A deep learning framework for automated breast cancer diagnosis using intelligent segmentation and classification

Breast cancer is the most commonly diagnosed cancer among women worldwide, accounting for a significant proportion of new cases. Deep learning (DL) has emerged as a powerful tool for the detection and diagnosis of breast cancer, particularly through the analysis of histological images, a critical component of automated diagnostic systems that directly impact patient management. The BreakHis dataset and the Wisconsin Breast Cancer Database (WBCD) are widely used publicly available resources for deep learning–based analyses of breast cancer histological images in cross-disciplinary healthcare research. A computer-assisted approach employs colour normalisation to reduce the effects of the differences in the distribution of breast histopathology images. In this paper, breast tumour areas of interest are segmented utilising Attention-Guided Deep Atrous-Residual U-Net at the segmentation stage. Subsequently, patches are processed to form feature vectors VGG19 and ResNet50 for the extraction of deep features from the patches. Also, to fine-tune these models even further, the breast cancer datasets are employed, and Levy Flight-based Red Fox Optimisation is used to extract features from the pre-trained models without further training. The Efficient Capsule Network is used to improve the feature representation and classification capabilities. AGDATUNet-LFRFO-ECN, which was suggested in the study, performed better than other models when tested on the WBCD dataset, with a sensitivity of 99.17%, specificity of 99.08%, and accuracy of 99.23%. What’s more, the AGDATUNet-LFRFO-ECN outperformed the available models on BreakHis with a sensitivity of 99.81%, a specificity of 99.79%, and an accuracy of 99.82%, which are the state-of-the-art. • Apply adaptive clustering methods to improve tumor detection in healthcare systems • Implement color normalization to minimize variation in histology image datasets • Use deep feature extraction for precise tumor region classification in breast cancer • Enhance diagnostic accuracy through efficient feature representation and analysis • Demonstrate superior performance over existing models in healthcare datasets