Article Preview
Top1. Introduction
Magnetic resonance (MR) imaging is currently an indispensable diagnostic imaging technique in the study of the human brain (Neeraj et al., 2010). It’s a non-invasive technique that provides fairly good contrast resolution for different tissues and generates an extensive information pool about the condition of the brain. Such information has dramatically improved the quality of brain pathology diagnosis and treatment. However this big amount of data makes manual interpretation impossible and necessitates the development of automated image analysis tools. Computing technologies and systems may be classifed into the categories of imperative, autonomic, and cognitive from the bottom up according to theories of cognitive informatics (Wang, 2009).
There is a variety of automated diagnostic tools that are developed by applying sophisticated signal/image processing techniques utilizing transforms and, may be, subsequently applying some computational intelligent techniques. In one possible methodology, the process of automatic segregation of normal/abnormal subjects, based on brain MRIs, is illustrated as a three-step process: feature extraction, feature selection and nonlinear classification.
To extract features from the MR brain images several image analysis methods are used: e.g. Gabor filters, Independent Component Analysis (ICA) (Moritz et al., 2000), techniques employing statistical feature extraction (like mean, median, mode, quartiles, standard deviation, kurtosis, skewness, etc.) (Begg et al., 2005), Fourier Transform (FT) based techniques (Bracewell, 1999), Wavelet Transform (WT) based techniques (Mallat, 89; Kharrat et al., 2009), etc. while Fourier Transform provides only frequency analysis of signals, Wavelet Transforms provide time-frequency analysis, which makes it a useful tool for time-space-frequency analysis and particularly for pattern recognition.
We use Genetic Algorithm (GA) to find minimum features subset giving optimum discrimination between extracted features. GA proves to be the most efficient compared with classical algorithms (Siedlecki et al., 1989) including sequential forward selection (SFS), sequential backward selection (SBS), sequential floating forward selection (SFFS) and sequential floating backward selection (SFBS).
We apply machine learning algorithms to obtain the classification of images under two categories, either normal or abnormal (Chaplot et al., 2006; El-Dahsan et al., 2009; Zacharaki et al., 2009). Support Vector Machines (SVMs) are widely used for classification tasks due to their appealing generalization properties and their computational efficiency.
The rest of the paper is organised as follows. Section 2 presents the Wavelet transform for feature extraction. Section 3 is devoted for feature selection employed for Genetic Algorithm. Image Classification is presented in Section 4. The performance evaluation, the feasibility and superiority of the proposed approach is conducted in Section 5. Finally, the section 6 presents our conclusions.