Despite its proven record as a breast cancer screening tool, mammography remains labour intensive and has recognized limitations, including low sensitivity in women with dense breast tissue. In the last ten years, Neural Network advances have been applied to mammography to help radiologists increase their efficiency and accuracy. In this research, a novel breast cancer detection framework is developed and evaluated on a medical dataset. The proposed detection and classification model passes through following five major phases: pre-processing, segmentation, feature extraction, feature selection, and classification. First, the pre-processing is performed for the given input image using the median filtering technique. Then, the preprocessed image is subjected to segmentation via K-means clustering. Local Binary Pattern, Haralick Features, Contrast, Correlation, Sum of squares: Variance, Inverse Difference Moment, Entropy, Information measures, Gray Level Run Length Features are extracted. Finally, the classification process is carried out via a hybrid classification approach, which is constructed by blending the deep belief network and neural network. Experimental results demonstrate that the model ss a powerful tool for diagnosing breast cancer by means of classifying the mammogram images with overall sensitivity of 1 and 0.99 AUC. To enhance the accuracy of detection, the activation function of DBN is optimized using a new Customized Individual Activity and Information Sharing based Team Work Optimization (CISA-TOA) model. This CISA-TOA is the conceptual improvement of the standard Teamwork Optimization Algorithm (TOA). Finally, a comparative evaluation is carried out to validate the efficiency of the projected model.