Classification of Brain MRI Image of Normal And Convulsion Pediatric By Using Texture Analysis

Abstract

This study is an effort to study brain - white matter (WM), gray matter (GM), cerebrospinal fluid (CSF) -

in MRI images which shows normal appearance if the patients were normal or had convulsion, through

computerized analysis. The main objective of this study was to identify brain tissue as normal and

abnormal “convulsion” pediatric using textural analysis. The textural features were extracted from the

whole MRI image using a window of 3×3pixel wide, where the first order statistics (mean, standard

deviation, Coefficient of variation etc…) were extracted using one pixel interlace. The features in the

window were classified as GM or WM or CSF as well as normal or abnormal using K-means and

discriminate analysis. The images were collected from MRI brain scans for 41 patients presented for

examination in the period from July 2014 to December 2014, 5 patients is control and 36 patients as cases

underwent epilepsy protocol and T1, T2 and FLAIR images obtained. Segmentation analysis and ratios

are compared between structures of the both diseased and control groups. the classification technique

were adopted as a method of pattern identification the image into normal or elliptic classes. Linear

discriminate analysis using steps wise were used to classify the WM, GM and CSF into the predefined

classes normal and convulsion one. The result of this classification showed that the white matter of the

convulsion cases showed a high classification accuracy of 99.5% and demarked as separate class from the

rest of the class’s weather it normal or abnormal. In conclusion the white matter of the convulsion cases

texturally were very different from the normal one while from visual perception point of view they appear

similar