Deep Learning based Analysis for Automated Detection and Classification of Brain Tumor

Abstract

Reproduction of quick and indefinite cells within brain cause a tissue which is generally known as a brain tumor. A number of individuals remain untreated as it does not show any hard symptoms at an initial stage. For identification of such disease, many neurologists suggest Computer Tomography Scan, Magnetic Resonance Imaging etc. which can be time consuming process and expensive too. To avoid so, various computer assisted methods have been suggested by researchers to overcome the drawbacks of traditional approaches. Deep learning has been considered as one of the reliable approaches for identification and classification of brain tumor disease that can prevent an individual from death due to its strong features capability for providing quick and better results at an early stage as compared to the traditional approaches. This research study has considered 3264 images from kaggle having 2764 tumor images and 500 with healthy ones and proposed a model that comprises of Visual Geometry Group (VGG) having 16 layers in collaboration with the concept of transfer learning to perform the diagnosis and classification of brain tumor disease. The proposed model has delivered an accuracy of 98.16%, precision of 99.09%, recall of 98.73% and F1-score of 98.91% which is far better when compared to the existing approaches.