Brain Tumour Images Classification Using Support Vector Machine and Pre-Trained Convolutional Neural Networks

Abstract

Brain tumour classification allows for successful treatment. The aim of this paper is to identify two categories of brain tumors using a benchmark dataset of Magnetic resonance image (MRI) brain tumor. Due to a lack of huge training data, this paper using data augmentation so that training data is increased. In this aspect, the original images are processed with median filtering and the obtained results are used as total training dataset In this study, we propose a novel method for classifying pictures of brain tumors and cancer using pre-trained convolution neural networks (CNN) and support vector machines (SVM). In this appraoch, the features are derived using two pre-trained convolution neural networks models namely Alexnet and ResNet. Support vector machine (SVM) is used to classify tumor picture after characteristics have been extracted. This paper is addressing issues like: to reduce training time, improve image classification accuracy and prevent over fitting. Instead of training with same dataset, using of segmented images as training data will achieves good performance. We trained our architectures with segmented images and limited pre-processing for three epochs to test classification accuracy and time consumption. The experimental performance attained is 97.34% with transfer learning.