Multi-Scale Vision Transformer with Optimized Feature Fusion for Mammographic Breast Cancer Classification.

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Machine Learning

Background:

• Breast cancer is a leading cause of mortality in women globally.
• Accurate and efficient diagnostic methods are crucial for early detection and treatment.
• Traditional deep learning models face challenges with feature redundancy and suboptimal fusion.

Purpose of the Study:

• To develop a novel deep learning framework for enhanced breast cancer diagnosis and classification.
• To overcome limitations of existing models in feature extraction, fusion, and selection.
• To improve the accuracy and efficiency of mammogram analysis.

Main Methods:

• Utilized MAX-ViT for multi-scale feature extraction and hierarchical representation learning.
• Introduced a gated attention fusion module (GAFM) for dynamic feature integration.
• Employed Harris Hawks Optimization (HHO) for efficient feature selection and XGBoost for classification.

Main Results:

• Achieved high performance metrics on the King Abdulaziz University Mammogram Dataset.
• Reported 98.2% accuracy, 98.0% precision, 98.1% recall, 98.0% F1-score, 98.9% AUC, and 95% MCC.
• Demonstrated superior performance compared to existing state-of-the-art models.

Conclusions:

• The proposed fusion-based deep learning framework significantly improves breast cancer diagnosis and classification.
• The integrated approach of MAX-ViT, GAFM, HHO, and XGBoost offers a robust solution.
• Validated the framework's effectiveness and potential for clinical application.