<p>One of the main diseases afflicting women globally is breast cancer, which also has a major impact on death rates. Improving patient survival and treatment results depends heavily on early and precise detection. Despite numerous deep learning approaches proposed in recent years, many models still fall short in achieving reliable accuracy, potentially resulting in incorrect or delayed diagnoses. To tackle this problem, we present a brand-new technique called RADGPWSCN-GCRA. This model analyzes mammogram images from the MIAS and INBreast datasets. First, image quality is improved using a smart filtering method to reduce noise. Then, regions likely to contain tumors are accurately separated. Finally, a combined model detects important patterns and relationships within the images for identifying tissue as either normal, benign, or cancerous. The model is optimized for optimal performance using a method inspired by nature. The model achieved very high accuracy 99.9% on both datasets, showing its reliability and effectiveness. Our method combines advanced image analysis and pattern recognition techniques to deliver highly accurate breast cancer detection. It shows strong potential for supporting real-world clinical diagnosis.</p>