An Efficient Method to Classify Brain Tumor using CNN and SVM

Brain Tumor is one of the most sophisticated diseases for the human body that happens when the brain cells start increasing unconditionally. Before giving treatment, the main challenge is to detect and classify tumors from brain MRI images. Researchers have been working really hard for ages to find the best method with higher accuracy for implementing them in real life medical image classification. The main problem is that when a classifier deals with huge amount of data it becomes difficult to classify them accurately. To solve this a CNN-SVM based method is proposed to classify brain tumor with higher accuracy. Firstly, a convolutional neural network having 19 layers is constructed using three convolutional 2D layers, three max-pooling layers, two fully-connected layers, three batch-normalization layers with activation functions reLu. Secondly softmax is used as a classifier and implemented over a dataset containing 3064 images on three class of tumor images (glioma tumors, meningioma tumors, and pituitary tumors). After that, another classifier named support vector machine is used to improve the accuracy of the CNN model using the features extracted from the model. The final accuracy of this proposed CNN-SVM based method is found 97.1%.