Improved DeTraC Binary Coyote Net-Based Multiple Instance Learning for Predicting Lymph Node Metastasis of Breast...

Area of Science:

• Computational pathology
• Medical imaging analysis
• Machine learning in oncology

Background:

• Early detection of lymph node metastasis in breast cancer is critical for improving patient prognosis and treatment efficacy.
• Accurate identification of metastatic lymph nodes is a significant challenge in breast cancer diagnostics.

Purpose of the Study:

• To introduce and evaluate the Improved Decompose, Transfer, and Compose Binary Coyote Net-based Multiple Instance Learning (ImDeTraC-BCNet-MIL) method.
• To enhance the accuracy of predicting lymph node metastasis from Whole Slide Images (WSIs) in breast cancer.

Main Methods:

• WSIs were segmented into patches using Otsu and double-dimensional clustering.
• A novel multiple instance learning approach (ImDeTraC-BCNet-MIL) was developed for feature construction and prediction.
• The method was applied for feature generation in both training and testing phases to identify lymph node metastasis.

Main Results:

• The ImDeTraC-BCNet-MIL model achieved high performance on benchmark datasets.
• Accuracy reached 95.3% on Camelyon16 and 99.8% on Camelyon17.
• Precision and recall rates were also notably high, demonstrating robust predictive capabilities.

Conclusions:

• The ImDeTraC-BCNet-MIL method effectively enhances the early detection of lymph node metastasis in breast cancer.
• This approach represents a significant advancement in computational pathology for cancer diagnostics.