis of colon cancer. This is a challenging task due to the large variations of polyps in color, texture, and lighting conditions, along with subtle differences between the polyp and its surrounding area. Recently, vision Transformers have shown robust abilities in modeling global context for polyp segmentation. However, they face two major limitations: the inability to learn local relations among multi-level layers and inadequate feature aggregation in the decoder. To address these issues, we propose a novel decoder architecture aimed at hierarchically aggregating locally enhanced multi-level dense features. Specifically, we introduce a novel module named Dense Attention Gate (DAG), which adaptively fuses all previous layers' features to establish local feature relations among all layers. Furthermore, we propose a novel nested decoder architecture that hierarchically aggregates decoder features, thereby enhancing semantic features. We incorporate our novel dense decoder with the PVT backbone network and conduct evaluations on five polyp segmentation datasets: Kvasir, CVC-300, CVC-ColonDB, CVC-ClinicDB, and ETIS. Our experiments and comparisons with nine competing segmentation models demonstrate that the proposed architecture achieves state-of-the-art performance and outperforms the previous models on four datasets. The source code is available at: https://github.com/krushi1992/Dense-Decoder.

Multi-Layer Dense Attention Decoder for Polyp Segmentation

TL;DR

This work proposes a novel decoder architecture aimed at hierarchically aggregating locally enhanced multi-level dense features and introduces a novel module named Dense Attention Gate (DAG), which adaptively fuses all previous layers’ features to establish local feature relations among all layers.

Abstract

Detecting and segmenting polyps is crucial for expediting the diagnosis of colon cancer. This is a challenging task due to the large variations of polyps in color, texture, and lighting conditions, along with subtle differences between the polyp and its surrounding area. Recently, vision Transformers have shown robust abilities in modeling global context for polyp segmentation. However, they face two major limitations: the inability to learn local relations among multi-level layers and inadequate feature aggregation in the decoder. To address these issues, we propose a novel decoder architecture aimed at hierarchically aggregating locally enhanced multi-level dense features. Specifically, we introduce a novel module named Dense Attention Gate (DAG), which adaptively fuses all previous layers' features to establish local feature relations among all layers. Furthermore, we propose a novel nested decoder architecture that hierarchically aggregates decoder features, thereby enhancing semantic features. We incorporate our novel dense decoder with the PVT backbone network and conduct evaluations on five polyp segmentation datasets: Kvasir, CVC-300, CVC-ColonDB, CVC-ClinicDB, and ETIS. Our experiments and comparisons with nine competing segmentation models demonstrate that the proposed architecture achieves state-of-the-art performance and outperforms the previous models on four datasets. The source code is available at: https://github.com/krushi1992/Dense-Decoder.
Paper Structure (14 sections, 12 equations, 2 figures, 3 tables)

This paper contains 14 sections, 12 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Method
  4. Transformer -- Encoder
  5. Multi-layer Dense Decoder
  6. Loss Function
  7. Experiments
  8. Datasets and Models
  9. Evaluation Metric
  10. Implementation Details
  11. Learning Ability
  12. Generalization Ability
  13. Ablation Study
  14. Conclusion

Figures (2)

  • Figure 1: The overall architecture of the multi-layer dense decoder-based polyp segmentation network. It comprises a PVT encoder represented by a collection of orange blocks. The dense decoder consists of multiple horizontally stacked decoding layers, each composed of a series of decoding blocks interconnected, denoted as green blocks. Each decoding block encompasses two modules: DAG (Dense Attention Gate) and CAM (Camouflage Attention Module). The block diagrams of both modules are depicted in the figure outlined within the green box.
  • Figure 2: Different types of decoding structures for the polyp segmentation. (a) U-shaped decoding structure ronneberger2015u. (b) U-shaped decoding structure with previous layer integration strategy patel2021enhanced. (c) Separate decoding for low-level and high-level features dong2021polyp. (d) multi-layer dense decoder architecture proposed on our model.