Deep feature-based automatic classification of mammograms

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Oncology

Background:

• Breast cancer is a leading cause of death in women globally.
• Early detection is crucial but challenging.
• Mammography is key for diagnosis, but manual interpretation is labor-intensive.

Purpose of the Study:

• To develop an automated computer-aided diagnosis (CADx) system for breast cancer classification.
• To enhance the accuracy and efficiency of mammogram analysis.
• To assist radiologists in differentiating benign and malignant tumors.

Main Methods:

• Proposed a deep ensemble transfer learning model for automatic feature extraction.
• Utilized a neural network classifier (nntraintool) for classification.
• Pre-processed mammogram images before inputting into the ensemble model.

Main Results:

• Achieved an accuracy of 0.88 for breast cancer classification.
• Obtained an Area Under the Curve (AUC) of 0.88.
• Demonstrated robust feature extraction and classification capabilities.

Conclusions:

• The proposed deep learning methodology is a promising CADx system.
• The approach offers a robust solution for automated breast cancer classification.
• This system can aid in improving diagnostic outcomes for mammography.