Breast cancer image analysis utilizing feature fusion-based hybrid classification network influenced by adaptive u-net with generative adversarial segmentation network

Nowadays, breast cancer is a very dangerous sickness among women though well-timed identification relatively enhances the rate of survival. To win over the breast tumor, an effective diagnosis mechanism is very necessary. The removal of tumor cells from the affected area stay as challenging. Classifying the breast tumor is very challenging for pathologists because of the tumor cell's heterogeneous nature. Hence, it is important to adopt an efficient breast tumor classification architecture employing a deep learning strategy. It supports the premature detection of breast cancer and proper treatments that help to prolong the chances of patients’ survival. Initially, the segmentation procedure is executed in the input images using an Adaptive U-Net Generative Adversarial Network (AUNet-GAN). Moreover, the variables of AUNet-GAN are tuned employing the Enhanced Golf Optimization Algorithm (EGOA). The major principle of segmentation is to change its illustration into a meaningful format of the image. Further, the segmented images are fed to the Feature Fusion-based Hybrid Network (FFHNet)-aided classification of the breast cancer model. This classification process is utilized to attain better benign and malignant. Later, several experimental observations are executed to estimate the efficacy rate of the designed method over the baseline models and also diverse efficacy metrics like accuracy, False Negative Rate (FNR), Prevalence Threshold (PT), dice coefficient, etc. The accuracy of the offered approach for datasets 1 and 2 is 94.30 and 95.44. Accordingly, the accuracy values of the existing works like CNN, ResNet, RNN, and VGG16 for dataset 1 are 89.33, 91.70, 91.70, and 90.99 and also for dataset 2 is 89.54, 82.91, 90.99, and 92.75. Thus, the empirical outcomes of this study help to provide proper treatment for breast cancer and the usage of classifiers efficiently improves the accuracy and strength of detection for breast cancer models.