Accuracy Analysis of Deep Learning Methods in Breast Cancer Classification: A Structured Review

Area of Science:

• Medical Imaging
• Computational Pathology
• Artificial Intelligence in Oncology

Background:

• Histopathological imaging is crucial for breast cancer diagnosis but is time-consuming due to image complexity.
• Early breast cancer detection is vital for effective medical intervention.
• Deep learning (DL) offers potential for automated diagnosis but faces challenges like overfitting and imbalanced data.

Purpose of the Study:

• To systematically review and analyze current research on deep learning for classifying histopathological breast cancer images.
• To assess recent approaches and identify cutting-edge DL methods in this field.
• To provide a foundation for developing more sophisticated DL techniques.

Main Methods:

• Systematic literature review of studies on DL for histopathological breast cancer image classification.
• Inclusion of literature from Scopus and Web of Science (WOS) databases.
• Analysis of papers published up to November 2022.

Main Results:

• Deep learning methods, especially convolution neural networks (CNNs) and their hybrid forms, are the most advanced approaches.
• These methods demonstrate capability in classifying complex histopathological images.
• Pre-processing, ensemble, and normalization techniques can enhance DL performance and address data issues.

Conclusions:

• CNNs and hybrid DL models represent the state-of-the-art for automated breast cancer histopathological image classification.
• Further research is needed to compare existing DL approaches and develop novel techniques.
• Addressing overfitting and data imbalance remains critical for robust DL solutions.