Enhancing Breast Cancer Detection and Classification Using Light Attention Deep Convolutional Network: A Novel Approach Integrating Deep Learning and Medical Imaging

Breast cancer is a severe disease and invasive malignancy, affecting the female population globally. It stands as the main cause of mortality among women. Timely detection of breast cancer leads to proper treatment planning and saves the lives of those struggling with it. In this context, mammogram images are the main contributor to effective breast cancer detection. So, researchers made several breast cancer detection methods, however, these methods did not ensure accurate detection because of poor image resolution, high false positive rates, low contrast, etc. Also, these methods are ineffective at precisely detecting subtle and complex patterns in mammographic images. Therefore, this research develops a novel Light Attention Deep Convolutional Network approach that improves breast cancer detection. In the preprocessing phase, collected images are resized and their pixel values are scaled to a standardized range. Histogram equalization is employed to ensure image quality by adjusting contrast levels, and noise is removed using a filtering method. The developed model prioritizes features that are indicative of early-stage cancer and malignant growth by using a light attention mechanism that assigns different levels of attention to various image regions. The validation of the proposed approach is conducted on two diverse datasets namely Breast mammography images with Masses and Mammography Breast Cancer Detection. The performance analysis ensures its ability to precisely detect and classify key features and attain better performance than the existing methods. The experimental outcomes highlight that the developed model achieves a higher accuracy of 98.8%. This demonstrates its ability for deployment in real-world, automated breast cancer screening applications.