Deep learning has the advantage of improved medical image processing research. Breast cancer is the most common cancer in women, and several apps have been established to improve early detection. Mammography is the most common method of screening for breast cancer, which has a high mortality rate, for women worldwide. The robustness of deep learning processes to big data strengthens the analytical abilities of Machine Learning (ML) models through feature selection on mixed image databases. Although Existing CNN-based systems have achieved higher performance than machine learning-based systems in classifying mammography images, there are several issues. These problems include ignorance of semantic features, limitations in the analysis of current image blocks, loss of blocks in low-contrast mammography images, and segmentation. These problems lead to mammography patches, computational costs, conclusions based on recent patches, and misinformation that differences in patch intensities cannot be recovered. To overcome these issues, we proposed the method Deep Dense net Convolution Generative Neural Network (D2CGNN) for using the Mammography images early detection and classification to improving the accuracy. Initially we collected the mammogram images from standard repository for malignant detection. The first pre-processing stage is based on Simple Decision Median Filter (SDMF) is used for image resizing, enhancing appearance, and reducing image noise to filter from the dataset. Then in the second step, there is a segmentation step using pre-processed image Visual Geometry Group19 (VGG) to segment the image with the aim of detecting masses in mammograms and extracting ROIs from the image. After segmentation to entering the third stage of feature selection based on Absolute Maximum Support Feature Selection (AMSFS) is used to select features based on the maximum weight of the support section. And the VGG-19 network model is used with reduced convolutional and max pooling layers to provide a large feature dataset to classify mammograms efficiently as benign or malignant for early detection for using most support feature weights. Finally, classification is used for better accuracy. Early detection based on Deep Dense net Convolution Generative Neural Network (D2CGNN) is used to evaluate the normal and abnormal images that D2CGNN generates to the training data to help the classifiers avoid overfitting. Regarding validation accuracy, D2CGNN is better than the image classification of previous methods.

Deep Learning, mammograms, Breast cancer, feature weights, VGG-19, D2CGNN, feature selection, classification, ROIs.