<p>Cancer is the world’s leading cause of death. At present, the challenges which medical doctors face in fighting cancer are increasing. Thus, there is a real need for accurate CAD systems not only for detecting cancer, but also for predicting treatment response, to help in cancer diagnosis and treatment later on. In this work, a well-known algorithm which is Convolutional Neural Network (CNN) will be considered. The mechanism/strategy to deal with a patch of images for women breasts that have been screened mammography indicating whether breast cancer exists or not and what is the action taken by the specialist for that specific patient, in order to build a system design based on deep learning approach to predict the treatment response for the patients later on. The methodology is based on Convolutional Neural Network (CNN to classify the manually collected mammograms into two main stages; before treatment (Stage-1) and after treatment (Stage-2), and then extract keywords from the clinical test report for each patient mammogram to detect that it is normal or abnormal. After that, we will have a classified dataset to deal with in a special manner related to the system requirements, to come up with simulation results after doing coding on the classified dataset. The proposed systems excelled in the MobileNet architecture model by applying 10&#xa0;K cross-validation in the two stages for the dataset of 213 patients with 1705 images and after data augmentation to reach 5100 images. This model’s results came up with 93.7% accuracy for the first stage, and 93.0% accuracy for the second stage. The proposed approach has great benefits for radiologists and physicians as a new mechanism to support and evaluate the predicting response of breast cancer disease treatment, using a simple, accurate and effective technique.</p>