Characterization of White Matter Lesions on Brain Magnetic Resonance Images Using Texture Analysis

Abstract

Texture analysis studies have been increasingly explored in neuroradiology in the recent yeares, as it enabling diseases characterization and quantification of disease distribution. The aim of this study was to characterize white matter lesions on the brain in MR images using texture analysis. This analytical study had been conducted at Antalya medical center in the period from October 2018 to March 2021 by using study of 1646 brain MR images selected conveniently for patients having Glioma, Multiple Sclerosis (MS) and Small Vessels Diseases (SVD) , then these imeges entered to interactive data language program to extract the textural features from the normal tissues and the white matter lesions, then these extracted features entered to statistical package social science for analysis. And the results reveal that; the best textural features for discrimination of the Glioma from the normal tissues in all MR imaging sequences were the mean, entropy, Gray-Level Nonuniformity and Run-Length Nonuniformity. While for the MS plaques the features were the mean, variance, entropy and Run Percentage textural features. Also the best features for differentiation the SVD from normal tissues in all MRI sequences were the mean, variance, energy, entropy, Gray-Level Nonuniformity and Run Percentage. In conclusion texture analysis statistics successfully discriminate the white matter lesions from normal brain tissues in all MRI sequences, firstly when using the first order statistics on the Glioma, MS and SVD the imaging sequence that shows the highest sensitivity in discrimination the lesion from normal tissues were T2=98.2%, T1+C=99.3% and T1+C=97.5% respectively; and when using the higher order statistics the highest sensitivity were in T1+C=93.6%, T1+C=99.3% and T1=97.5% correspondingly.