Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Towards Interpretable Attention Networks for Cervical Cancer Analysis

Ruiqi Wang, Mohammad Ali Armin, Simon Denman, Lars Petersson, David Ahmedt-Aristizabal

TL;DR

This work tackles interpretable cervical cell classification in multi-cell images by contrasting traditional CNNs (ResNet, DenseNet) with attention-based architectures (Residual Attention Networks, Residual Channel Attention). It employs integrated gradients to produce attribution explanations, and evaluates on the SIPaKMeD dataset, revealing that DenseNet-121 with residual channel attention delivers the best performance while yielding focused, region-specific explanations. The study demonstrates that multi-cell context enhances classification and that channel-attention mechanisms can isolate informative cell groups, offering interpretable insights for clinical use. It also points to future directions involving graph-based representations to capture inter-cell relations within multi-cell images.

Abstract

Recent advances in deep learning have enabled the development of automated frameworks for analysing medical images and signals, including analysis of cervical cancer. Many previous works focus on the analysis of isolated cervical cells, or do not offer sufficient methods to explain and understand how the proposed models reach their classification decisions on multi-cell images. Here, we evaluate various state-of-the-art deep learning models and attention-based frameworks for the classification of images of multiple cervical cells. As we aim to provide interpretable deep learning models to address this task, we also compare their explainability through the visualization of their gradients. We demonstrate the importance of using images that contain multiple cells over using isolated single-cell images. We show the effectiveness of the residual channel attention model for extracting important features from a group of cells, and demonstrate this model's efficiency for this classification task. This work highlights the benefits of channel attention mechanisms in analyzing multiple-cell images for potential relations and distributions within a group of cells. It also provides interpretable models to address the classification of cervical cells.

Towards Interpretable Attention Networks for Cervical Cancer Analysis

TL;DR

This work tackles interpretable cervical cell classification in multi-cell images by contrasting traditional CNNs (ResNet, DenseNet) with attention-based architectures (Residual Attention Networks, Residual Channel Attention). It employs integrated gradients to produce attribution explanations, and evaluates on the SIPaKMeD dataset, revealing that DenseNet-121 with residual channel attention delivers the best performance while yielding focused, region-specific explanations. The study demonstrates that multi-cell context enhances classification and that channel-attention mechanisms can isolate informative cell groups, offering interpretable insights for clinical use. It also points to future directions involving graph-based representations to capture inter-cell relations within multi-cell images.

Abstract

Recent advances in deep learning have enabled the development of automated frameworks for analysing medical images and signals, including analysis of cervical cancer. Many previous works focus on the analysis of isolated cervical cells, or do not offer sufficient methods to explain and understand how the proposed models reach their classification decisions on multi-cell images. Here, we evaluate various state-of-the-art deep learning models and attention-based frameworks for the classification of images of multiple cervical cells. As we aim to provide interpretable deep learning models to address this task, we also compare their explainability through the visualization of their gradients. We demonstrate the importance of using images that contain multiple cells over using isolated single-cell images. We show the effectiveness of the residual channel attention model for extracting important features from a group of cells, and demonstrate this model's efficiency for this classification task. This work highlights the benefits of channel attention mechanisms in analyzing multiple-cell images for potential relations and distributions within a group of cells. It also provides interpretable models to address the classification of cervical cells.

Paper Structure

This paper contains 14 sections, 1 equation, 2 figures, 2 tables.

Table of Contents

  1. INTRODUCTION
  2. METHODOLOGY
  3. Traditional Deep Convolutional Neural Networks
  4. Residual Convolutional Networks (ResNet)
  5. Dense Convolutional Networks (DenseNet)
  6. Attention-based frameworks
  7. Residual Attention Networks
  8. Residual Channel Attention
  9. Explainability and interpretability
  10. EVALUATION
  11. Datasets and experimental setup
  12. Evaluation metric and implementation
  13. Experimental results and discussion
  14. CONCLUSIONS

Figures (2)

  • Figure 1: Multi-cell images of the five categories considered.
  • Figure 2: Attribution Prediction Gradient Visualization. Columns from left to right: the original image, gradient and integrated gradient overlay, gradient and integrated gradient.