Digital Enabled Diagnosis System for Mammogram Images Classification Using Machine Learning

Breast tumor is the most commonly detected tumors and remain one of the leading causes of cancer-related mortality among women worldwide. Although mammography is widely adopted for early breast cancer screening, existing diagnostic systems are often time-consuming and highly dependent on expert radiologists. Moreover, mammogram images frequently suffer from noise and low contrast, which can result in inaccurate feature extraction and misclassification. Many existing studies also report high classification error, limited robustness, and poor performance on imbalanced datasets, restricting their practical deployment. To address these challenges, a digital enabled diagnosis system for early breast cancer detection from mammograms is proposed. The system is designed as a flexible and efficient predictive framework capable of classifying breast conditions into malignant, benign, and normal categories. The proposed approach follows a structured five-stage pipeline. First, two-dimensional median filtering (2DMF) is employed to enhance image quality with low computational cost. Next, a Boltzmann trust network is used for accurate tumor localization and segmentation. Optimal feature selection is achieved using a hybrid spiral intelligent ambiguous rough set (HSIARS) method, which reduces variance and classification error. Finally, classification is performed using an optimized gradient executive bundling (Optimized GEB) model, significantly improving prediction accuracy and overall diagnostic performance.