The use of magnetic resonance images (MRI) is a frequently used tool in disease detection. The use of healthcare professionals to examine MRI images and to identify diseases are among traditional methods. Therefore, one way to improve clinical healthcare is to present and analyze medical images more efficiently and intelligently. Brain tumors can be of different types, and accordingly, they can cause serious health problems in adults and children. Such bulks can occur anywhere in the brain in different sizes and densities. This is not a standardized situation due to its nature. The diagnoses are revealed by the experts by analyzing the tumor images manually. In the proposed model, it is aimed at automating the process and reducing human errors in the system. The model is based on the deep learning technique, which is a probabilistic neural network to identify unwanted masses in the brain. In this study, a model has been created with VGG and CNN (Convolutional Neural Network) architectures, which are among the deep learning techniques. The performance values of the model outputs, accuracy, error rates and specificity separators are discussed comparatively.