ABSTRACT
This chapter mainly focuses on the detection and classification of brain tumor. Detection and classification of brain tumor is a fundamental problem in computer-aided diagnosis for biomedical applications. The medical image classification difficulties can be efficiently solved with the help of transfer learning, considering the deep learning perception. Transfer learning is normally expressed through the use of pre-trained models. In this chapter, two pre-trained models VGG-16 and VGG-19 have been optimized for classifying the normal and brain tumor having patient from magnetic resonance imaging image data set. The proposed methodology comprises two important steps. First, the images are pre-processed and then the pre-processed images are classified using VGG-16 and VGG-19. The experiment is carried out on a data set of 253 images. Of 253, there are 98 images which contain no tumor and the remaining 155 images have brain tumor symptoms. The achieved model accuracy is 86.62% using the optimized VGG-16 model. On the other hand, the achieved model accuracy is 85.32% using the optimized VGG-19 model. 15% of image data are used for testing purpose, and 85% of image data are used for training purpose. The achieved area under the curve (AUC) in the training and validation stage is 85.80% and 87%, respectively, when optimized VGG-16 is applied. The achieved AUC in the training and validation stage is 83.73% and 85%, respectively, when optimized VGG-19 is applied.
