Brain Tumor Classification in MRI Images Using Transfer Learning

Abstract

AbstractBio-medical image processing is a challenging and most emerging field in the computer based medical diagnosis. Brain tumors are most common and harmful disease these days, over a thousands people suffering annually with brain tumors. Among different types of brain tumors Gliomas are most common (around 30% of all brain tumor). Magnetic Resonance imaging (MRI) is a well received imaging approach to identify brain tumors. The classification of brain MRI is a difficult task. Due to large number of MRI images manual analysis is tedious and pron to errors. So there is a need of automatic classification method. The proposed method focus on a binary classification problem that seeks to differentiate between normal and tumorous MRI images by using the idea of deep transfer learning with the help of a pre-trained VGG16 model. The proposed method achieve an accuracy of 91.8% and consume less time as compared to previous traditional methods. Image augmentation techniques are also used to increase the size of training data set. The proposed method suggest that transfer learning based approach achieve good results despite limited data availability in medical field.KeywordsDeep learningConvolutional neural networkTransfer learning MRIBrain tumorImage classification