Accurate breast cancer diagnosis strategy (BCDS) based on deep learning techniques

Breast cancer (BC) survival rates and the patient's quality of life are boosted by early detection and timely therapy. It is the most prominent cancer and the primary trigger for deaths due to cancer in women around the world. As a result, a variety of artificial intelligence-based computer-assisted procedures are being included in the conventional diagnostic workflow. This study proposes an accurate Breast Cancer Diagnosis Strategy (BCDS) based on deep learning techniques. A framework for BCDS will be presented to consolidate and improve BC detection by defining three stages of BCDS: (i) Preprocessing Stage (PS), (ii) Classification Stage (CS), and (iii) Ensemble Voting Stage (EVS). In PS, three preprocessing operations which are image resizing using bilinear interpolation, data augmentation using Conditional- Convolutional Generative Adversarial Network (C-DCGAN) with Adversarial Feedback Loop (AFL) and data enhancement using Multiscale Retinex with Color Restoration (MSRCR) algorithm will be performed to enhance images and increase the performance of diagnostic model. In CS, an ensemble learning-based technique that includes three classifiers called Xception, Inception-ResNet-V2, and Visual Geometry Group (VGG16) will be applied to accurately diagnose BC patients. Finally, in EVS, majority voting and weighted random forest based on accurate voting techniques will be provided to get the most optimal diagnosis. In the benchmark BreakHis dataset, test results illustrated that the three fine-tuned classifiers (Xception, Inception-ResNet-V2, and VGG16) of BCDS provide accuracy values equal 97%, 98%, and 99.28% for multi-classification. These fine-tuned classifiers yield accuracy scores of 99%, 99%, and 100% based on binary jobs. Results indicate that the BCDS model achieves 100% accuracy for binary tasks and 99.89% accuracy for multi-classification tasks. Physicians can utilize BCDS as a decision-support framework, especially in nations of poverty when resources and knowledge are a handful. Early and accurate identification of the tumor's type lessens the possibility of a botched treatment and lowers the death rate from tumors in the breast.