Brain Tumors Classification using Deep Models and Transfer Learning

Abstract Brain tumor classification using magnetic resonance imaging (MRI) along with medical knowledge results in a better decision to treat a patient. Also, the classification of some types of tumors is often a challenging problem due to the need for a detailed analysis of tumor texture. Therefore, machine learning approaches and specialists' experience can be very beneficial. This paper aims to explore the potential of deep learning structures in classifying different types of brain tumors. Our approach involves using a 50-layer ResNet deep network, which has shown promising results in various image classification tasks. For more consideration, transfer learning technique also is employed to evaluate the performance of the proposed algorithm. The presented algorithms are compared with the other deep networks such as convolutional neural network (CNN), recurrent neural network (RNN), and dictionary learning-based classifier. The studies show that the ResNet-50-based deep model performs better than the mentioned classifier categories in different evaluation criteria such as accuracy, sensitivity, and robustness and has an effective role in medical diagnosis.