Histopathological Image Diagnosis for Breast Cancer Diagnosis Based on Deep Mutual Learning

Area of Science:

• Medical Imaging Analysis
• Computational Pathology
• Artificial Intelligence in Oncology

Background:

• Breast cancer (BC) is a prevalent and potentially fatal disease affecting millions globally.
• Early and accurate diagnosis is crucial for effective treatment and improved patient outcomes.
• Deep learning (DL) shows significant promise in medical image analysis, including histopathology.

Purpose of the Study:

• To enhance breast cancer detection by integrating qualitative and quantitative data using DL.
• To investigate the efficacy of deep mutual learning (DML) for breast cancer diagnosis.
• To differentiate between aggressive and benign breast cancer using various imaging modalities.

Main Methods:

• Development of deep convolutional neural networks (DCNNs) for assessing BC histopathological images.
• Application of deep mutual learning (DML) to merge qualitative and quantitative imaging data.
• Evaluation of the DML model on diverse datasets: Break His-200×, BACH, and PUIH.

Main Results:

• The DML model achieved high accuracy rates: 98.97% (Break His-200×), 96.78% (BACH), and 96.34% (PUIH).
• The DML approach demonstrated superior performance compared to other methodologies.
• Improved localization results without compromising classification performance, indicating enhanced utility.

Conclusions:

• Deep mutual learning offers a highly accurate and effective method for breast cancer detection from histopathological images.
• The DML model's ability to enhance localization and classification makes it a valuable tool for diagnostics.
• Further development is planned to integrate this diagnostic model into clinical settings.