Brain tumors classification from MRI images: A comparative study between different neural networks

The detection of brain tumors through the analysis of images is becoming increasingly common for promptly treating patients. Among the different types of imaging techniques, Magnetic Resonances Imaging (MRI) is probably the most popular one in the pre- and post-treatment to estimate the structure of tumors. Thus, they also represent a useful means for supporting intelligent techniques in the identification of brain tumors, enabling machine learning models to completely automate the classification task. In this paper, we propose a new methodology for classifying brain tumors through the analysis of MRI images. In particular, our approach relies on a feature extraction technique to obtain representative data, which are used as input for two predictive models, a Convolutional Neural Network (CNN) and a Residual Neural Network (ResNet). We discuss experimental evaluation performed over a ground-truth dataset and show a comparative analysis between proposed models in the classification of tumors according to their type.