BRAIN TUMOR CLASSIFICATION USING SVM AND KNN MODELS FOR SMOTE BASED MRI IMAGES

Abstract

A brain tumor is a disease by which many people are affected. It is differentiated into two types-mainly benign and malignant tumors. The benign is a non-cancerous tumor and it can be removed by surgery. Hence, there is more chance of curing the affected person. On the other hand, the malignant tumor is cancerous. It will spread all over the body and it is very harmful and it is difficult to cure the affected person. In the proposed system, the brain tumor MR images are taken and the tumors are first segmented using Otsu’s threshold technique using MATLAB processing tools. Segmented images further transmitted to discrete wavelet transform (DWT) to get the features of the images. Identified images features are further applied to principle component analysis (PCA) for dimensionality reduction. Further, synthetic minority over-sampling technique (SMOTE) is used to balance the samples in the dataset classes. The proposed work has been tested with K-nearest neighbor(KNN)and support vector machine(SVM) models for predicting the classification accuracy. From the results obtained, it is clear that the performance of the proposed work is better improved with SMOTE sampling technique.