Voxel-based morphometry detects patterns of atrophy that help differentiate progressive supranuclear palsy and Parkinson's disease

Abstract

Progressive supranuclear palsy (PSP) and Parkinson's disease (PD) are neurodegenerative diseases with distinctive pathological appearances. Early clinical diagnosis can be difficult. MRI may help differentiate PSP from PD, but the differences are often only obvious with advanced disease. It would be useful to have an unbiased assessment of difference to guide visual assessment of MRI as an aid to clinical diagnosis. Voxel-based morphometry (VBM) offers nonbiased, observer-independent morphometric MRI analysis. Our objectives were to assess structural differences between PSP, PD, and normal controls and test the clinical utility of the results. T1-weighted MR images in 12 patients with clinically diagnosed PSP, 12 with PD, and 12 age- and sex-matched controls were normalized to a common stereotaxic space and segmented into gray matter (GM) and white matter (WM) then analyzed using VBM. MRI scans were reviewed by a neuroradiologist blinded to the clinical diagnosis and assigned to the “non-PSP” or “PSP” group based on regional differences highlighted using VBM. VBM revealed significant group differences between PSP and PD as well as PSP and controls, with tissue reduction demonstrated in the region of the cerebral peduncles and midbrain. With these regional differences as a guide, neuroradiological diagnosis achieved a sensitivity of 83% and a specificity of 79%. VBM did not detect dramatic changes in frontal regions despite significant frontal cognitive decline in the PSP group. Pathology in the basal ganglia rather than tissue loss in the frontal lobes could be responsible for this. This information may help in the differentiation of PSP in clinical practice.

Introduction

Progressive supranuclear palsy (PSP) is a neurodegenerative condition which in its classical form presents with postural instability and falls, a supranuclear gaze palsy with early abnormalities of saccadic eye movements, symmetrical bradykinesia, and axial rigidity Lees, 1987, Steele et al., 1964. PSP is characterized pathologically by the presence of tau-positive neurofibrillary tangles, neuropil threads, tufted astrocytes and oligodendroglial coiled bodies in brainstem structures, including the midbrain, cerebellum, and basal ganglia as well as premotor and motor cortical areas (Verny et al., 1996). In the early stages or when the disease is atypical, it can be difficult to distinguish from other neurodegenerative diseases such as multiple system atrophy (MSA), corticobasal degeneration (CBD), and Parkinson's disease (PD).

Various imaging techniques have been used to study PSP and its clinical differentials. Early studies using MRI highlighted signal change and atrophy in the putamen and midbrain as well as an increase in third ventricular size as markers of the disease Drayer et al., 1986, Savoiardo et al., 1994. More recent cross-sectional imaging studies have retrospectively tested the use of these measures as potential aids to diagnosis Schrag et al., 2000, Warmuth-Metz et al., 2001. Over 70% of clinically diagnosed PSP cases can be correctly identified using these markers. Diffusion-weighted MRI studies of the basal ganglia can differentiate cases of PD and PSP with a sensitivity of 90% (Seppi et al., 2003).

The majority of imaging studies have relied on visual observations or linear measurements of regions chosen on the basis of previous imaging work or knowledge of the areas pathologically affected at postmortem in PSP. These studies rely on prior assumptions of regional MRI abnormalities, and hence an inherent bias is introduced by selecting a limited number of brain regions for study.

An MRI comparison of the local composition of different brain tissue types prioritizes a measure of structural differences between subjects. Voxel-based morphometry (VBM) is an unbiased technique designed to detect statistically significant differences in these tissues between subject groups (Ashburner and Friston, 2000). VBM removes positional and global volume differences through spatial normalization and then detects differences in tissue density by comparing local intensities of tissue after smoothing. VBM avoids some of the shortcomings of other methods by being operator-independent and semiautomated, hence avoiding a priori assumptions as to which brain areas should be assessed. VBM has recently demonstrated frontal atrophy in PSP when compared to healthy controls (Brenneis et al., 2004). No studies comparing PSP to another bradykinetic rigid syndrome have been undertaken, and the potential application of VBM results in clinical practice has not been explored. Our first objective was to apply VBM analysis of MRI in PSP and PD to study the morphological alterations in PSP- and PD-affected brains in vivo.

While VBM is cited as supportive of region of interest studies Maguire et al., 2000, Vargha-Khadem et al., 1998 and as a method of identifying affected brain areas with a view to new avenues of research, the clinical utility of VBM findings in bradykinetic rigid syndromes has not been evaluated. Our second objective was to apply the VBM findings as a guide to neuroradiological distinction of PSP from PD and controls.