Deep feature selection using local search embedded social ski-driver optimization algorithm for breast cancer...

Area of Science:

• Medical Imaging
• Artificial Intelligence
• Computational Biology

Background:

• Breast cancer is a leading cause of mortality in women, underscoring the need for effective early detection methods.
• Computer-aided detection systems enhance radiologists' ability to identify abnormalities in mammograms.
• Classifying malignant breast masses from mammograms remains a significant challenge in medical research.

Purpose of the Study:

• To propose a novel method for classifying breast masses using mammograms.
• To improve the accuracy and efficiency of breast cancer detection through automated analysis.
• To develop a reliable model for differentiating malignant from healthy breast tissue.

Main Methods:

• Feature extraction from mammograms using an attention-aided VGG16 model.
• Optimization of extracted features using the Social Ski-Driver (SSD) algorithm with Adaptive Beta Hill Climbing.
• Classification of the optimal feature subset using the K-nearest neighbors (KNN) classifier.

Main Results:

• The proposed model achieved 96.07% accuracy in classifying breast masses on the DDSM database.
• Effective identification and differentiation of malignant and healthy breasts were demonstrated.
• High performance was achieved using only 25% of the features extracted by the attention-aided VGG16 model.

Conclusions:

• The developed method shows significant potential for accurate and efficient breast cancer detection.
• The integration of deep learning, meta-heuristics, and feature selection offers a robust approach to medical image analysis.
• The publicly available code facilitates further research and development in computer-aided breast cancer diagnosis.