Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Comparative Analysis and Ensemble Enhancement of Leading CNN Architectures for Breast Cancer Classification

Gary Murphy, Raghubir Singh

TL;DR

The paper tackles the problem of identifying which CNN architectures are most effective for breast cancer classification on histopathology images by conducting a large-scale, standardized cross-model comparison. It introduces a comprehensive methodology that includes pre-generation and serialization of datasets, systematic augmentation studies, and a diverse set of standalone CNNs plus novel ensemble architectures that combine three CNNs with multiple classifiers. Key findings show that top standalone models can reach very high accuracies (e.g., up to 99.75% on BreakHis) and that ensembles offer marginal yet consistent improvements, with Bach Online challenge results around 89%. The proposed framework, including automated result curation and robust data conditions, is transferable to other medical imaging tasks and enables rapid, reproducible model selection and optimization.

Abstract

This study introduces a novel and accurate approach to breast cancer classification using histopathology images. It systematically compares leading Convolutional Neural Network (CNN) models across varying image datasets, identifies their optimal hyperparameters, and ranks them based on classification efficacy. To maximize classification accuracy for each model we explore, the effects of data augmentation, alternative fully-connected layers, model training hyperparameter settings, and, the advantages of retraining models versus using pre-trained weights. Our methodology includes several original concepts, including serializing generated datasets to ensure consistent data conditions across training runs and significantly reducing training duration. Combined with automated curation of results, this enabled the exploration of over 2,000 training permutations -- such a comprehensive comparison is as yet unprecedented. Our findings establish the settings required to achieve exceptional classification accuracy for standalone CNN models and rank them by model efficacy. Based on these results, we propose ensemble architectures that stack three high-performing standalone CNN models together with diverse classifiers, resulting in improved classification accuracy. The ability to systematically run so many model permutations to get the best outcomes gives rise to very high quality results, including 99.75% for BreakHis x40 and BreakHis x200 and 95.18% for the Bach datasets when split into train, validation and test datasets. The Bach Online blind challenge, yielded 89% using this approach. Whilst this study is based on breast cancer histopathology image datasets, the methodology is equally applicable to other medical image datasets.

Comparative Analysis and Ensemble Enhancement of Leading CNN Architectures for Breast Cancer Classification

TL;DR

The paper tackles the problem of identifying which CNN architectures are most effective for breast cancer classification on histopathology images by conducting a large-scale, standardized cross-model comparison. It introduces a comprehensive methodology that includes pre-generation and serialization of datasets, systematic augmentation studies, and a diverse set of standalone CNNs plus novel ensemble architectures that combine three CNNs with multiple classifiers. Key findings show that top standalone models can reach very high accuracies (e.g., up to 99.75% on BreakHis) and that ensembles offer marginal yet consistent improvements, with Bach Online challenge results around 89%. The proposed framework, including automated result curation and robust data conditions, is transferable to other medical imaging tasks and enables rapid, reproducible model selection and optimization.

Abstract

This study introduces a novel and accurate approach to breast cancer classification using histopathology images. It systematically compares leading Convolutional Neural Network (CNN) models across varying image datasets, identifies their optimal hyperparameters, and ranks them based on classification efficacy. To maximize classification accuracy for each model we explore, the effects of data augmentation, alternative fully-connected layers, model training hyperparameter settings, and, the advantages of retraining models versus using pre-trained weights. Our methodology includes several original concepts, including serializing generated datasets to ensure consistent data conditions across training runs and significantly reducing training duration. Combined with automated curation of results, this enabled the exploration of over 2,000 training permutations -- such a comprehensive comparison is as yet unprecedented. Our findings establish the settings required to achieve exceptional classification accuracy for standalone CNN models and rank them by model efficacy. Based on these results, we propose ensemble architectures that stack three high-performing standalone CNN models together with diverse classifiers, resulting in improved classification accuracy. The ability to systematically run so many model permutations to get the best outcomes gives rise to very high quality results, including 99.75% for BreakHis x40 and BreakHis x200 and 95.18% for the Bach datasets when split into train, validation and test datasets. The Bach Online blind challenge, yielded 89% using this approach. Whilst this study is based on breast cancer histopathology image datasets, the methodology is equally applicable to other medical image datasets.
Paper Structure (42 sections, 7 figures, 7 tables, 3 algorithms)

This paper contains 42 sections, 7 figures, 7 tables, 3 algorithms.

Table of Contents

  1. Introduction
  2. Breast Cancer
  3. CNNs
  4. CNNs in Breast Cancer Classification
  5. Related Works
  6. Table of Related Works
  7. Existing comparison of works
  8. Aims and Contributions
  9. Datasets tested
  10. BreakHis Dataset
  11. BAch 2018 Dataset
  12. Models Tested
  13. Standalone Models
  14. Bespoke Model Architecture
  15. Ensemble Architectures
  16. ...and 27 more sections

Figures (7)

  • Figure 1: Basic block architecture of AlexNet b22
  • Figure 2: ResNet50 Simplified Block Architecture showing the 34 layers with learnable parameters. The arced lines show the 'shortcuts' introduced in the model to enable its residual capability. Redrawn and transposed from vertical to horizontal b27.
  • Figure 3: Bespoke CNN Model Architecture
  • Figure 4: End-to-end Architecture - Pre-processing, CNN models in a loop, Ensemble Models in a loop - inferences
  • Figure 5: Dense block classification variants
  • ...and 2 more figures