There have been several methods proposed so far using diffusion tensor imaging (DTI) for the assessment of normal-appearing brain tissue (NABT) injury in multiple sclerosis (MS). However, for these methods, the analyses of the NABT injury at the cellular level, wherein histological examinations can be used, still present challenging problems. We developed a method of segregating NABT into the following anatomical structures using lambda chart analysis associated with a two-dimensional Gaussian deconvolution of diffusion characteristic functions: 1) structures primarily composed of small neurons and glia; 2) structures primarily composed of large neurons; 3) structures primarily composed of short axons; and 4) structures primarily composed of long axons. Each segregated structure that had a distinctive diffusion characteristic was subjected to the statistical inference of DTI-derived parameters for 14 patients with conventional relapsing-remitting MS (RRMS) and 20 age-matched healthy volunteers. In all of the structures, the trace values were significantly higher and the fractional anisotropy values were significantly lower in the RRMS patients than in the healthy volunteers. Furthermore, the volume fractions of the structures primarily composed of short axons markedly decreased, whereas those of the structures primarily composed of small neurons and glia markedly increased. These results suggest that axonal loss and glial proliferation predominantly occurred in the subcortical white matter and adjacent deep cortical layer, namely, the juxtacortical region. This cell-oriented analysis of NABT injury using DTI confirmed in vivo the histological observation that the juxtacortical region is the most vulnerable site in MS. (C) 2007 Elsevier Inc. All rights reserved.