A hybrid method for MRI brain image classification

https://doi.org/10.1016/j.eswa.2011.02.012Get rights and content

Abstract

Automated and accurate classification of MR brain images is of importance for the analysis and interpretation of these images and many methods have been proposed. In this paper, we present a neural network (NN) based method to classify a given MR brain image as normal or abnormal. This method first employs wavelet transform to extract features from images, and then applies the technique of principle component analysis (PCA) to reduce the dimensions of features. The reduced features are sent to a back propagation (BP) NN, with which scaled conjugate gradient (SCG) is adopted to find the optimal weights of the NN. We applied this method on 66 images (18 normal, 48 abnormal). The classification accuracies on both training and test images are 100%, and the computation time per image is only 0.0451 s.

Highlights

► We proposed an MR brain image classifier whose accuracy is 100%. ► SCG is used to find the optimal weights of NN. ► The computation time per image is 0.0451s.

Introduction

Magnetic resonance imaging (MRI) is an imaging technique that produces high quality images of the anatomical structures of the human body, especially in the brain, and provides rich information for clinical diagnosis and biomedical research. The diagnostic values of MRI are greatly magnified by the automated and accurate classification of the MRI images. In recent years, researchers have proposed a lot of approaches for this goal, which fall into two categories.

One category is supervised classification, such as support vector machine (SVM) (Chaplot, Patnaik, & Jagannathan, 2006) and k-nearest neighbors (k-NN) (Fletcher-Heath, Hall, Goldgof, et al., 2001). The other category is unsupervised classification, such as self-organization feature map (SOFM) (Chaplot et al., 2006) and fuzzy c-means (Maitra & Chatterjee, 2008). While all these methods achieved good results, and the supervised classification performs better than unsupervised classification in terms of classification accuracy (success classification rate).

The goal of this study was to design a more efficient and accurate classifier to distinguish normal and abnormal brain MRIs. Moreover, we will present a detailed description of the classifier so that other researchers test and validate our method.

Our proposed method first employs wavelet transform to extract features from MRIs. Second, principle component analysis (PCA) technique was applied to reduce the dimensions of features. Third, the reduced features are sent to a back propagation neural network (BPNN), where scaled conjugate gradient (SCG) is adopted to find the optimal weights of the BPNN.

The structure of this paper is organized as follows. A short description of our method is presented in Section 2. Wavelet transform based methods for feature extraction is presented in Section 3. Principle component analysis technique for feature reduction is introduced in Section 4. Section 5 briefs the structure and training approach of the back propagation neural network. Experiments in Section 6 demonstrate the effectiveness and rapidness of our proposed algorithm. We conclude this paper in Section 7.

Section snippets

Methodology

Our method consists of three stages as shown in Fig. 1: feature extraction, feature reduction, and NN-based classification. This is a canonical and standard classification method which has already been proven as the best classification method (Ghosh, Shankar, & Meher, 2009).

Advantages of wavelet transform

The most conventional tool of signal analysis is Fourier transform (FT), which breaks down a time domain signal into constituent sinusoids of different frequencies, thus, transforming the signal from time domain to frequency domain. However, FT has a serious drawback. It lost the time information of the signal. For example, analyst can not tell when a particular event took place from a Fourier Spectrum. Thus, the classification will decrease as the time information is lost.

Gabor adapted the FT

Feature reduction

Excessive features increase computation times and storage memory. Furthermore, they sometimes make classification more complicated, which is called the curse of dimensionality. It is required to reduce the number of features.

Principal component analysis (PCA) is an efficient tool to reduce the dimension of a data set consisting of a large number of interrelated variables while retaining most of the variations. It is achieved by transforming the data set to a new set of ordered variables

Structure

Neural networks are widely used in pattern classification since they do not need any information about the probability distribution and the a priori probabilities of different classes. A single-hidden-layer backpropagation neural network is adopted with sigmoid neurons in the hidden layer and linear neuron in the output layer.

The training vectors were presented to the NN, which is trained in batch mode. The network configuration is NI × NH × 1, i.e., a two-layer network with NI input neurons, NH

Experiments and discussions

The experiments were carried out on the platform of P4 IBM with 3 GHz main frequency and 2G memory, running under Windows XP operating system. The algorithm was developed via the wavelet toolbox, the neural network toolbox, and the statistical toolbox of Matlab 2009b (The Mathworks ©). The programs can be run or tested on any computer platforms where Matlab is available.

Conclusions

In this paper we have developed a novel hybrid classifier to distinguish normal and abnormal brain MRIs. The method obtained 100% classification accuracy on both training and test images of the selected datasets, and the computation time for each image is only 0.0451 s.

Future work should focus on the following aspects. First, the proposed method should be employed for MRI images with other contrast mechanisms such as T1-weighted, proton-density-weighted, and diffusion-weighted MRIs. Second, the

Acknowledgments

The research was conducted with following supports: (1) National Natural Science Foundation of China (#60872075); (2) National Technical Innovation Project Essential Project Cultivate Project (#706928) and (3) Nature Science Fund in Jiangsu Province (#BK2007103). We gratefully appreciate to Dr. Zhenyu Zhou who reviewed the whole paper.

References (15)

There are more references available in the full text version of this article.

Cited by (301)

View all citing articles on Scopus
View full text