In today’s fast-paced world, lifestyle choices have substantially changed, influencing overall health. Among the various methods available for breast cancer screening, mammography remains the most widely used technique. Recent advancements have introduced computer-aided diagnostic systems that improve mammogram quality and assist in detecting suspicious regions. This paper aims to detect the most applicable pre-processing algorithms for optimizing the interpretation and classification of mammographic images. These algorithms are essential in ensuring precise segmentation and classification processes. Mammographic images are acquired using the Direct File Upload method and subjected to pre-processing approaches including Gaussian filtering (GF), contrast-limited adaptive histogram equalization (CLAHE), and unsharp masking (USM). These enhancements improve image clarity before proceeding to segmentation and classification. A modified U-Net (U-shaped convolutional neural network) model is employed for segmentation. In contrast, a deep convolutional neural network (DCNN) classifier evaluates different combinations of GF, CLAHE, and USM to determine their effectiveness. The Findings suggest that applying multiple pre-processing methods enhances lesion detection and aids in distinguishing benign and malignant cases. The proposed approach achieves a high classification precision of 98.39%, exceeding traditional procedures and demonstrating the potential of AI-driven techniques in breast cancer detection.
