Correction for intracranial volume in analysis of whole brain atrophy in multiple sclerosis: the proportion vs. residual method