Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CAD-Unet: A Capsule Network-Enhanced Unet Architecture for Accurate Segmentation of COVID-19 Lung Infections from CT Images

Yijie Dang, Weijun Ma, Xiaohu Luo, Huaizhu Wang

TL;DR

This paper addresses the challenge of segmenting COVID-19 lung infections in CT images, where boundaries are often indistinct and lesions vary in shape and size. It proposes CAD-Unet, a capsule-augmented Unet that runs a capsule encoder path in parallel with a Unet encoder, coupled to fuse features, and uses dual decoders plus attention gates and a hybrid loss to boost boundary precision. Empirical results on four public datasets show CAD-Unet achieving state-of-the-art or competitive performance for both binary and multi-class segmentation, with significant improvements in key metrics and solid ablation support for its components. The approach offers a practical, efficient framework for accurate lesion segmentation that can aid clinical decision-making and is accompanied by open-source code.

Abstract

Since the outbreak of the COVID-19 pandemic in 2019, medical imaging has emerged as a primary modality for diagnosing COVID-19 pneumonia. In clinical settings, the segmentation of lung infections from computed tomography images enables rapid and accurate quantification and diagnosis of COVID-19. Segmentation of COVID-19 infections in the lungs poses a formidable challenge, primarily due to the indistinct boundaries and limited contrast presented by ground glass opacity manifestations. Moreover, the confounding similarity between infiltrates, lung tissues, and lung walls further complicates this segmentation task. To address these challenges, this paper introduces a novel deep network architecture, called CAD-Unet, for segmenting COVID-19 lung infections. In this architecture, capsule networks are incorporated into the existing Unet framework. Capsule networks represent a novel network architecture that differs from traditional convolutional neural networks. They utilize vectors for information transfer among capsules, facilitating the extraction of intricate lesion spatial information. Additionally, we design a capsule encoder path and establish a coupling path between the unet encoder and the capsule encoder. This design maximizes the complementary advantages of both network structures while achieving efficient information fusion. \noindent Finally, extensive experiments are conducted on four publicly available datasets, encompassing binary segmentation tasks and multi-class segmentation tasks. The experimental results demonstrate the superior segmentation performance of the proposed model. The code has been released at: https://github.com/AmanoTooko-jie/CAD-Unet.

CAD-Unet: A Capsule Network-Enhanced Unet Architecture for Accurate Segmentation of COVID-19 Lung Infections from CT Images

TL;DR

This paper addresses the challenge of segmenting COVID-19 lung infections in CT images, where boundaries are often indistinct and lesions vary in shape and size. It proposes CAD-Unet, a capsule-augmented Unet that runs a capsule encoder path in parallel with a Unet encoder, coupled to fuse features, and uses dual decoders plus attention gates and a hybrid loss to boost boundary precision. Empirical results on four public datasets show CAD-Unet achieving state-of-the-art or competitive performance for both binary and multi-class segmentation, with significant improvements in key metrics and solid ablation support for its components. The approach offers a practical, efficient framework for accurate lesion segmentation that can aid clinical decision-making and is accompanied by open-source code.

Abstract

Since the outbreak of the COVID-19 pandemic in 2019, medical imaging has emerged as a primary modality for diagnosing COVID-19 pneumonia. In clinical settings, the segmentation of lung infections from computed tomography images enables rapid and accurate quantification and diagnosis of COVID-19. Segmentation of COVID-19 infections in the lungs poses a formidable challenge, primarily due to the indistinct boundaries and limited contrast presented by ground glass opacity manifestations. Moreover, the confounding similarity between infiltrates, lung tissues, and lung walls further complicates this segmentation task. To address these challenges, this paper introduces a novel deep network architecture, called CAD-Unet, for segmenting COVID-19 lung infections. In this architecture, capsule networks are incorporated into the existing Unet framework. Capsule networks represent a novel network architecture that differs from traditional convolutional neural networks. They utilize vectors for information transfer among capsules, facilitating the extraction of intricate lesion spatial information. Additionally, we design a capsule encoder path and establish a coupling path between the unet encoder and the capsule encoder. This design maximizes the complementary advantages of both network structures while achieving efficient information fusion. \noindent Finally, extensive experiments are conducted on four publicly available datasets, encompassing binary segmentation tasks and multi-class segmentation tasks. The experimental results demonstrate the superior segmentation performance of the proposed model. The code has been released at: https://github.com/AmanoTooko-jie/CAD-Unet.

Paper Structure

This paper contains 24 sections, 4 equations, 8 figures, 12 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Segmentation architecture based on Unet
  4. Capsule networks for object segmentation
  5. The proposed approach
  6. Network architecture
  7. Convolutional capsule layer
  8. Two parallel encoder pathways
  9. Attention gate
  10. Dual decoders
  11. Hybrid loss function
  12. Experiments and results
  13. Dataset
  14. Evaluation metrics
  15. Experimental setup
  16. ...and 9 more sections

Figures (8)

  • Figure 1: Detailed structure of the proposed CAD-Unet architecture.
  • Figure 2: Schematic description of convolutional capsule layer
  • Figure 3: The structure of ResBlock
  • Figure 4: Attention gate block, where $g$ is the gating signal and $x$ is the input feature maps. $A$ is the spatial attention obtained, which is applied to all channels of the input feature maps $x$.
  • Figure 5: Visual comparison of a binary segmentation model trained with different segmentation architectures for binary COVID-19 segmentation using dataset 2, dataset 3 and dataset 4.
  • ...and 3 more figures