Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

ACS-SegNet: An Attention-Based CNN-SegFormer Segmentation Network for Tissue Segmentation in Histopathology

Nima Torbati, Anastasia Meshcheryakova, Ramona Woitek, Diana Mechtcheriakova, Amirreza Mahbod

TL;DR

ACS-SegNet tackles semantic tissue segmentation in histopathology by integrating CNN and ViT features in a dual-encoder framework, fused with Convolutional Block Attention Modules (CBAM). By combining a SegFormer encoder (ViT) with a ResNet encoder (CNN) and a UNet-like decoder, the method captures both global context and local details. On GCPS and PUMA, ACS-SegNet achieves top-tier performance with IoU and Dice metrics (e.g., 76.79% IoU / 86.87% Dice on GCPS and 64.93% IoU / 76.60% Dice on PUMA), and an ablation confirms the benefit of CBAM. The approach is parameter-efficient (~50M params) compared with state-of-the-art hybrids, and code is publicly available, enabling practical adoption in histopathology workflows.

Abstract

Automated histopathological image analysis plays a vital role in computer-aided diagnosis of various diseases. Among developed algorithms, deep learning-based approaches have demonstrated excellent performance in multiple tasks, including semantic tissue segmentation in histological images. In this study, we propose a novel approach based on attention-driven feature fusion of convolutional neural networks (CNNs) and vision transformers (ViTs) within a unified dual-encoder model to improve semantic segmentation performance. Evaluation on two publicly available datasets showed that our model achieved μIoU/μDice scores of 76.79%/86.87% on the GCPS dataset and 64.93%/76.60% on the PUMA dataset, outperforming state-of-the-art and baseline benchmarks. The implementation of our method is publicly available in a GitHub repository: https://github.com/NimaTorbati/ACS-SegNet

ACS-SegNet: An Attention-Based CNN-SegFormer Segmentation Network for Tissue Segmentation in Histopathology

TL;DR

ACS-SegNet tackles semantic tissue segmentation in histopathology by integrating CNN and ViT features in a dual-encoder framework, fused with Convolutional Block Attention Modules (CBAM). By combining a SegFormer encoder (ViT) with a ResNet encoder (CNN) and a UNet-like decoder, the method captures both global context and local details. On GCPS and PUMA, ACS-SegNet achieves top-tier performance with IoU and Dice metrics (e.g., 76.79% IoU / 86.87% Dice on GCPS and 64.93% IoU / 76.60% Dice on PUMA), and an ablation confirms the benefit of CBAM. The approach is parameter-efficient (~50M params) compared with state-of-the-art hybrids, and code is publicly available, enabling practical adoption in histopathology workflows.

Abstract

Automated histopathological image analysis plays a vital role in computer-aided diagnosis of various diseases. Among developed algorithms, deep learning-based approaches have demonstrated excellent performance in multiple tasks, including semantic tissue segmentation in histological images. In this study, we propose a novel approach based on attention-driven feature fusion of convolutional neural networks (CNNs) and vision transformers (ViTs) within a unified dual-encoder model to improve semantic segmentation performance. Evaluation on two publicly available datasets showed that our model achieved μIoU/μDice scores of 76.79%/86.87% on the GCPS dataset and 64.93%/76.60% on the PUMA dataset, outperforming state-of-the-art and baseline benchmarks. The implementation of our method is publicly available in a GitHub repository: https://github.com/NimaTorbati/ACS-SegNet
Paper Structure (12 sections, 3 equations, 2 figures, 3 tables)

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

Table of Contents

  1. Introduction
  2. Method
  3. Model
  4. Datasets
  5. GCPS
  6. PUMA
  7. Experiments
  8. Results & Discussion
  9. GCPS Results
  10. PUMA Results
  11. Combined Results
  12. Conclusion

Figures (2)

  • Figure 1: Block diagram of the proposed model. The model consists of two parallel encoders: a SegFormer encoder and a ResNet encoder. Features from the four stages of the SegFormer encoder are first upsampled and concatenated (C. blocks) with the corresponding ResNet encoder features, and then fused through the Convolutional Block Attention Module (CBAM). The fourth block of the ResNet encoder serves as the bottleneck. For each path, the feature maps are illustrated with their corresponding dimensions. Input Image: (digital image of) H&E-stained tissue specimen of melanoma cancer.
  • Figure 2: Example of tissue segmentation results of the proposed method from the GCPS dataset (top) and the PUMA dataset (bottom). Tumor (cancerous) regions are shown in red, blood vessels in green, stroma in blue, and non-cancerous regions in black. Input image: (digital image of) H&E-stained tissue specimens of gastric cancer (top) and of melanoma (bottom).