Correlates of MS disability assessed in vivo using aggregates of MR quantities

Patients.

We studied 23 patients (17 women and six men), 21 of whom had relapsing-remitting and two secondary progressive MS.13 Their mean age was 37.1 years (SD, 8.9 years), median duration of the disease was 5 years (range, 1 to 21) and median Expanded Disability Status Scale (EDSS) score14 was 2.0 (range, 0.0 to 6.0). All patients were relapse- and steroid-free for at least 3 months before study entry; 15 of those with relapsing-remitting MS were being treated with interferon beta. On the day MRI scans were obtained, patients were assessed clinically by a single neurologist who was unaware of the MRI results. Local Ethical Committee approval and written informed consent from all the subjects were obtained before study initiation.

Image acquisition.

All patients underwent scanning on two occasions separated by an interval of about 72 hours. During the first session, several scans were obtained using a 1.5-T scanner. First, we performed dual-echo turbo spin echo (repetition time [TR], 3,300 ms; echo time [TE], 16/98 ms; echo train length, 5). Second, two-dimensional gradient echo (TR, 600 ms; TE, 12 ms; flip angle, 20°) with and without an off-resonance radiofrequency saturation pulse (offset frequency, 1.5 kHz; Gaussian envelope duration, 7.68 ms; flip angle, 500°) was done.

Third, pulsed gradient spin-echo single-shot echoplanar pulse sequence (PGSE-EPI) (interecho spacing, 0.8 ms; TE, 123 ms) was performed, with diffusion gradients applied in eight noncollinear directions, chosen to cover three-dimensional space uniformly.15 The duration and maximum amplitude of the diffusion gradients were 25 ms and 21 mTm⁻¹ giving a maximum b factor in each direction of 1,044 s/mm². In order to optimize the measurement of diffusion, only two b factors were used (b₁ ≈ 0, b₂ = 1,044 s/mm²). Fat saturation was performed using a four radiofrequency binomial pulse train to avoid a chemical shift artifact. A birdcage head coil of approximately 300-mm diameter was used for signal transmission and reception. Finally, T1-weighted conventional spin echo (TR, 768 ms; TE, 15 ms) was performed 5 minutes after injection of 0.1 mmol/kg gadolinium–diethylenetriaminepentaacetic acid.

For turbo spin echo, gradient echo, and conventional spin echo sequences, 24 contiguous axial slices were acquired (5 mm slice thickness, 256 × 256 matrix, and 250 × 250 mm field of view). The slices were positioned to run parallel to a line that joins the most inferoanterior and inferoposterior parts of the corpus callosum. For PGSE-EPI scans, 10 axial slices (5 mm slice thickness, 128 × 128 matrix, and 250 × 250 mm field of view) were acquired, with the same orientation of the dual echo scans, with the second-to-last caudal slice positioned to match exactly the central slices of the dual-echo, gradient echo, and T1-weighted sets. This brain portion was chosen because the periventricular area is a common location for MS lesions and these central slices are less affected by the distortions due to B₀ field inhomogeneity.

During the second session, MRSI scans of the brain were obtained using another MR system operating at 1.5 T. A turbo spin echo sequence with axial slices parallel to the line connecting the anterior and posterior commissures was acquired. These images were used to select an intracranial volume of interest measuring approximately 90 mm (anteroposterior) × 20 mm (craniocaudal) × 90 mm (left to right), centered on the corpus callosum. Two-dimensional spectroscopic images were obtained using a 90°–180°–180° pulse sequence (TR, 2,000 ms; TE, 272 ms; field of view, 250 × 250 mm; phase encoding steps, 32 × 32, with one signal average per step), as previously described.14

Image analysis and postprocessing.

All image analysis and postprocessing was conducted by an experienced observer, unaware of the patients’ clinical characteristics. T2 and T1 lesion volumes were measured using a segmentation technique based on local thresholding.16 The two gradient echo images (with and without the saturation pulse) were first coregistered,17 and then MT ratio (MTR) images were derived pixel by pixel, as previously described.4 PGSE-EP images were first corrected for distortion induced by eddy currents.17 Then, the diffusion tensor was calculated for each pixel according to the equation described by Basser et al.18 After “diagonalization” of the matrix, mean diffusivity (D) was derived for every pixel. The diffusion maps were then interpolated to the same matrix as the dual echo images. Histograms of MTR and D maps were created as previously described.4,7⇓ For all the histograms, the average MTR and D values were calculated as well as the heights and locations of the peaks of the histograms. Postprocessing of the MRS images was performed as previously described.19 Metabolite resonance intensities of N-acetylaspartate (NAA), choline, and creatine were determined automatically from peak areas relative to a spline-corrected baseline. Results were expressed as the intravoxel ratios of NAA and choline to total creatine.

Statistical analysis.

Univariate correlations between MR measures and disability were assessed using the Spearman rank correlation coefficient. Multivariable linear regression models were used to generate composite MR scores. Each of these composite scores was computed using a linear combination of MR parameters, chosen a priori based on biologic considerations. The weight of each MR parameter resulted from the coefficients estimated by the regression model. The magnitude and the significance of the different correlations between EDSS and composite MR scores were evaluated by a nonparametric Spearman correlation analysis, because EDSS does not satisfy the assumptions of continuity and normality for a valid inference in linear regression models, and the sample size studied was too small for asymptotic properties to be applied.