© RSNA, 2013
Purpose: To investigate the association between the development of thalamic and cortical atrophy and the conversion to clinically definite multiple sclerosis (CDMS) in patients with clinically isolated syndrome (CIS).
Materials and Methods: This prospective study was approved by the institutional review board. Informed consent was given by 216 CIS patients, and patients were treated with 30 µg of intramuscular interferon β1a once a week. They were assessed with a magnetic resonance (MR) imaging examination at baseline, 6 months, 1 year, and 2 years. Patients were evaluated within 4 months of an initial demyelinating event, had two or more brain lesions on MR images, and had two or more oligoclonal bands in cerebrospinal fluid. MR imaging measures of progression included cumulative number and volume of contrast agent–enhanced (CE) new and enlarged T2 lesions, and changes in whole-brain, tissue-specific global, and regional gray matter volumes. Regression and mixed-effect model analyses were used.
Results: Over 2 years, 92 of 216 patients (42.6%) converted to CDMS; 122 (56.5%) CIS patients fulfilled McDonald 2005 criteria and 153 (70.8%) fulfilled McDonald 2010 criteria for MR imaging dissemination in time and space. The mean time to first relapse was 3.1 months, and mean annual relapse rate was 0.46. In mixed-effect model analysis, the lateral ventricle volume (P = .005), accumulation of CE (P= .007), new total T2 (P = .009) and new enlarging T2 lesions (P = .01) increase, and thalamic (P = .009) and whole-brain (P = .019) volume decrease were associated with development of CDMS. In multivariate regression analysis, decrease in thalamic volumes and increase in lateral ventricle volumes (P = .009) were MR imaging variables associated with the development of CDMS.
Conclusion: Measurement of thalamic atrophy and increase in ventricular size in CIS is associated with CDMS development and should be used in addition to the assessment of new T2 and CE lesions.
MR Imaging Acquisition and Analysis
MR imaging was performed at baseline, 6 months, 1 year, and 2 years by using standardized protocol with a 1.5-T imager (Gyroscan; Philips Medical Systems, Best, the Netherlands). Axial brain images were obtained by using fluid-attenuated inversion recovery (FLAIR) with 1.5-mm thickness (repetition time msec/echo time msec/inversion time msec, 11000/140/2600; flip angle, 90°; 256 × 181 matrix). Axial T1-weighted three-dimensional spoiled gradient-recalled images were acquired with 1-mm section thickness (25/5; flip angle, 30°; 256 × 204 matrix). Both FLAIR and spoiled-gradient recalled images had no gaps. In addition, patients underwent postcontrast T1 spin-echo 3-mm section thickness imaging 5 minutes after injection of a single dose of 0.1 mmol/kg of Gd-DTPA contrast agent (12/450).
The T2 and number of CE lesions were obtained by using FLAIR, and lesion volumes were measured on T1 postcontrast images with a semiautomated edge detection contour and threshold technique previously described (34). By using software (FMRIB Linear Image Registration Tool; Oxford Centre for Functional Magnetic Resonance Imaging of the Brain, University of Oxford, Oxford, England; http://www.fmrib.ox.ac.uk/fsl) (35), all follow-up FLAIR and T1 spin-echo postcontrast images for a given subject were coregistered to the baseline FLAIR image by using a six-degrees-of-freedom rigid-body model. All subsequent lesion analysis was performed by using coregistered images. For each time point, T2 lesion activity analysis was performed with the aid of a subtraction image: the image from the previous time point was subtracted from the corresponding current image. The result was then smoothed with a Gaussian kernel of 0.5 mm. Cross-sectional regions of interest were overlaid on the subtraction image to facilitate the identification of new and enlarging T2 lesions. The cumulative number of T2 and CE lesions was obtained by summing the total number of these lesions from baseline to 6 months, 6 months to 1 year, and 1 to 2 years of the study.
For baseline analyses of global and tissue-specific atrophy measures, software was used (SIENAX, version 2.6; Oxford Centre for Functional Magnetic Resonance Imaging of the Brain) (36), and corrections for T1 hypointensity misclassification were performed with an in-painting program, developed in house, on T1-weighted three-dimensional spoiled gradient-recalled images. Normalized volumes of whole-brain, gray matter, white matter, cortex, and lateral ventricles were measured, as described previously (37).
For longitudinal changes of the whole-brain volume, we applied software (SIENA; Oxford Centre for Functional Magnetic Resonance Imaging of the Brain) to calculate the percentage change in brain volume (36). To quantify longitudinal gray matter, white matter, cortical, and lateral ventricle percentage change in volume, we used a modified hybrid of SIENA and SIENAX software (Oxford Centre for Functional Magnetic Resonance Imaging of the Brain). We used a brain and skull-constrained coregistration technique to place both baseline and follow-up images into a joint space that was halfway between the two images at all time points in the study. We then combined the baseline and follow-up intracranial volume masks via union and valid voxel masks via intersection, which ensured that the same imaging volume was analyzed at both time points. Finally, we segmented the images with a modified longitudinal software (FMRIB Automated Segmentation Tool; Oxford Centre for Functional Magnetic Resonance Imaging of the Brain) (38) that used a four-dimensional random Markov field to prevent misclassification between time points when longitudinal intensity changes were lacking or minimal. For each tissue compartment, total tissue volume was calculated from partial volume maps for both baseline and follow-up, and percentage volume change was derived directly from the images.
Absolute tissue volumes for the total subcortical deep gray matter, caudate nucleus, putamen, globus pallidus, thalamus hippocampus, nucleus accumbens, and amygdala at each time point were estimated from in-painted T1-weighted three-dimensional spoiled gradient-recalled images with a model-based segmentation and registration tool (FMRIB Integrated Registration and Segmentation Tool, version 1.2; Oxford Centre for Functional Magnetic Resonance Imaging of the Brain) (39). Normalized volumes were obtained by multiplying the estimated volumes from this tool by the volumetric scaling factor from SIENAX, and percentage volume changes were obtained between different time points (37).