Deep Learning Models Show Promising Potential in Breast Cancer Diagnosis

Strategic Insights -

A recent study published in the Frontiers in Medicine journal has found that deep learning models, particularly convolutional neural networks (CNNs) and transformer-based architectures, demonstrate excellent performance in classifying breast cancer pathology images. The research highlights the potential of these models in enhancing clinical decision-making in breast cancer management. According to the study, CNN-based models such as ResNet50, RegNet, and ConvNeXT achieved near-perfect performance in binary classification tasks, with an accuracy of 99.2% and an AUC of 0.999. The study also found that transformer-based foundation models, such as UNI, possess strong feature extraction capabilities, but their zero-shot performance on this specific task was limited. However, with simple fine-tuning, they quickly achieved excellent results.

Key Takeaways:

  • The study found that CNN-based models such as ResNet50, RegNet, and ConvNeXT achieved near-perfect performance in binary classification tasks, with an accuracy of 99.2% and an AUC of 0.999.
  • The best-performing model in the binary classification task was ConvNeXT, which attained an accuracy of 99.2% (95% CI: 98.3%-1), a specificity of 99.6% (95% CI: 99.1%-1), an F1-score of 99.1% (95% CI: 98.0-1%), a Cohen's Kappa coefficient of 0.983 (95% CI: 0.960-1), and an AUC of 0.999 (95% CI: 0.999-1).
  • In the eight-class classification task, the best-performing model was the fine-tuned foundation model UNI, which attained an accuracy of 95.5% (95% CI: 94.4-96.6%), a specificity of 95.6% (95% CI: 94.2-96.9%), an F1-score of 95.0% (95% CI: 93.9-96.1%), a Cohen's Kappa coefficient of 0.939 (95% CI: 0.926-0.952), and an AUC of 0.998 (95% CI: 0.997-0.999).
  • The study found that using foundation model encoders directly without fine-tuning resulted in generally poor performance on the classification task.
  • The research concluded that deep learning models are highly effective in classifying breast cancer pathology images, particularly in binary tasks where multiple models reach near-perfect performance.
  • The study highlights the potential of deep learning models in enhancing clinical decision-making in breast cancer management.

Statistics:

  • 99.2% accuracy achieved by ConvNeXT in binary classification task
  • 0.999 AUC achieved by ConvNeXT in binary classification task
  • 95.5% accuracy achieved by fine-tuned foundation model UNI in eight-class classification task
  • 0.998 AUC achieved by fine-tuned foundation model UNI in eight-class classification task
  • 685,000 deaths in 2020 due to breast cancer (GLOBOCAN 2020)

Sources:

  • Comparative analysis of convolutional neural networks and transformer architectures for breast cancer histopathological image classification. Frontiers in Medicine, 2025;12:1606336.
  • Journal of Engineering. July 14, 2025; p 1159.