Sixty patients clinically confirmed to have MS (17) and 50 controls were scanned with the same MRI unit at a tertiary care facility. None of the patients with MS had other major medical illnesses, were younger than 20 years or older than 60 years, used corticosteroids within 4 weeks, or had a history of substance abuse. Forty-two patients had the relapsing-remitting and 18 had the secondary progressive MS clinical disease course. (18) Physical disability was assessed by the Expanded Disability Status Scale (EDSS) within 1 week of the MRI by a single experienced neurologist blind to the MRI findings. (19) Scores ranged from 0 to 8.0 (mean ± SD, 3.7 ± 1.9). The duration of MS ranged from 0.5 to 38 years (mean ± SD, 10.6 ± 9.4 years). The average number of lifetime courses of high-dose intravenous methylprednisolone taken by all 60 patients with MS was 1.8. Four (7%) were receiving bimonthly intravenous methylprednisolone for disease progression. Controls included normal volunteers recruited from hospital staff and consecutive patients referred to the MRI center for dizziness, headaches, and seizure disorder, who had normal neurologic and MRI findings. An experienced observer reviewed the scans of controls to ensure normal findings and discarded 3 controls due to the presence of bright lesions on T2-weighted images (53 control scans screened, 50 retained). Visual determination of gray matter T2 intensity was not used to exclude control scans. Patients with MS and controls were sex-matched (68% women and 70% women, respectively) and age-matched (mean ± SD age, 42 ± 9 years and 42 ± 10 years, respectively) (P>.9).

MAGNETIC RESONANCE IMAGING

Fast spin-echo T2WI (repetition time [TR]/echo time [TE]/number of signal averages [NSA] = 2300/120/2; 6-mm slice thickness; 0.6 mm slice gap; echo train length 18), fast spin-echo fluid-attenuated inversion-recovery (FLAIR) images (TR/TE/NSA/inversion time = 8000/120/2/2200; 5-mm slice thickness; interleaved; echo train length 20), and T1-weighted images (TR/TE/NSA = 585/20/1, 5-mm slice thickness, interleaved) were obtained in the axial plane on an ACS-NT MRI scanner (Philips Medical Systems, Best, the Netherlands). The in-plane spatial resolution was approximately 1 ± 1 mm. The FLAIR protocol was detailed previously. (20) Images were transferred to Sun workstations (Sun Microsystems, Moutainview, Calif) on which images were analyzed quantitatively at the Buffalo Neuroimaging Analysis Center (Buffalo, NY). A trained observer who was blind to clinical information performed the quantitative MRI analysis. Based on a localization technique, (21) standardized circular regions-of-interest (ROIs) were placed on T2WI in the substantia nigra pars compacta, substantia nigra pars reticulata, red nucleus, anterior thalamus, posterior thalamus, head of the caudate nucleus, and in the cerebrospinal fluid (CSF) of the right lateral ventricular body. To sample the structure while minimizing partial volume effects, the ROIs were 2-mm in diameter for the substantia nigra pars reticulata, pars compacta, and red nucleus and were 5 mm for the other structures. Since the lateral ventricular size varied among subjects, the largest circular ROI (not exceeding 5 mm) was placed in the ventricle without including the adjacent parenchyma or choroid plexus. Freehand ROIs were also manually traced for the putamen and globus pallidus. One axial slice was used for each measurement (the slice showing the largest part of the structure). Care was taken to avoid placing small hyperintensities (MS plaques or perivascular spaces) into ROIs. To correct for potential interscan variations in system scaling and gain, mean signal intensity in each ROI was divided by mean signal intensity of lateral ventricular CSF as adapted from Pujol et al. (22) Because T2 intensity was expressed as a ratio to the right lateral ventricular CSF, it is important to note that absolute right vs left CSF intensities did not differ in patients (P = .44) or controls (P = .26), and neither right (P = .44) nor left (P = .27) absolute CSF intensities differed between the groups. The data from the anterior and posterior thalami ROIs were averaged to produce a value for the thalamus. The data from the substantia nigra pars reticulata and pars compacta ROIs were averaged to produce a value for the substantia nigra. Separate measurements were taken from each hemisphere, which were then collapsed across both hemispheres by calculating the mean in each ROI (to reduce the number of statistical variables).

We assessed the validity of right/left collapsing of the variables and assessed the symmetry of the gray matter signal changes. The Pearson correlation r values between right and left T2 intensities were 0.92 for the caudate, 0.88 for the thalamus and putamen, 0.90 for the red nucleus, 0.87 for the substantia nigra, and 0.92 for the globus pallidus. Paired sample t tests showed no significant right vs left difference for the caudate, thalamus, and red nucleus. While the putamen, substantia nigra, and globus pallidus showed significant right vs left differences in T2 intensities, the effect sizes were quite small: 0.3 or less (effect sizes = difference between means divided by pooled SDs). Thus, in general, the T2 intensity in the gray matter ROIs showed symmetry in the MS group.

After collapsing the ROIs, a total of 6 ROIs were used for further analysis (substantia nigra, red nucleus, thalamus, caudate, putamen, globus pallidus). The typical ROI placement in a patient with MS is shown in Figure 1. The fast spin-echo T2 sequence was used in the analysis of T2 hypointensity. Previous studies analyzing iron deposition have used conventional spin-echo to detect T2 shortening. (21-24) In this study, we used fast spin-echo T2 because of the faster scanning time and increased clinical utility in the evaluation of patients with MS (see "Comment" section).

The total T2 hyperintense parenchymal plaque lesion load was determined by manual tracing of lesions on FLAIR images and was the sum of the volume of each lesion seen on each FLAIR axial slice (area multiplied by slice thickness, nongapped images); artifacts and other normal hyperintensities seen in the normal population on FLAIR images were avoided. (20) To assess central atrophy, third ventricular width was measured from FLAIR images using our previously established method. (25) The analysis of gray matter and CSF T2 intensity, T2 hyperintense lesion volume, and third ventricular width was performed using EasyVision software (Release 2.1.2; Philips Medical Systems, Best, the Netherlands). Since hypointense T1 lesions are difficult to delineate manually, the analysis of T1 lesion volume was performed using a semiautomated edge finding and local thresholding technique (Java Image, Version 1.0; Xinapse Systems, Leicester, England [http://www.xinapse.com]). The operator clicks on the edge of hypointense area and the program examines a region 5 x 5 pixels around the mouse click and computes the maximum intensity gradient within the region. The pixel with the highest intensity gradient is then used as the starting point for contour following, thus outlining the region where the intensity is locally lower than at the starting pixel.

RELIABILITY

The same individual reanalyzed the MRI scans of 10 randomly chosen patients with MS at least 2 weeks after the initial analysis. Intraobserver coefficients of variation for the T2 intensity measurements ranged from 1.0% to 2.7% as follows: caudate, 1.2% (right)/1.6% (left); anterior thalamus, 2.1%/1.2%; posterior thalamus, 1.5%/2.2%; red nucleus, 1.4%/1.8%; substantia nigra-pars reticulata, 2.4%/2.7%; pars compacta, 1.3%/1.8%; putamen, 1.3%/1.1%; globus pallidus, 1.0%/1.0%; ventricular CSF, 1.2%. The intraobserver coefficients of variation were 1.2% for total T2 hyperintense parenchymal lesion volume, 1.7% for total T1 hypointense parenchymal lesion volume, and 5.7% for third ventricular width. A second trained observer analyzed the same 10 patients for gray matter and CSF T2 intensity; the interobserver coefficients of variation ranged from 0.6% to 2.9%. To test the stability of the T2 intensity measurement technique, 2 healthy volunteers, aged 26 (man) and 31 (woman) years, each underwent the MRI protocol twice (1 week apart). The scan/rescan intrasubject coefficients of variation for the various gray matter intensities ranged from 1.2% to 2.5%.

ANALYSIS

Group differences were assessed by independent sample t tests. The Pearson r statistic was used for correlations between continuous variables and the Spearman rank correlation test was used to compare continuous data with ordinal ratings. We used a conservative threshold for statistical significance (P<.01) in all univariate comparisons and controlled for multiple correlations via regression models. Within the MS group, linear and logistic regression models were used to predict conventional MRI findings (total hyperintense parenchymal lesion volume, third ventricular width) and clinical parameters (EDSS, disease duration, relapsing-remitting [secondary progressive] clinical course) using only T2 intensity measures as predictors that differed significantly between patients with MS and controls (ie, abnormal T2 hypointensity). All models controlled for age by entering age in block 1 and holding age in the final model. Otherwise, each model used a forward stepwise selection procedure, with P to enter .05 and P to exit .10. Two types of analyses were performed. First, regression models with individual gray matter ROI T2 intensities were fitted to detect significant predictors of total T1 or T2 parenchymal lesion volume, third ventricular width, or clinical parameters while controlling only for age. Second, total T1 and T2 parenchymal lesion volume and third ventricular width were added to the clinical parameter models.

RESULTS