Effect of Slice Thickness on Texture Analysis of Multiple Sclerosis

Abstract

Texture analysis accuracy of images acquired by magnetic resonance imaging depends on various image quality parameters, such as magnetic field strength, imaging sequence, noise, spatial resolution and slice thickness. In this paper, the effect of slice thickness on texture analysis of histologically confirmed multiple sclerosis subjects is studied by statistical means. Original consecutive 1 mm slices have been averaged by Matlab to simulate 3 mm slices at the same location. Both slice types are analyzed by MaZda, a texture analysis tool, to calculate 292 texture parameters. Wilcoxon’s signed rank test is applied to test differences in texture parameters between regions of interest. Our results indicate that tissue separability is nearly equal with the studied slice thicknesses, depending slightly on the imaging sequence. Most of the texture parameters are well separable with both slice thicknesses.