Characterization of Cerebral Stroke in CT Image Using Texture Analysis

Abstract

The aim of this study to characterize the brain tissues using texture analysis technique that depend on the disparity of grey level in CT images. the study consisted of 200 patients with CT brain examinations performed with multidetector helical CT system Toshiba Sensation 64 slice with KVp 120 and mAs 350 in Alzytouna Specialized Hospital in period from December 2014 to August 2017, were the CT images characterize to hemorrhagic stroke and Ischemic Stroke; the image was read by IDL in JIPG format and by clicks on areas represents the brain tissues classified to grey matter, white matter, CSF and hemorrhagic stroke Ischemic Stroke area, with window 3×3 pixel was generated and textural feature for the classes center were generated. These textural features for First order statistics include; (coefficient of variation, stander deviation, variance, signal, energy, and entropy). These features were assigned as classification center using the Euclidian distances to classify the whole image. The result of the study showed that classification Map that created using linear discriminant analysis functions where the three different tissue classes of brain tissues to gray and white matter, CSF and hemorrhagic or stroke area were clearly separated according to calculated texture at P<0.05, and CL=95%. For hemorrhagic area showed that the result of the classification of the hemorrhagic tissues were very different from result of the tissues with classification accuracy of 97.2 %, sensitivity 99.1 % and specificity 96.3 %. And for ischemic area showed that the Gray Level variation and features give classification accuracy of ischemic stroke 97.6%, gray matter 95.2%, white matter 97.3% and the CSF classification accuracy 98.0%. Theoverall classification accuracy of brain tissues 97.0%. Texture analysis depending on the relative attenuation coefficient of tissues i.e. the CT No in HU could serve the diagnostic field and overcoming the visual diagnosis that comes with different interpretation and also would have promising future to avoid invasive technique if the base line for individual tissues being determined and algorithmic aided computer have been applied