Mess-based Prelu-DenseNet53: An early-stage invasive carcinoma breast cancer prediction model

Performing Breast Cancer (BC) classification is essential for effective diagnosis since BC is an extremely risky disease. But, the challenges of irrelevant feature extraction are faced by the conventional BC classification owing to the insufficient prognostic and predictive power that resulted in the offender’s misidentification. To overcome such challenges, a novel classification framework utilizing Max Entropy-based Ebola Snakes-Sobel-Parameterized-Rectified Linear unit-DenseNet 53 (MESS-PReLu-DenseNet 53) is proposed in this paper. The proposed system undergoes various phases like image resizing, noise and background removal, contrast enhancement, segmentation, feature extraction, along with classification. The input images’ size is modified in the image resizing phase; also, in the subsequent noise removal phase, the unwanted noises are removed utilizing Adaptive Local Power-Law Transformation (ALP-LT). Afterward, the background images are removed; then, the image quality is enhanced by preserving the image’s edges utilizing the Non-Subsampled Contourlet Transform fused Multi-scale Retinex (NSCT-MR) technique. After that, by utilizing the Max Entropy-based Ebola Snakes-Sobel (MESS) approach, the enhanced images thus obtained are segmented for detecting spiculation. Meanwhile, the more appropriate and useful data are extracted from the pre-processed image. For effective BC-type classification, the detected spiculation and the extracted features are fed into the Parameterized-Rectified Linear unit-DenseNet 53 (P-ReLu-DenseNet53). The experiential results exhibited that higher accuracy is attained by the proposed model; also, it efficiently classifies the BC types without misprediction.