| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Neuroradiology |
1 From the Departments of Radiology (J.C.J.B., G.J.L.N., J.A.C., E.B., F.B.), Pathology (W.K.), and Neurology (C.P.), MR Center for MS Research, and Department of Biostatistics and Epidemiology (H.J.A.), VU Medical Center, PO Box 7057, 1007 MB Amsterdam, the Netherlands; Experimental in vivo NMR, Image Sciences Institute, University Medical Center Utrecht, Utrecht, the Netherlands (E.L.A.B.); Department of Biomedical NMR, Faculty of Biomedical Engineering, University of Technology, Eindhoven, the Netherlands (K.N.I.); and Netherlands Brain Bank, Amsterdam, the Netherlands (R.R.). Received September 27, 2003; revision requested December 5; revision received February 6, 2004; accepted March 18. Supported by the Stichting Vrienden MS Research; J.C.J.B. supported by grant no. 97–307 MS from the Stichting Vrienden MS Research. Address correspondence to J.C.J.B. (e-mail: j.bot@vumc.nl).
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
MATERIALS AND METHODS: Formaldehyde-fixed cervical spinal cord specimens from 11 patients with MS—three men and eight women (mean age at death, 66 years ± 11.3 [standard deviation])—and two female control subjects without neurologic disease (83 and 41 years of age at death) were examined at 4.7 T. Relaxation time measurements and MTR mapping were performed. Analyses included the whole cord area and region-of-interest measurements. Histopathologic analyses included semiquantitative myelin and quantitative axonal analysis.
RESULTS: Compared with control specimens (P < .001, analysis of variance), specimens from patients with MS had smaller cord areas (mean area, 59.0 mm2 ± 12.5 vs 72.7 mm2 ± 10.0), significant prolongation of T1 (mean prolongation, 30%) and T2 (mean prolongation, 13%), and decreased MTRs (mean, 10.5%). Within MS specimens, 58% of the white matter area displayed signal intensity abnormalities on intermediate-weighted MR images. The number of axons in normal-appearing white matter in MS specimens was, on average, 46% lower than the number of axons in white matter in control specimens. All quantitative MR parameters correlated well with demyelination; the correlation with T2 relaxation time was the strongest (r = 0.77, Spearman and Kendall nonparametric correlations). By contrast, quantitative MR parameters correlated less well with axonal density; the correlation with T2 relaxation time was the strongest (r = –0.44, Spearman and Kendall nonparametric correlations). Multilevel analysis, corrected for age and MS phenotype, could not result in a model explaining axonal density on the basis of quantitative MR parameters when myelin density was included as a predictor.
CONCLUSION: Changes in quantitative MR imaging parameters in the cervical spinal cord in MS are mainly determined by demyelination and do not reflect axonal disease well.
© RSNA, 2004
Index terms: Magnetic resonance (MR), high-field-strength imaging, 341.12141, 341.121417, 341.12146 • Magnetic resonance (MR), magnetization transfer, 341.121417 • Magnetic resonance (MR), tissue characterization, 341.12146 • Spinal cord, MR, 341.12141, 341.121417, 341.12146 • Sclerosis, multiple, 10.871
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Recent studies have involved investigating spinal cord abnormalities in greater detail by using high-spatial-resolution MR imaging combined with postmortem histopathologic analysis. Nijeholt et al (11) described a strong correlation between the extent of spinal cord MS abnormalities, as depicted on high-spatial-resolution intermediate-weighted MR images, and demyelination. Another quantitative study (9) revealed that axonal damage in the spinal cord occurs largely independently of lesions on T2-weighted MR images. This lack of sensitivity to axonal disease of MR imaging is unfortunate because axonal disease may represent the substrate of persistent disability in patients with MS (12,13).
More specific methods of MR imaging, including MR spectroscopy, magnetization transfer imaging, and measurement of relaxation time, have been developed. Although MR spectroscopy is sensitive for detecting neuronal damage, it has limited spatial resolution and cannot easily be applied to spinal cord imaging. It has been shown that the results of relaxation time mapping and magnetization transfer imaging correlate better with the degree of demyelination and/or axonal loss in the brain and that these measurement techniques may also be applied to spinal cord imaging (10,14–17).
The purpose of our postmortem study was to correlate quantitative MR imaging data (ie, relaxation times and magnetization transfer ratios [MTRs]) with the histopathologic findings of demyelination and axonal disease in cervical spinal cord specimens from patients with MS and control subjects.
|
MATERIALS AND METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
MR Imaging Protocol
Specimens were fixed in 10% formalin shortly after autopsy because they could not be imaged immediately. In four cases, repeated MR examinations were performed at various random points in time (between 20 and 160 days) after death. No further decreases in T1 or T2 relaxation times or decreases in MTR were observed after 45 days of fixation (J.C.J.B., E.L.A.B., K.N., and F.B. working in consensus; data not shown). The images used for the present study were, therefore, all obtained between 61 and 117 days after initial fixation to minimize possible differences between specimens caused by the effects of formalin fixation.
Specimens were imaged in a transverse plane at 4.7 T (Unity INOVA; Varian, Palo Alto, Calif) by using a custom-made solenoidal coil. For all high-spatial-resolution experiments, 25 1-mm sections were obtained with a field of view of 2 x 2 cm (with a 128 x 128 matrix that was zero-filled to 256 x 256, resulting in 7-µm pixels). Conventional MR imaging included a high-spatial-resolution intermediate-weighted spin-echo sequence (repetition time msec/echo time msec, 3000/15; number of signals acquired, eight) for anatomic images. Quantitative MR imaging included an inversion-recovery-array sequence (>5000/15; number of signals acquired, two) involving inversion times of 10, 150, 300, 600, 1200, and 1800 msec for T1 relaxation time measurements; a spin-echo sequence (repetition time, 4000; number of signals acquired, four) with a variable echo time of 10, 16, 25, and 40 msec for T2 relaxation time measurements; and a gradient-echo sequence (1300/9; number of signals acquired, four) with a flip angle of 5° that was performed with and without a Gaussian-shaped off-resonance prepulse (–4.7 kHz off resonance; duration, 12 msec; nominal flip angle, 429°) for magnetization transfer measurements.
T1 maps were calculated by using a hill-climbing algorithm in which signals from the inversion-recovery array were fitted on a voxel-by-voxel basis by using three parameters according to the following equation: SI = A + B exp(–TI/T1), where SI is absolute signal intensity, A is proton density, factor B accounts for imperfections of the inversion pulses (18), TI is inversion time in milliseconds, and T1 is in milliseconds. T2 maps were calculated by using a hill-climbing algorithm in which the signal from the echo time array was fitted on a voxel-by-voxel basis by using two parameters according to the following equation: SI = C exp(–TE/T2), where C represents proton density multiplied by a constant factor, TE is echo time in milliseconds, and T2 is in milliseconds. MTR maps were calculated according to the following equation: [(Mo – Ms)/Mo] · 100, where Mo represents signal intensity on images obtained without off-resonance pulses and Ms represents signal intensity on images obtained with off-resonance pulses.
Image Analysis
On the intermediate-weighted MR images, location (posterior, anterior, or lateral column) and type of signal intensity abnormalities were determined in consensus by J.C.J.B., E.B., and G.L.N., each of whom had a minimum of 3 years of experience with spinal cord MR imaging. Lesions were classified as high-signal-intensity lesions when they were clearly demarcated and their signal intensity was higher than that of gray matter and as intermediate-signal-intensity lesions when they showed vaguely defined areas of mild signal intensity increase (ie, signal intensity equal to or less than that of central gray matter) (11). Normal-appearing white matter (NAWM) was defined as the remaining white matter within the MS specimen that showed no visually appreciable signal intensity increase.
Lesion size was expressed as a percentage of the total cord surface area (as determined with manual tracing of the cord contour) that showed abnormalities; different white matter columns were quantified separately. T1 and T2 relaxation times and MTR were measured for both the entire cervical spinal cord cross-sectional area (for comparison of patients with MS and control subjects) and within regions of interest (ROIs) (for comparison of differences between tissue types [ie, white matter, NAWM, intermediate-signal-intensity lesions, and high-signal-intensity lesions]). The standard MR imaging ROI size was 0.35 mm2. ROIs were placed by J.C.J.B.
Histopathologic Evaluation
After MR imaging, specimens were cut into six equal pieces of approximately 1 cm each. Thereafter, 5-µm-thick slices were cut at 1-cm intervals (six per subject) and stained with Klüver (BDH Laboratory Supplies, England) for myelin and NE-14 (phosphorylated neurofilaments) (Sigma-Aldrich, the Netherlands) for axons. MR images and histologic slices were matched by using anatomic landmarks such as the shape of the central gray matter and the shape of the focal abnormalities. Care was taken to select ROIs on histologic slices that were identical to those placed on MR images (J.C.J.B. and W.K. in consensus).
The number of axons was counted on NE-14–stained tissue slices within the regions defined on MR images by using a digitizing video overlay system (Qprodit; Leica Microsystems, Cambridge, England). A final on-screen magnification of x1200 (x400 objective) was used, and the typical field size of axonal measurements was 0.002621 mm2. The number of fields assessed was based on the running mean, but at least six fields were counted within each of the MR imaging–specified ROIs (9). Axonal measurements were performed by two pathology department students who were trained in the Qprodit system, with supervision from J.C.J.B., W.K., and E.B.
The axonal diameter was also determined with the Qprodit software by using the smallest diameter through the center of the axons. Qprodit was also used to calculate axonal distance (ie, the shortest distance between centers of axons). The amount of residual myelin was estimated, as a percentage relative to the normal density of myelin figures, on Klüver-stained slices by an experienced neuropathologist (W.K., with 25 years of experience) (19). A myelin density score of 90% or greater was set as the cutoff value for normal myelin density. (Note that owing to technical limitations, it is not possible to standardize the staining intensity for myelin, so a quantitative approach is thus precluded.)
A total of 222 of 250 ROIs were analyzed (J.C.J.B., H.J.A.) for comparison between MR imaging and histologic findings. These 222 ROIs consisted of 158 ROIs in MS specimens (50 ROIs in posterior, 53 in anterior, and 55 in lateral column) and 64 ROIs in the white matter of control specimens (20 in posterior, 20 in anterior, and 24 in lateral portions of the spinal cord). Twenty-eight of the initially selected ROIs could not be included in the final analysis owing to inaccurate T1- or T2-relaxation-time–fitting routines or suboptimal quality of the corresponding histologic staining (J.C.J.B., W.K., F.B.). ROIs in MS specimens included 82 NAWM areas, 40 areas of intermediate signal intensity increase, and 36 areas with high-signal-intensity abnormalities.
Statistical Analysis
For data that are not normally distributed, medians rather than means are reported, together with interquartile ranges instead of standard deviations. Differences between MS and control specimens regarding normally distributed quantitative MR imaging and histopathologic parameters were calculated by using a one-way analysis of variance test. For ROI subgroup analysis, we used Tukey post-hoc comparisons when appropriate. The Wilcoxon W test was used to evaluate differences between non–normally distributed data. Correlations between MR imaging and histopathologic parameters were calculated by using Spearman and Kendall nonparametric correlations. A P value of less than .05 was considered to indicate a statistically significant difference. Multilevel analysis was used to model myelin or axonal density by using the MR imaging parameters as predictors. This allowed correction for different numbers of axons within patients.
|
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
|
|
|
90%), and 24 (29%) showed low-but-near-normal myelin density (score, 80%–90%). Median myelin density in these NAWM ROIs was 70% (interquartile range, 10) (Table 3). Progressive demyelination was observed in MS specimens with increasing signal intensity on intermediate-weighted MR images (Fig 3).
|
|
With a decreased number of axons, the axonal diameter and interaxonal distance were increased in patients with MS as compared with these values in control subjects, regardless of MR imaging findings. Within MS specimens, the number of axons per square millimeter and the axonal diameter differed between NAWM and both intermediate-signal-intensity and high-signal-intensity abnormalities, although these values did not differ between the two types of focal lesions (Table 3, Fig 4).
|
|
|
|
Depending on ROI classification, T2 relaxation time in MS specimens increased progressively compared with that in control specimens. T2 relaxation time values differed significantly among all ROI classes, including between white matter and NAWM ROIs (Table 5).
MS specimens showed, on average, lower MTR (total cord area measurement), with a reduction of 10.5% compared with the MTR in control specimens (F1 = 43.9, P < .001) (Table 4). With decreasing cord area, a steady decrease in whole-cord-averaged MTR values was measured (Fig 5).
Progressive lowering of MTR was observed in MS abnormalities, with the MTR within intermediate-signal-intensity and high-signal-intensity abnormality ROIs being significantly lower than the MTR within white matter and NAWM ROIs (Table 5).
Relationship between Quantitative MR Imaging Parameters and Histopathologic Appearance
The relationship between quantitative MR imaging parameters and demyelination was quite strong, as expected (Table 5). Within the MS group, we found that with increasing demyelination, T1 and T2 relaxation times increased (Spearman 
= 0.71, P < .001 for T1 relaxation time; = 0.77, P < .001 for T2 relaxation time) (Fig 6) and MTR values decreased ( = –0.76, P < .001). A multilevel analysis model predicting myelin density (corrected for number of axons) showed independent contribution from T2 relaxation times and MTR. The model showed an increase of 0.22 msec (95% confidence interval: 0.13, 0.31) in T2 relaxation time and a decrease of 0.11% (95% confidence interval: 0.0086, 0.21) in MTR for each percentage point decrease in myelin density.
|
= –0.39 for T1 relaxation time, = –0.44 for T2 relaxation time, and = 0.41 for MTR) (Fig 7).
|
|
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
Similarly, myelin density also correlated very well with all quantitative MR parameters included. Of all quantitative MR parameters, T2 relaxation time was found to be the strongest independent predictor for myelin density.
The relationship between axonal and MR parameters was less straightforward. Although the lack of a strong correlation between axonal features and quantitative MR parameters may be expected owing to the confounding effect of demyelination, it is less obvious why this failed for NAWM as well. As previously reported (9,20), considerable loss of axons (in this study, 46%) was found in the NAWM of the spinal cord in MS specimens, while quantitative MR parameters showed only minimal changes (3%–7%), with only T2 relaxation time differing statistically from that of white matter in control specimens. We can only speculate that the lack of accompanying MR imaging changes was caused by shrinkage of the tissue, leading to atrophy and reestablishing pseudonormal biophysical properties in the remaining tissue. This phenomenon is apparently different from the situation in brain tissue, where axonal loss is accompanied not only by atrophy but also by widening of extracellular spaces, leading to abnormal relaxation mechanisms, expressed on T1-weighted brain MR images as so-called black holes (15,21).
In the control spinal cord specimens, no MR imaging abnormalities were detected, and myelin density for all ROIs was normal. In MS specimens, however, we found (as previously reported) signal intensity abnormalities ranging from intermediately increased signal intensity and poor demarcation to very clearly demarcated areas of high-signal-intensity abnormalities reflecting demyelination on intermediate-weighted MR images (11,22). The correlation between the appearance of the spinal cord on intermediate-weighted MR images and axonal parameters was not obvious. In addition, within NAWM ROIs, conventional MR imaging lacked sensitivity in depicting pathologic changes, which may include both demyelination and axonal loss. Areas classified as NAWM in MS cervical spinal cord specimens at MR imaging already showed considerable demyelination (median, 70% of residual myelin present) and extensive axonal loss (average, 46% of axons lost compared with axonal density of control specimens). As was the situation within NAWM, changes in relaxation times and MTR in lesions were dominated by myelin changes and did not reflect axonal disease well.
Another important finding of our study was that cord atrophy (ie, decreases in cord area) correlated well with T1 relaxation time increase (and, to a lesser extent, with MTR decrease and T2 relaxation time increase). Given the fact that these MR imaging parameters primarily represent demyelination, it appears that cord atrophy itself is a feature representing not only axonal loss but also the loss of myelin. This suggests that cord area measurement, which has been proposed as a clinically relevant marker for disease progression (23–25), may not be regarded as a direct measure of the number of axons present in the spinal cord.
When interpreting the results obtained in our study, some limitations have to be considered. First, selection bias may have been introduced. Although a large number of ROIs (n = 222) were analyzed, these were obtained from relatively small numbers of patients with MS (n = 11) and control subjects (n = 2). Patients with MS had, on average, long disease duration and were mostly in the final stage of their disease; the cervical cord specimens examined were, as a result, severely affected by MS disease. Second, possible effects of the presence and extent of gliosis on quantitative MR measurements were not taken into account in this study. Without a reliable and standardized staining procedure and quantification method, owing to heterogeneity in the spinal cord with regard to the appearance of gliosis (there were various combinations of numbers and sizes of reactive astrocytes and amounts and densities of fibrillary gliosis), we were not confident in including gliosis as a quantifiable histopathologic parameter. Last, our results differ from those of in vivo studies because we used formalin-fixed tissue and a high magnetic field strength, which are both known to influence MTR and relaxation times (26). We assume, however, that contrast between different tissues at MR imaging (14) and the correlations between MR imaging and histopathologic findings remain unaffected.
In conclusion, quantitative MR techniques revealed differences in relaxation times and magnetization exchange properties between control subjects and patients with MS, as well as between types of MS abnormalities. The degree of MR imaging abnormality is dominated by demyelination rather than by axonal disease.
|
ACKNOWLEDGMENTS |
|---|
|
FOOTNOTES |
|---|
Authors stated no financial relationship to disclose.
Author contributions: Guarantors of integrity of entire study, F.B., C.P., J.A.C.; study concepts, F.B., C.P., J.A.C.; study design, J.C.J.B., F.B., C.P., J.A.C., K.N.I., R.R.; literature research, J.C.J.B., F.B., C.P., K.N.I.; clinical studies, C.P., W.K., R.R.; experimental studies, J.C.J.B., E.L.A.B., K.N.I., W.K., F.B.; data acquisition, J.C.J.B., E.L.A.B., E.B., W.K., R.R.; data analysis/interpretation, J.C.J.B., E.L.A.B., K.N., F.B., W.K., H.J.A., C.P., G.J.L.N.; statistical analysis, J.C.J.B., H.J.A.; manuscript preparation, J.C.J.B.; manuscript definition of intellectual content and editing, J.C.J.B., F.B., C.P., J.A.C., G.J.L.N.; manuscript revision/review and final version approval, all authors
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMATERIALS AND METHODS RESULTS DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  M. Rovaris, F. Barkhof, M. Calabrese, N. De Stefano, F. Fazekas, D. H. Miller, X. Montalban, C. Polman, M. A. Rocca, A. J. Thompson, et al. MRI features of benign multiple sclerosis: Toward a new definition of this disease phenotype Neurology, May 12, 2009; 72(19): 1693 - 1701. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 K. M. Zackowski, S. A. Smith, D. S. Reich, E. Gordon-Lipkin, B. A. Chodkowski, D. R. Sambandan, M. Shteyman, A. J. Bastian, P. C. van Zijl, and P. A. Calabresi Sensorimotor dysfunction in multiple sclerosis and column-specific magnetization transfer-imaging abnormalities in the spinal cord Brain, May 1, 2009; 132(5): 1200 - 1209. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 C. Gilmore, J. Geurts, N Evangelou, J. Bot, R. van Schijndel, P. Pouwels, F Barkhof, and L Bo Spinal cord grey matter lesions in multiple sclerosis detected by post-mortem high field MR imaging Multiple Sclerosis, February 1, 2009; 15(2): 180 - 188. [Abstract] [PDF]
|
|
|
|
 |
|
 F. Agosta, E. Pagani, D. Caputo, and M. Filippi Associations Between Cervical Cord Gray Matter Damage and Disability in Patients With Multiple Sclerosis Arch Neurol, September 1, 2007; 64(9): 1302 - 1305. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
O. Ciccarelli, C.A. Wheeler-Kingshott, M.A. McLean, M. Cercignani, K. Wimpey, D.H. Miller, and A.J. Thompson Spinal cord spectroscopy and diffusion-based tractography to assess acute disability in multiple sclerosis Brain, August 1, 2007; 130(8): 2220 - 2231. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Agosta, M. Absinta, M. P. Sormani, A. Ghezzi, A. Bertolotto, E. Montanari, G. Comi, and M. Filippi In vivo assessment of cervical cord damage in MS patients: a longitudinal diffusion tensor MRI study Brain, August 1, 2007; 130(8): 2211 - 2219. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 H. Vrenken, J.J.G. Geurts, D.L. Knol, C.H. Polman, J.A. Castelijns, P.J.W. Pouwels, and F. Barkhof Normal-appearing white matter changes vary with distance to lesions in multiple sclerosis. AJNR Am. J. Neuroradiol., October 1, 2006; 27(9): 2005 - 2011. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y. Ge Multiple Sclerosis: The Role of MR Imaging AJNR Am. J. Neuroradiol., June 1, 2006; 27(6): 1165 - 1176. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S.M. Hesseltine, M. Law, J. Babb, M. Rad, S. Lopez, Y. Ge, G. Johnson, and R.I. Grossman Diffusion Tensor Imaging in Multiple Sclerosis: Assessment of Regional Differences in the Axial Plane within Normal-Appearing Cervical Spinal Cord AJNR Am. J. Neuroradiol., June 1, 2006; 27(6): 1189 - 1193. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 W Rashid, G R Davies, D T Chard, C M Griffin, D R Altmann, R Gordon, A J Thompson, and D H Miller Increasing cord atrophy in early relapsing-remitting multiple sclerosis: a 3 year study J. Neurol. Neurosurg. Psychiatry, January 1, 2006; 77(1): 51 - 55. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
F. Barkhof and R. C. McKinstry Quantifying spinal cord demyelination with magnetic transfer imaging Neurology, May 24, 2005; 64(10): 1677 - 1678. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
