An efficient magnetic resonance brain image classifier using tetrolet transform and kernel support vector machine based on OTSU binarization

Abstract

Classification of tumor from cancer causing to non-cancer causing, it plays a main role in diagnosing the disease effectively without any flaws. In this proposed paper, a novel methodology is presented in order to classify a given Magnetic resonant brain image as normal or abnormal using OTSU binarization segmentation with help of tetrolate transform. By replacing wavelet transform with tetrolate transform the classification is made efficient in case of images with geometric shapes. A good number of features are extracted by using OTSU binarization from edge-based segmentation, the more number of features makes the classification for effective and accurate. The image is finely segmented pixel by pixel for good accuracy and about 12 features like We calculate four different type of accuracy like RBF (Radial Basis Function), linear, polygonal and quadratic based on image features. We performed our proposed methods with four different kernels LIN (Linear), HPOL (Homogeneous Polynomial), and IPOL and GRB (Gaussian Radial Basis function) kernel to achieve the highest classification accuracy. The work is added with advancement by using Graphical user interface (GUI), which makes the user comfortable in accessing the method where most of the users are from clinical background and are not aware of any software and their usag.