Brain MRI Slices Classification Using Least Squares ... - CiteSeerX

Vol. 1, No. 1, Issue 1, Page 21 of 33 Copyright © 2007, TSI® Press Printed in the USA. All rights reserved

Brain MRI Slices Classification Using Least Squares Support Vector Machine H. Selvaraj1, S. Thamarai Selvi2, D. Selvathi3, L. Gewali1 1

University of Nevada, Las Vegas, NV, USA Anna University, MIT Campus, Chennai, India 3 Mepco Schlenk Engineering College, Sivakasi, India 2

Received 1 January 2007; revised 2 February 2007, accepted 3 March 2007 Abstract This research paper proposes an intelligent classification technique to identify normal and abnormal slices of brain MRI data. The manual interpretation of tumor slices based on visual examination by radiologist/physician may lead to missing diagnosis when a large number of MRIs are analyzed. To avoid the human error, an automated intelligent classification system is proposed which caters the need for classification of image slices after identifying abnormal MRI volume, for tumor identification. In this research work, advanced classification techniques based on Least Squares Support Vector Machines (LS-SVM) are proposed and applied to brain image slices classification using features derived from slices. This classifier using linear as well as nonlinear Radial Basis Function (RBF) kernels are compared with other classifiers like SVM with linear and nonlinear RBF kernels, RBF classifier, Multi Layer Perceptron (MLP) classifier and K-NN classifier. From this analysis, it is observed that the proposed method using LSSVM classifier outperformed all the other classifiers tested. Keywords Classification, MRI, LS-SVM, SVM, RBF, MLP, K-NN.

1. INTRODUCTION The field of medical imaging gains its importance with increase in the need of automated and efficient diagnosis in a short period of time. Computer and Information Technology are very much useful in medical image processing, medical analysis and classification. Medical images are usually obtained by X-rays and recent years by Magnetic Resonance (MR) imaging. Magnetic Resonance Imaging (MRI) is used as a valuable tool in the clinical and surgical environment because of its characteristics like superior soft tissue differentiation, high spatial resolution and contrast. It does not use harmful ionizing radiation to patients [1, 2]. Magnetic Resonance Images are examined by radiologists based on visual interpretation of the films to identify the presence of tumor abnormal tissue. The shortage of radiologists and the large volume of MRI to be analyzed make such readings labor intensive, cost expensive and often inaccurate. The sensitivity of the human eye in interpreting large numbers of images decreases with increasing number of cases, particularly when only a small number of slices are affected. Hence

there is a need for automated systems for analysis and classification of such medical images. The MRI may contain both normal slices and defective slices. The defective or abnormal slices are identified and separated from the normal slices and then these defective slices are further investigated for the detection of tumor tissues. Matthew C. Clarke et al. [3] developed a method for abnormal MRI volume identification with slice segmentation using Fuzzy C-means (FCM) algorithm. Luiza Antonie [4] proposed a method for Automated Segmentation and Classification of Brain MRI in which an SVM classifier was used for normal and abnormal slices classification with statistical features. The latest development in data classification research has focused more on Least Squares Support Vector Machines (LS-SVMs) because several recent studies have reported that LS-SVM generally are able to deliver higher classification accuracy than the other existing data classification algorithms [5][6]. In this paper, the potential benefit of using an LSSVM based approach with linear and nonlinear type of kernels are investigated for the automated

dimensional space. The dominant feature which makes SVM very attractive is that classes which are nonlinearly separable in the original space can be linearly separated in the higher dimensional feature space. Thus SVM is capable of solving complex nonlinear classification problems. Important characteristics of SVM are its ability to solve classification problems by means of convex quadratic programming (QP) and also the sparseness resulting from this QP problem. The learning is based on the principle of structural risk minimization. Instead of minimizing an objective function based on the training samples (such as mean square error), the SVM attempts to minimize the bound on the generalization error (i.e., the error made by the learning machine on the test data not used during training). As a result, an SVM tends to perform well when applied to data outside the training set. SVM achieves this advantage by focusing on the training examples that are most difficult to classify. These “borderline” training examples are called support vectors. A least squares version of SVM (LS-SVM) is introduced by Suykens in [19, 20] with the idea of modifying Vapnik’s SVM formulation by adding a least squares term in the cost function. This variant circumvents the need to solve a more difficult QP problem and only requires the solution of a set of linear equations. This approach significantly reduces the complexity and computation in solving the problem.

classification of brain MRI slices. This is for separating abnormal slices from the data collection containing both normal and abnormal slices. The purpose is to perform segmentation process for tumor calculation only on abnormal slices. This results in significant cost and time saving by avoiding the segmentation task for large number of normal slices. The categorization of slices into normal and abnormal is done using statistical features of images such as mean, variance, and cooccurrence based textural features of images such as energy, entropy, difference moment, inverse difference moment and correlation. For comparative analysis, SVM Classifier with linear and nonlinear type of kernels, the RBF, MLP and K-NN classifiers are also implemented using the same data sets. The motivation behind this paper is to develop a machine classification process for evaluating the classification performance of different classifiers to this problem in terms of statistical performance measure. The paper is organized as follows. Preliminaries dealing with LS-SVM techniques are presented in Section 2. Section 3 discusses the proposed methodology using LS-SVM for MRI image slices classification. Implementation of the proposed approach is given in Section 4. The performance of different classifiers is discussed in Section 5. Finally, the conclusion is presented in Section 6.

In this paper, we treat slice classification as a two class pattern classification problem. We apply all the MRI slices to classifier to determine whether the tumor is present or not. We refer to these two classes throughout as “normal” and “abnormal”

2. REVIEW OF LS-SVM LEARNING FOR CLASSIFICATION The Support Vector Machine algorithm was first developed in 1963 by Vapnik and Lerner [7] and Vapnik and Chervonenkis [8] as an extension of the Generalized Portrait algorithm. This algorithm is firmly grounded in the framework of statistical learning theory – Vapnik Chervonenkis (VC) theory, which improves the generalization ability of learning machines to unseen data [9] [10]. In the last few years Support Vector Machines have shown excellent performance in many real-world applications including hand written digit recognition [11], object recognition [12], speaker identification [13], face detection in images [14] and text categorization [15]. SVM is a classification algorithm based on kernel methods [16], [17] and [18]. In contrast to linear classification methods, the kernel methods map the original parameter vectors into a higher (possibly infinite) dimensional feature space through a nonlinear kernel function. Without the need to compute the nonlinear mapping explicitly, dotproducts can be computed efficiently in higher

slices. Let vector x ∈ R denote a pattern to be classified, and let scalar y denote its class label y ∈ {± 1} ). In addition, let (i.e., n

{(xi , yi ), i = 1,2,......, l} denote a given set of l

training examples. The problem is how to construct a classifier [i.e., a decision function f(x)] that can correctly classify an input pattern x that is not necessarily from the training set.

Linear SVM classifier Let us begin with the simplest case, in which the training patterns are linearly separable. That is, there exists a linear function of the form f(x) = w T x + b

(1)

such that for each training example xi, the function yields f ( xi ) ≥ 0 for y i = +1, and f(xi)