Hybrid Feature Selection And Densestnet Approach For Breast Cancer Identification And Categorization

Early detection is key to taking proactive measures to eliminate breast cancer, which is a frequent cause of mortality for females in many developing nations. This study introduces a novel approach to breast cancer screening that makes use of cutting-edge computational methods. To improve precision, this study employs robust deep-learning models with astute image- processing approaches. The Breast Cancer dataset underwent pre-processing using adaptive filtering and histogram equalization. To produce attributes like texture, edges, forms, and sizes, data augmentation was done using a method termed cycle Generative adversarial network (GANS). In addition, by utilizing VGG16 MODEL preprocessing for enhanced feature extraction, it enhanced breast cancer detection by decreasing the likelihood of over fitting a dataset with few samples. Then, a new hybrid optimization method called Hybrid Red Deer plus Sparrow Search Optimization, which combines the Red Deer Algorithm with the Sparrow Search Algorithm, was used to choose the most relevant subset of features. Hybrid Red Deer with Sparrow Optimization (HRDSO) is the foundation of a new architecture called DenseXe Net by integrating both DensetNet and enhanced ResNeXt, whose objective was to improve the performance of BC prediction. Compared with various alternative approaches and evaluated against different performance metrics, the proposed model achieves 97.58% accuracy for BC detection. Breast segmentation was performed using MATALAB evaluation.