Exploring Brain Tumor Classification Using Deep Learning

Abstract

Diagnosis at a beginning period and recognition of the type of cancer can assist doctors and health experts in determining the most appropriate treatment. The target of this is to research is to build a reliable means and appropriate method for classifying human brain cancers that uses magnetic resonance imaging (MRI) to distinguish between the many forms of Glioblastoma, malignant tumors, and gland tumours are examples of brain tumors. In order to enhance and achieve accurate results can make preprocessing methods like resize MR images, cropping and data augmentation to avoid over fitting. By using deep learning pre-defined models as ResNet, VGG16, MobileNet and Inception. And transfer-based learning CNN that supported with calculation of dice, sensitivity and specificity we founded that by using dice with CNN model the achieved accuracy was 99.9%.