Abstract
Objective: To assess the potential of registered volumetric MRI in measuring rates of atrophy in MS.
Background: Pathologic and imaging studies suggest that the development of permanent neurologic impairment in MS is associated with progressive brain and spinal cord atrophy. Atrophy has been suggested as a potential marker of disease progression. Conventional atrophy measurements requiring manual outlining are time-consuming and subject to reproducibility problems. Registration of serial MRI may offer a useful alternative in that cerebral losses may be measured directly from automated subtraction of brain volumes.
Methods: Twenty-six patients with MS and 26 age- and gender-matched controls had two volumetric brain MR studies 1 year apart. Baseline brain and ventricular volumes were measured using semiautomated techniques, and follow-up scans were registered to baseline. Rates of cerebral atrophy were calculated directly from the registered scans.
Results: Baseline brain volumes in the MS group were smaller (mean difference 78 mL [95% CI 13 to 143; p = 0.02]) and ventricular volumes greater (mean difference 12 mL [95% CI 6 to 18; p < 0.001]) than controls. The rate of cerebral atrophy in the MS group (0.8% per year) was over twice that of controls (0.3%), and the rate of ventricular enlargement was five times greater than the controls (1.6 versus 0.3 mL/year).
Conclusion: Progressive cerebral atrophy is an important feature of MS. Registration-based measurements are sensitive and reproducible, allowing progressive atrophy to be detected within 1 year and may have potential as a marker of progression in monitoring therapeutic trials.
Therapeutic agents for the treatment of MS are available. Most therapeutic trials focus on measuring clinical change in disability and, as a surrogate marker for disease progression, total lesion volume on MRI. A median increase in total lesion volume of 5% to 10% per year has been shown in relapsing-remitting patients on unenhanced T2-weighted scans.1 A more pathologically specific method may be provided by measurement of hypointense areas on T1-weighted scans.2 However, methods used for measuring lesion volumes (manual outlining in particular) can be time-consuming and may be subject to problems of reproducibility.3 Furthermore, the correlation between T2 lesion volume and disability has been disappointing. There is, therefore, a need for a surrogate MRI marker that is sensitive and reproducible but also tracks disease progression, as opposed to disease activity, over time and correlates with progressive disability. To facilitate use in large clinical trials, the measurements involved also should be at least semiautomated.
Recent pathologic studies remind us that axonal transection is a common consequence of demyelination.4 These findings raise the possibility that axonal destruction is the primary factor responsible for permanent neurologic dysfunction. Axonal loss cannot be assessed directly in vivo; however, a macroscopic concomitant of those losses, namely, cerebral or spinal cord atrophy, can be measured using MRI. Previous cross-sectional imaging studies report reduced cerebral, hemispheric, and brainstem volumes and enlarged ventricles in MS patients compared with normal controls.5-9 Furthermore, imaging studies show statistically significant correlations between disability and atrophy of the cerebellum and spinal cord.7,9-13 The measurement methods used include estimating total or regional brain volumes from either a representative number of axial MR slices,10 stereologic (point counting) techniques,7 manual segmentation of volumes,6 or semiautomated estimates of brain–intracranial volume ratios.11-13 Relatively few longitudinal MRI studies of cerebral atrophy in MS have been performed, and all have used the technique of measuring as consistently as possible a cerebral region on the first scan, measuring the region on the second scan, and then subtracting the two numbers to get a volume or area of difference.
The heterogeneity of disease in MS means that measurement of any one cerebral subregion may miss significant atrophy occurring elsewhere in the brain. Detection of small diffuse volume changes from serial MRI is difficult with conventional methods that rely on manual outlining of regions of interest or on semiautomated thresholding because the results are critically dependent on the reproducibility of the outlining of the structure being measured. Such measurements are time-consuming, observer dependent, and susceptible to segmentation error. Image subtraction offers an attractive alternative method of assessing small amounts of diffuse atrophy directly from serial scans. For subtraction to produce meaningful results, the baseline and follow-up scans need to be positionally matched (registered). However, even with careful positioning of individuals at the time of scanning, differences in the position and orientation of the acquired brain images usually are much greater than any changes being measured. Recent developments in registration methods have improved the accuracy of postacquisition positional matching of serial three-dimensional MR brain images. Whole-brain MRI volumes can now be registered to within a fraction of a voxel of each other, thereby allowing subtraction images to reveal volume changes.14,15
Once the serial scans from an individual are accurately registered, cerebral volume change may be derived directly.16,17 Unlike manual outlining, this method has much less critical dependence on accurate identification of brain tissue, since brain volume changes are measured by integrating the shifts in brain–CSF boundaries that result from cerebral atrophy. The integral, the brain boundary shift integral (BBSI), represents the total volume traversed by the boundaries of the brain in going from the first scan to the second scan. Small segmentation errors do not affect the BBSI. Hence, this provides a direct measure of change in total cerebral volume, which may be useful in MS.
This technique was applied to the measurement of rates of cerebral atrophy over a 1-year period in patients with MS.
Subjects and methods.
Subjects.
Twenty-six patients with clinically definite MS and 26 age- and gender-matched healthy controls were studied. Each individual underwent two volumetric MRI studies approximately 1 year apart. The MS group consisted of nine patients with primary progressive MS, six with secondary progressive MS, six with relapsing-remitting MS, and five with benign disease.18 All patients underwent a neurologic examination and evaluation of the Kurtzke Expanded Disability Status Scale (EDSS) before undergoing MRI.19 A significant change in EDSS over the year was defined as an increase of 1.0 if baseline EDSS was less than or equal to 5.0 and an increase of 0.5 if baseline EDSS was greater than 5.0. None of the patients was experiencing an acute relapse at the time of assessment. A summary of demographic features of the patients and controls is shown in table 1. All subjects gave written informed consent to take part in the study, which was approved by the local research ethics committee.
Baseline details of MS and control groups
MRI.
T1-weighted, volumetric MRI images were acquired on a 1.5-T Signa unit (General Electric Medical Systems, Milwaukee, WI) with a fast spoiled GRASS (FSPGR) technique and 28 × 18 cm field of view, yielding 124 contiguous 1.5-mm thick coronal slices through the head with a 256 × 192 image matrix. (Acquisition parameters were repetition time [TR]/echo time [TE]/inversion time [TI]/number of excitations [NEX]/flip angle[FLIP]—15.6/4.2/450/1/20 apart from 13 of the controls, where the parameters were TR/TE/NEX/FLIP—35/5/1/35.)
Image analysis.
Image data were transferred to a Sun workstation (Sun Microsystems, Inc., Mountain View, CA), and all analyses were performed blind to patient details and diagnosis. Control and patient scans were interspersed and randomly presented to the operator. Rates of brain atrophy were calculated by accurately registering each subject’s pair of scans. Regions defining whole brain were first obtained using semiautomated iterative morphologic techniques20 with the image intensity threshold for the boundary between brain and CSF set at 60% of mean brain intensity. The inferior cutoff through the brainstem was taken at the level of the lowest point of the cerebellum. Ventricular volumes comprised the lateral ventricles and temporal horn of the lateral ventricles.
The registration algorithm determines the rotations and translations required to obtain an optimum match over the whole brain. The optimization procedure minimizes the SD of the ratio of signal intensity from each voxel. In this way, all brain voxels contribute to accurate registration.15 The figure illustrates the accuracy of registration achieved. Changes in cerebral volume were calculated directly from the registered image by measuring the BBSI.17,21 Rates of atrophy were expressed as percentage change in brain volume per year.
Figure. (A–C) Registered T1-weighted MRI of a normal control subject. The brain has been accurately registered and a representative coronal slice is shown. However, all slices are equally well registered. (A) Baseline scan; (B) second scan acquired 1 year later; (C) superimposed difference image where black indicates loss of signal and white gain in signal showing no significant change within brain between the scans. (D–F) Registered T1-weighted MRI of a MS patient. (D) Baseline scan; (E) second scan acquired 1 year later; (F) superimposed difference image. Ventricular enlargement is shown in black; change in signal from new lesions (brainstem) also are shown, but these do not contribute to the measurement of atrophy by the brain boundary shift integral.
Reproducibility.
inter-rater reproducibility was assessed by measuring brain and ventricular volumes of 10 subjects presented to two investigators blind to patient details and diagnosis. Similarly, intra-rater reproducibility was assessed with five randomly chosen subjects measured twice. Reproducibility in each case was expressed as the coefficient of variation (SD divided by the mean).
To assess reproducibility of the BBSI measures used in this study, eight randomly chosen subjects were reanalyzed using segmentation, registration, and quantification by a second observer blind to the results of the first. Reproducibility was expressed as the root mean square difference in BBSI.
Statistical analysis.
Data were analyzed using Microsoft Excel 95 (Microsoft Corporation, Redmond, WA), SPSS version 8.0 (SPSS, Inc., Chicago, IL), and Stata version 5.0 (Stata Corp, College Station, TX). Two sample t-tests, or the Mann-Whitney test (where the assumptions for the t-test were not fulfilled), were used to compare measures of brain and ventricular volume and median rates of brain atrophy and ventricular growth between patients and controls. Where the t-test suggested a statistically significant difference between patients and controls, the effects of disease type was investigated using analysis of variance (ANOVA). The appropriateness of the ANOVA was tested using Bartlett’s test. Where Bartlett’s test suggested unequal variances, the Kruskal-Wallis test was used. Where the ANOVA indicated a significant between group difference on the F test (p < 0.05), linear regression models were fitted to investigate differences between subgroups. The appropriateness of the models were checked by examining residuals, fitted values, and Cook’s distances. Spearman’s rank correlation coefficient was used to examine the relationships between disability measured by EDSS, disease duration, and cerebral volume.
Results.
Reproducibility.
The mean coefficient of variation for inter-rater reproducibility of brain volumes was 0.54% (range 0.01 to 1.1). The mean (±SD) of the 10 brain volumes for the two measurers were 1416.6 cm3 (±102) and 1413.6 cm3 (±95), respectively. Mean intra-rater reproducibility was 0.46% (range 0.1 to 0.7) with mean (±SD) brain volumes for the five scans measured of 1351.1 cm3 (±65) and 1350.0 cm3 (±66). The mean coefficient of variation for inter-rater reproducibility of ventricular volumes was 0.32% (range 0.01 to 1.2) with means (±SD) of 19.5 cm3 (±10.6) and 19.6 cm3 (±10.7). Mean (±SD) ventricular volumes for intra-rater reproducibility were identical at 30.0 cm3 (±12.8) for each set of measurements. The mean coefficient of variation was 0.02% (range 0.002 to 0.05). Registration was successful in all cases. The root mean square difference (similar to mean absolute difference) in atrophy measured using the BBSI was 1.9 cm3 (only 0.16% of mean brain volume) compared with a root mean square difference in atrophy measured by segmentation of 10.5 cm3 in the same subjects (0.89% of brain volume).
Demographic data.
There was no significant difference (t = 0.98, p = 0.3) between the ages of the controls and MS patients (see table 1).
The follow-up interval was not significantly different between the two groups, with interval means (±SD) of 363 days (±62) for the controls and 344 days (±51) for the MS group.
Cross-sectional measures.
The mean baseline brain volume of 1165 ± 118 mL of the MS group was 78 mL (95% CI 13 to 143 mL; p = 0.02) smaller than the mean of 1243 ± 117 mL for the control patients. The mean ventricular volume at baseline of 27.5 ± 11.9 mL for the MS group was 12 mL (95% CI 6 to 18; p < 0.001) greater than the mean of 15.4 ± 9.7 mL for the control group (table 2).
Mean and SD of baseline brain and ventricular volumes
There was a negative correlation between disease duration and baseline brain volume in the MS group (r = −0.7, p = 0.0003).
The ANOVA models showed no significant differences between the different MS subgroups on baseline brain or ventricular volume after accounting for differences in gender distribution.
Longitudinal measures.
Rates of brain atrophy and ventricular growth are shown in table 3. The median rate of brain atrophy for the MS group at 0.8% per year was significantly greater than the median rate of 0.3% per year in the control group. Median ventricular growth for the MS group was 1.6 mL/year, which also was significantly greater than the median rate of 0.3 mL/year in the control group. There was no significant difference between subgroups of MS patients on any of these measures. Only 9 of the 26 MS patients showed a significant increase in EDSS over the study period, and no significant correlation was found between change in EDSS score and change in brain or ventricular volume.
Median rates (interquartile range) of brain atrophy and ventricular growth
Discussion.
This study has used registration of serial volumetric MRI to show significantly increased rates of cerebral atrophy in MS over 1 year. The MS group had rates of global brain atrophy and ventricular enlargement that were two to five times greater than an age- and gender-matched control group. The MS group had smaller mean baseline brain volumes (6% smaller) and larger mean ventricular volumes (80% greater) compared with the control group.
Registration of serial MRI provides a reproducible and sensitive measurement of diffuse cerebral volume loss. The method does not require sophisticated image acquisition: a standard volumetric T1-weighted sequence is used, acquired under 10 minutes. All analyses are performed after acquisition and so do not increase scanning time. The main advantage of the technique is that accurate positional matching allows digital subtraction of serial imaging that is not critically dependent on the accuracy of segmentation. Each individual forms their own control and change in brain volume is calculated directly from the registered MRI. The method gives a measure of net volume loss over the whole brain, so that small positional shifts or shape changes do not affect the total calculated atrophy. The technique has good reproducibility, with a root mean square difference of atrophy (error) of only 0.16% of mean brain volume compared with 0.9% with segmentation. Measurement variability of 0.16% is significantly less than the changes in brain volume we expect in MS. This level of reproducibility compares well with previously described segmentation or stereologic measurement techniques: intra-observer variability typically is quoted at between 2% and 6%.7,22
This study did not directly compare atrophy measurements with other potential MRI markers of progression such as T2 or T1 lesion volumes. It is, therefore, not possible to assess the relative sensitivity to change of these different measures. To facilitate comparisons, these different measurements need to be performed in the same patient cohort.
Reduced global and regional cerebral volumes and increased ventricular size in MS patients have been reported in previous cross-sectional MRI studies.5,8,9,11-13 The size of any differences are dependent on the composition and severity of the patient group studied. Our results generally are in agreement with others’ studies.6 Similar group differences have been reported in recent studies using different atrophy measures. Brain-to-intracranial volume ratios were found to be 6% smaller in 140 relapsing-remitting patients with MS compared with 16 healthy control subjects.12,13 A cross-sectional MRI study7 of 10 control subjects and 40 patients (20 relapsing-remitting and 20 secondary progressive) using stereologic measurements of atrophy found normalized cerebral hemisphere volumes to be 56 mL smaller in the MS group, which is comparable to our study. However, these differences were not statistically significant (p = 0.047), and the authors of the study speculate that the lack of significance resulted from the large variance in their data.7 Nonetheless, they did find significant differences in white matter volumes.
Corrections for intracranial volume are important to adjust for differences in head size that may occur, particularly when subject and control groups are not well matched for gender. Both brain and intracranial volumes are 15% larger in men than in women, and unless the study appropriately controls for these differences, they may obscure any pathologic change. Rates of atrophy expressed as a percentage of baseline brain volume are less likely to be subject to gender effects, but age matching may be more important. Our patient and control groups were, therefore, closely matched for both gender and age.
The significant differences in baseline brain and ventricular volumes between the MS and control groups and the significant negative correlation between brain volume and disease duration, imply an ongoing increased rate of atrophy in the MS group. This was confirmed by the longitudinal measurements, which showed that over the year of the study, ventricular enlargement occurred at five times the rate of the control group and global atrophy progressed at twice the control rate. Ventricular and brain atrophy measurements were performed independently of one another using different techniques, and it is reassuring that the results support each other with a good correspondence between rates of brain atrophy and ventricular enlargement. These increased rates are (roughly) consistent with the difference in baseline brain and ventricular volumes given the disease duration in the MS patients. The baseline differences in ventricular volume (12 mL) and brain volume (78 mL) were 6 to 10 times the respective difference in annualized rates of atrophy. If a linear rate of atrophy was assumed (unlikely), this implies 6 to 10 years of pathologic rates of atrophy. Nonlinear progression is more likely, and further longitudinal studies with careful measurement of rates of change are needed to define the natural history of atrophic change in MS.
The rate of cerebral atrophy of 0.8% per year found in our mixed group of MS patients was similar to that of approximately 0.9% per year found in studies on relapsing-remitting patients.12 These rates of loss also are similar to those found in an 18-month study: cerebral volumes estimated from four axial slices showed rates of loss ranging between 1% and 3% over this period.10
We did not show a significant association between baseline EDSS and either brain or ventricular volume, despite finding a significant correlation between brain volume and disease duration. Other studies also failed to find a correlation between whole-brain volume and EDSS.6,10 By contrast, a significant negative correlation has been shown between white matter volume and disability as measured either by EDSS or the Scripps Neurologic Rating Scale.7 The lack of a correlation in our study may result from the heterogeneity of our MS group, with small patient numbers in each subtype, combined with limitations of the EDSS. In particular, the EDSS does not take into account cognitive impairment, which might be an expected correlate of cerebral atrophy, although this is not always the case.23,24 A positive relationship has been demonstrated between neuropsychological impairment and ventricular enlargement on CT and on MRI.25,26 By contrast, spinal cord atrophy in MS has been shown to correlate strongly with EDSS27,28 probably reflecting the heavy weighting of locomotor disability within the EDSS score.
The poor correlation between change in EDSS and rate of atrophy may reflect the minimal EDSS progression in our patients over the year (it failed to change in two-thirds of the patients). A longer follow-up period and a greater spread in change in EDSS scores might have allowed associations with structural change to be detected. An earlier 18-month study showed a higher rate of atrophy in patients who had a sustained deterioration in terms of EDSS compared with those who did not.10 However, cerebral atrophy is not specific to MS, and results may be confounded when other diseases are present, although this would be more of a problem with older subjects. Further studies that correlate atrophy with progression in disability are required to give face validity to this potential surrogate marker. In the case of cerebral atrophy, whether measured by registration or by any other technique, it is logical that future studies include measures of cognitive function.
These results show that registration of volumetric MRI can detect atrophy in MS over an interval of only 1 year. This adds to the accumulating evidence that atrophy, whether of spinal cord, cerebellum, or cerebrum, is an important consequence of disease progression in MS. Registration-based measures of cerebral atrophy from serial MRI are sensitive and reproducible and, perhaps in combination with measures of spinal cord atrophy, may prove to be effective surrogate markers of progression of value in therapeutic trials in MS. Their serial application over longer periods and in different subgroups, in combination with other MR techniques for evaluating MS disease, also should give insights into the pathogenic mechanisms underlying disability in MS.
Acknowledgments
Supported by a MRC Clinician Scientist Fellowship (N.C. Fox); an Alzheimer’s Disease Society Research Fellowship (R.J. Harvey); a grant from the MS Society of Great Britain and Northern Ireland (NMR Research Unit); and Biogen (S.M. Leary).
- Received July 12, 1999.
- Accepted November 8, 1999.
References
Letters: Rapid online correspondence
REQUIREMENTS
You must ensure that your Disclosures have been updated within the previous six months. Please go to our Submission Site to add or update your Disclosure information.
Your co-authors must send a completed Publishing Agreement Form to Neurology Staff (not necessary for the lead/corresponding author as the form below will suffice) before you upload your comment.
If you are responding to a comment that was written about an article you originally authored:
You (and co-authors) do not need to fill out forms or check disclosures as author forms are still valid
and apply to letter.
Submission specifications:
- Submissions must be < 200 words with < 5 references. Reference 1 must be the article on which you are commenting.
- Submissions should not have more than 5 authors. (Exception: original author replies can include all original authors of the article)
- Submit only on articles published within 6 months of issue date.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- Submitted comments are subject to editing and editor review prior to posting.
You May Also be Interested in
Cutaneous α-Synuclein Signatures in Patients With Multiple System Atrophy and Parkinson Disease
Dr. Rizwan S. Akhtar and Dr. Sarah Brooker