An Enhanced Mammogram Classification to Detect Breast Cancer using Boosted PSO Feature Selection and Multi-Feature Analysis

Breast cancer is a globally recognized public health issue, particularly impacting women. While mammogram screening is a widely used method for identifying cancerous cells, accurately distinguishing these cells from the surrounding tissue remains a challenge. A novel approach is presented in this study, to categorize mammograms. Initially, preprocessing is done by median filter with Contrast Limited Adaptive Histogram Equalization (CLAHE). Multi-Feature Analysis (MFA) is used for feature extraction to extract texture, shape, and volume information from the mammograms. A Modified Weighted Particle Swarm Optimization (MWPSO) algorithm selects the optimal features, which are fed to the Light Gradient Boosting Machine (Light GBM) for classification. The model's performance is evaluated through k-fold cross-validation and demonstrates high accuracy in mammogram classification. By combining the strengths of image processing and machine learning algorithms, the proposed methodology provides a powerful tool for mammogram classification.