An Efficient Automatic Brain Tumor Classification Using LBP Features and SVM-Based Classifier

Abstract

Brain tumor detection is a tedious task which involves a lot of time and expertise. With each passing year, the world has always witnessed an increase in the number of cases of brain tumor. It is thereby apparent; that it is becoming difficult for the doctors to detect tumors in MRI scans, not only because of the increase in numbers but also, because of the complexity of the cases. So, the research in this domain is still ongoing as the world is in search of an exemplary and flawless method for an automated brain tumor detection technique. In this paper, we introduced a novel architecture for brain tumor detection which detects whether the given MR image is malignant or benign. Preprocessing, segmentation, dimension reduction, and classification are the major phases of our proposed architecture. On the MR images, T2-weighted preprocessing is applied to convert into grayscale images. In the next stage, features are extracted from the preprocessed images by applying local binary pattern (LBP) technique. Principal component analysis (PCA) is used to discard uncorrelated features. This reduced feature set is fed to the support vector machine (SVM) classifier to predict whether the given MR image is normal (benign) or abnormal (malignant). Experimental results on benchmark MR image datasets exhibit that the proposed method gives promising accuracy when compared to the existing work though it is simple.