Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Breast Cancer Histopathology Classification using CBAM-EfficientNetV2 with Transfer Learning

Naren Sengodan

TL;DR

This work tackles breast cancer histopathology classification across multiple magnifications by integrating Convolutional Block Attention Modules (CBAM) with EfficientNetV2-XL and employing modality-specific transfer learning alongside CLAHE preprocessing. The proposed CBAM-EfficientNetV2-XL architecture demonstrates state-of-the-art accuracy, achieving 99.01% and a 98.31% F1-score at 400X on BreakHis, while maintaining computational efficiency suitable for real-time deployment. The study emphasizes attention-guided multi-scale feature learning and provides attention-map visualizations to interpret model focus, highlighting nuclei and tissue morphology as key discriminators. The approach, validated on BreakHis with auxiliary pre-training on ICIAR 2018 and PCam, offers a practical pathway to accurate, scalable breast cancer diagnostics in diverse clinical settings.

Abstract

Breast cancer histopathology image classification is critical for early detection and improved patient outcomes. 1 This study introduces a novel approach leveraging EfficientNetV2 models, to improve feature extraction and focus on relevant tissue regions. The proposed models were evaluated on the BreakHis dataset across multiple magnification scales (40X, 100X, 200X, and 400X). 2 Among them, the EfficientNetV2-XL with CBAM achieved outstanding performance, reaching a peak accuracy of 99.01 percent and an F1-score of 98.31 percent at 400X magnification, outperforming state-of-the-art methods. 3 By integrating Contrast Limited Adaptive Histogram Equalization (CLAHE) for preprocessing and optimizing computational efficiency, this method demonstrates its suitability for real-time clinical deployment. 3 The results underscore the potential of attention-enhanced scalable architectures in advancing diagnostic precision for breast cancer detection.

Breast Cancer Histopathology Classification using CBAM-EfficientNetV2 with Transfer Learning

TL;DR

This work tackles breast cancer histopathology classification across multiple magnifications by integrating Convolutional Block Attention Modules (CBAM) with EfficientNetV2-XL and employing modality-specific transfer learning alongside CLAHE preprocessing. The proposed CBAM-EfficientNetV2-XL architecture demonstrates state-of-the-art accuracy, achieving 99.01% and a 98.31% F1-score at 400X on BreakHis, while maintaining computational efficiency suitable for real-time deployment. The study emphasizes attention-guided multi-scale feature learning and provides attention-map visualizations to interpret model focus, highlighting nuclei and tissue morphology as key discriminators. The approach, validated on BreakHis with auxiliary pre-training on ICIAR 2018 and PCam, offers a practical pathway to accurate, scalable breast cancer diagnostics in diverse clinical settings.

Abstract

Breast cancer histopathology image classification is critical for early detection and improved patient outcomes. 1 This study introduces a novel approach leveraging EfficientNetV2 models, to improve feature extraction and focus on relevant tissue regions. The proposed models were evaluated on the BreakHis dataset across multiple magnification scales (40X, 100X, 200X, and 400X). 2 Among them, the EfficientNetV2-XL with CBAM achieved outstanding performance, reaching a peak accuracy of 99.01 percent and an F1-score of 98.31 percent at 400X magnification, outperforming state-of-the-art methods. 3 By integrating Contrast Limited Adaptive Histogram Equalization (CLAHE) for preprocessing and optimizing computational efficiency, this method demonstrates its suitability for real-time clinical deployment. 3 The results underscore the potential of attention-enhanced scalable architectures in advancing diagnostic precision for breast cancer detection.

Paper Structure

This paper contains 15 sections, 9 equations, 5 figures, 1 table.

Table of Contents

  1. Introduction
  2. Literature Review
  3. Methodology
  4. Model Architecture
  5. CBAM Architecture
  6. Channel Attention Module
  7. Spatial Attention Module
  8. EfficientNet Backbone
  9. Datasets
  10. Data Preprocessing
  11. Training and Hyperparameter Tuning
  12. Evaluation Metrics
  13. Experimental Results and Discussions
  14. Results
  15. Conclusion

Figures (5)

  • Figure 1: Model Architecture
  • Figure 2: CBAM
  • Figure 3: Spatial and Channel Attention
  • Figure 4: Data Augmentation and preprocessing techniques
  • Figure 5: ROC-AUC Curve of EfficientNetV2-XL with CBAM