Classification of magnetic resonance brain images using bi-dimensional empirical mode decomposition and autoregressive model

Abstract

Purpose: Automated classification of brain magnetic resonance (MR) images has been an extensively researched topic in biomedical image processing. In this work, we propose a new approach for classifying normal and abnormal brain MR images using bi-dimensional empirical mode decomposition (BEMD) and autoregressive (AR) model Methods: In our approach, brain MR image is decomposed into four intrinsic mode functions (IMFs) using BEMD and AR coefficients from multiple IMFs are concatenated to form a feature vector. Finally a binary classifier, least-squares support vector machine (LS-SVM), is employed to discriminate between normal and abnormal brain MR images. Results: The proposed technique achieves 100% classification accuracy using second-order AR model with linear and radial basis function (RBF) as kernels in LS-SVM. Conclusions: Experimental results confirm that the performance of the proposed method is quite comparable with the existing results. Specifically, the presented approach outperforms one-dimensional empirical mode decomposition (1D-EMD) based classification of brain MR images. © 2015, Korean Society of Medical and Biological Engineering and Springer.