Abstract
Objective: To evaluate 1) the ability of magnetization transfer ratio (MTR) histogram analysis to detect the extent of changes occurring outside MS lesions seen on conventional scans, 2) whether such changes vary in the different MS clinical phenotypes, 3) whether the changes are associated with the extent and severity of the macroscopic lesion load, and 4) the contribution to brain atrophy.
Methods: Dual-echo, T1-weighted, and MT scans of the brain were obtained from 77 patients with varying MS courses and 20 age- and sex-matched control subjects. To create MT histograms of the normal-appearing cerebral tissue, MS lesions were segmented from dual-echo scans, superimposed automatically, and nulled out from the coregistered and scalp-stripped MTR maps. Average MTR, peak height, and peak position were considered. T2 and T1 lesion loads, average lesion MTR, and brain volume were also measured.
Results: Average histogram MTR (p < 0.0001) and peak position (p < 0.0001) from patients with relapsing–remitting MS (RMMS) were lower than those from control subjects. Patients with primary progressive MS (PPMS) had lower average histogram MTR (p = 0.002) and histogram peak height (p = 0.01) than control subjects. Patients with secondary progressive MS (SPMS) had a lower peak height (p = 0.05) than those with RRMS. Average lesion MTR (p < 0.0001) correlated highly with the histogram MTR. Average histogram MTR (p < 0.0001) and T2 lesion load (p = 0.001) correlated highly with brain volume.
Conclusions: The amount of microscopic changes account for an important fraction of the lesion load in MS. They may contribute to the development of brain atrophy and tend to be more evident in patients with secondary progressive MS.
Subtle changes are known to occur in the so-called normal-appearing white matter (NAWM) from patients with MS, as shown by bioptic1 and postmortem studies.2-5 These changes include diffuse astrocytic hyperplasia, patchy edema, perivascular infiltration, abnormally thin myelin, and axonal loss.1-5 Although conventional MRI is very sensitive in detecting MS lesions and their changes over time,6-8 it does not give a complete picture of the overall burden of the disease because, by definition, it does not detect changes in NAWM.7 This, and the inability of conventional MRI to provide information regarding the different pathologic substrates of the MS lesions,7,9-11 may be the two main reasons for the lack or the paucity of the correlation found between the extent of the abnormalities seen on T2-weighted MR images and the MS clinical manifestations.6-8
The intrinsic nature of lesions visible on conventional MR images has been studied extensively using several new MR techniques,11-15 and encouraging correlations with the MS clinical manifestations have been found.14-17 On the contrary, the contribution of the changes occurring outside T2-visible lesions to the pathogenesis of MS has not yet been elucidated fully, although it is likely that they may contribute to the evolution of the disease.13,18,19
The pathologic abnormalities found in the NAWM from patients with MS have the potential to modify the relative proportions of mobile and immobile protons of the diseased tissue and, as a consequence, to determine a decrease of the magnetization transfer ratios (MTRs) from the white matter outside MS lesions visible on conventional MR images.7,20,21 Although previous MT studies only assessed changes in small portions of NAWM by using analysis of regions of interest (ROI),14,20,21 MTR histogram analysis (as introduced by van Buchem et al.22) allows evaluation of data from all the MR pixels of the brain tissue, thus providing a complete assessment of macro- and microscopic disease burden in MS. MTR histogram-derived measures from MS patients are different from those of healthy control subjects18,22-25 and are correlated with the clinical manifestations of MS.18,23,24 Because in these studies the whole brain tissue was used to create MTR histograms,18,22-25 it is difficult to disentangle the relative contributions of changes occurring within or outside MS lesions visible on conventional MR images. This study 1) evaluates the ability of MTR histogram analysis to detect the extent of the changes occurring outside MS lesions seen on conventional images, 2) estimates whether such changes are different in terms of extent and nature in the different MS clinical phenotypes, 3) defines whether they are associated with the extent and severity of the macroscopic disease burden (T2 and T1 lesion loads and average lesion MTR), and 4) assesses their contribution to brain atrophy.
Patients and methods.
Patients.
Patients included had clinically definite MS for at least 2 years.26 Their clinical disease phenotypes were classified as relapsing–remitting MS (RRMS; clearly defined disease relapses with either full recovery or sequelae, but without disease progression during the periods between the relapses), secondary progressive MS (SPMS; initial RR course followed by progression with or without occasional relapses, minor remissions, and plateaus), primary progressive MS (PPMS; disease progression from onset with occasional plateaus and temporary minor improvements), or benign MS (BMS) (patients fully functional in all neurologic systems 15 years after disease onset), according to the criteria of Lublin et al.27 None of the patients had had immunosuppressive or immunomodulating treatments for at least 12 months before entry into the study. In addition, they had neither relapses nor steroid treatment during the preceding 3 months. At the time MRI was performed, patients were assessed neurologically by a single physician who was unaware of the MRI results, and disability was measured using the Expanded Disability Status Scale (EDSS).28 Sex- and age-matched control subjects with no previous history of neurologic diseases and with a normal neurologic examination were also studied. Local ethical committee approval and written informed consent from all the patients and control subjects were obtained before study initiation.
Image acquisition.
Brain MR images were obtained using a scanner operating at 1.5 T (Magnetom SP63; Siemens, Erlangen, Germany). During a single session, the following images were acquired without moving the patient from the scanner: dual-echo conventional spin echo (CSE; repetition time [TR], 2,400 msec; first echo time [TE], 30 msec; second TE, 80 msec; number of acquisitions, 1), T1-weighted CSE (TR, 768 msec; TE, 15 msec; number of acquisitions, 2), two-dimensional gradient echo (GE; TR, 600 msec; TE, 12 msec; α = 20 deg) with and without an MT saturation pulse. The radio frequency saturation pulse was 1.5 kHz below the water frequency, with a gaussian envelope of duration of 16.4 msec, a bandwidth of 250 Hz, and an amplitude of 3.4 × 10-6 T. For the dual-echo and T1-weighted scans, 24 contiguous, interleaved, axial slices were acquired with a 5-mm slice thickness, a 256 × 256 matrix, and a 250-mm field of view, giving an in-plane spatial resolution of approximately 1 × 1 mm. MT images were obtained using the same acquisition parameters, except for the number of slices, which was 20. The set of slices for the MT images was positioned to obtain the same central 20 slices as for the dual-echo and T1-weighted images. The slices were positioned to run parallel to a line that joins the most inferoanterior and inferoposterior parts of the corpus callosum according to published guidelines,29 with “double obliquing” when necessary.
From the two GE images, with and without the saturation pulse, and after their coregistration (see next paragraph), MTR images were derived pixel by pixel according to the following equation: MTR = (M0 − MS)/M0 × 100, where M0 is the signal intensity for a given pixel without the saturation pulse and MS is the signal intensity for the same pixel when the saturation pulse is applied.
Image analysis.
Lesions were first identified by agreement by two experienced observers, without knowing to whom the images belonged, on the proton-density (PD) and T1-weighted hard copies. T2-weighted images were always used to increase confidence in lesion identification. For T1-weighted scans, only areas with a signal intensity between that of the gray matter and that of the CSF, and with corresponding lesions on both echoes of the dual-echo images, were considered hypointense lesions. Lesion volume measurements were then performed by a single observer, again without knowing to whom the images belonged. A semi-automated segmentation technique based on local thresholding was used for lesion segmentation,30 using the marked hard copies as a reference. The intraobserver coefficient of variation of this technique for measuring MRI abnormalities in MS is less than 5%.31 After image coregistration (described later), lesion outlines on PD-weighted images were superimposed onto the MTR maps and the average lesion MTR was calculated for each patient according to the following formula: 
where N is the number of lesions in a patient, Ai is the area of lesion i; and MTRi is the average MTR in lesion i.
Using five contiguous T1-weighted slices, with the most caudal at the level of the velum interpositum cerebri,32 we calculated a MRI measure of cerebral volume. On each slice, brain tissue was extracted from the skull and CSF spaces using the same strategy and technique31 used for lesion segmentation. Brain volume was calculated by multiplying the surface area of the segmented tissue by the slice thickness.
We also derived MTR histograms (with bins 1% in width) for the overall brain tissue not involved by lesions visible on the dual-echo scans. First, we removed the skull and other extracranial tissues from the PD-weighted and the GE images without and with the MT pulse, using the same local thresholding technique used for lesion segmentation.31 Second, the scalp-stripped GE images were coregistered and the MTR maps were obtained. Next, the MTR maps were coregistered with the corresponding scalp-stripped PD-weighted images. Coregistration of images was performed using a surface-matching technique that fits the contours corresponding to the CSF–dura interface of the two MR images. This technique estimates the spatial transformation needed to remap images from different MRI studies into the same spatial reference system, thus enabling a pixel-by-pixel correspondence between different MRI studies and ensuring the correct positioning of the ROI across multiple images.33,34 The optimal transformation is inferred from the minimization of the distance between the two discrete surfaces. Validation studies have shown that the accuracy of realignment was of the order of the image pixel size.34,35 Lastly, the macroscopic lesions segmented on PD-weighted images were superimposed automatically onto the coregistered MTR map, and the areas corresponding to the segmented lesions were nulled out. To reduce any additional image noise, we excluded from the analysis all the pixels with MTR values lower than 10%. This enables most of the pixels belonging to intraventricular CSF, other hypointense normal anatomic structures (e.g., blood vessels), and image background to be excluded. In this way, only pixels belonging to macroscopically normal brain tissue remained. To correct for the interpatient differences in brain volume, each histogram was normalized by dividing it by the total number of pixels included. For each MTR histogram, several parameters were analyzed. These included the height and the position of the histogram peak with respect to the x-axis (the MTR value most represented in the brain), the average MTR value, and the MTR values corresponding to the 25th, 50th, and 75th percentiles of the histogram (MTR25, MTR50, and MTR75), which indicate the MTR value at which the integral of the histogram is 25%, 50%, and 75% of the total respectively.
Statistical analysis.
Differences in MT metrics between MS patients and control subjects were evaluated using the two-tailed Student’s t-test for nonpaired data. The following four pairwise comparisons were decided a priori (a priori contrasts) to assess the differences in MTR histograms of the normal-appearing brain tissue (NABT) from the different clinical phenotypes studied: control subjects versus RRMS patients, control subjects versus PPMS patients, BMS patients versus RRMS patients, and RRMS patients versus SPMS patients. The numbers of a priori contrasts were determined by the available degrees of freedom, and their nature was decided on the basis of the clinical evolution of the disease (i.e., MS onset is either RRMS or PPMS; RRMS can evolve to SPMS or stabilize to BMS). A multivariate regression model with T2 lesion load, T1 lesion load, and average lesion MTR as independent variables was used to assess the MRI measures that best correlated with the NABT MTR. Another multivariate regression model was run with T2 lesion load, T1 lesion load, average lesion MTR, and NABT MTR as independent variables to assess the MRI measures that best correlated with brain volume. The effect of age, sex, disease duration, disease course, T2 and T1 lesion loads, average lesion MTR, and average NABT MTR on the probability of having low (EDSS score ≤ 4.0) or high (EDSS score < 4.0) disability scores was assessed using a multivariate logistic regression model.
Results.
Clinical data.
A total of 77 MS patients (49 women and 28 men; mean age, 37.3 ± 10.1 years) and 20 control subjects (13 women and 7 men; mean age, 35.4 ± 6.3 years) were entered in the study. Thirty-three patients had RRMS, 20 had SPMS, 13 had PPMS, and 11 patients had BMS. The demographic and clinical characteristics of the MS patients studied are reported in table 1.
Demographic and clinical characteristics of the patients with different MS clinical phenotypes studied
MRI results.
No abnormalities were found on the scans from control subjects. In table 2, the hyperintense T2 and hypointense T1 lesion loads together with their ratio (T1/T2 ratio), average lesion MTR, and brain volume from the overall MS population studied and for each of the four clinical phenotypes are reported. The conventional MRI characteristics of these patients are similar to previously published results.18,36,37 Patients with MS had lower brain volume than control subjects (418.4 ± 58.1 mL versus 482.5 ± 25.4 mL, p < 0.0001).
Lesion loads on T2- and T1-weighted scans in the different clinical phenotypes studied
In table 3, NABT MT metrics from control subjects, the overall MS population studied, and each of the four clinical MS phenotypes are reported. The figure shows the MTR histograms of the NABT obtained from the whole patient population and in each clinical subgroup compared with that from control subjects. Patients with MS had lower NABT MTR histogram metrics than control subjects. The p values of the comparisons were less than 0.0001 for the average NABT MTR, MTR25, MTR50, MTR75, and peak position, and less than 0.01 for histogram height.
Mean (±SD) MT metrics in NABT of control subjects and of different MS phenotypes studied
Figure (a through e). Normal-appearing brain tissue magnetization transfer ratio (MTR) histograms from the whole MS patient population and each clinical subgroup. In each graph, the histogram derived from control subjects is shown as a gray line for comparison. The entire MS population is presented in (a), relapsing–remitting MS in (b), secondary progressive MS in (c), benign MS in (d), and primary progressive MS in (e). Note that these average histograms do not necessarily show the same trends as the statistics presented in tables 3 and 4⇓ because the average histograms also reflect heterogeneity within the populations of the separate groups. NAWM = normal-appearing white matter.
In table 4, the p values of the four a priori contrasts for all the NABT MTR histogram metrics are reported. Average NABT MTR, MTR25, MTR50, MTR75, and peak position from patients with RRMS or from patients with PPMS were all significantly lower than those from control subjects. Compared with control subjects, patients with PPMS, but not those with RRMS, also had a significantly lower histogram peak height (p = 0.01). No difference was found between the NABT MTR histograms from patients with RRMS and BMS. Patients with SPMS had a lower MTR75 (p = 0.04) and peak height (p = 0.05) than those with RRMS.
Results of the a priori pairwise comparisons between control subjects and different MS clinical phenotypes
The average lesion MTR (r = 0.68, p < 0.0001) correlated highly with the average NABT MTR. All the other factors were excluded from the multiple regression model. NABT MTR (r = 0.70, p < 0.0001) was the parameter that correlated best with brain volume. Another significant contribution came from the T2 lesion load (p = 0.001), whereas the other factors were excluded. The disease course was the only factor influencing significantly the probability of belonging to the group with a higher disability score (p < 0.0001).
Discussion.
This study indicates that, in the brain of patients with MS, there are definite abnormalities that go undetected when using conventional MRI, and they are considered part of the NABT. It also shows that such abnormalities may be functionally relevant, because they may contribute to the development of brain atrophy and they tend to be more evident in patients with SPMS.
Three pieces of evidence already suggested that microscopic changes in the normal-appearing cerebral tissue from MS are not meaningless. First, MTR changes can be detected in the NAWM subsequently involved by new MS lesions.38,39 Second, the MTR values of the NAWM away from20,40 or around20 T2-visible lesions from patients with SPMS are lower than those of the corresponding NAWM regions from patients with less disabling RRMS. Third, NAWM from patients with SPMS has an N-acetylaspartate-to-creatine ratio that, on average, is 8.2% lower than the NAWM from patients with RRMS.19 In patients with RRMS, a progressive reduction of this ratio is detectable over time, and its decrease correlates strongly with changes in physical disability.19 However, all these studies evaluated only a small part of the NAWM,19,20,40 whereas the current study provides an overall assessment of the cerebral tissue spared from T2-visible lesions.
MTR is reduced in both lesions and NAWM in MS patients.7,14,18,20,21,40 In the current study we obtained MTR histograms of the NABT only, by removing accurately from the coregistered MTR maps all the pixels known to belong to T2-visible lesions. Thus, the characteristics of the histograms we obtained only reflect subtle changes outside T2 visible lesions. Focal changes in the normal-appearing cerebral tissue are expected to decrease the peak height and increase the number of pixels with low MTR values. The more severe the pathologic process in these focal abnormalities, the greater the relative increase in the histogram at low MTR. Mild but more widespread changes to the white matter would cause a larger reduction in the peak height, accompanied by a broadening of the peak at its left side because more of the tissue is affected, but with little or no increase at very low MTR. Thus the reduction in the mean MTR might be more subtle. In an extreme case, when most of the white matter is affected diffusely, it would also be possible for the peak position to move to the left, because little tissue would remain at truly normal MTR.
Because NAWM represents the largest part of the NABT included in our MTR histograms, we believe that microscopic white matter abnormalities rather than abnormalities in the gray matter (i.e., cerebral and cerebellar cortex, and basal ganglia) may be responsible for the MTR histogram changes we observed. Nevertheless, it is clear that lesions in or adjacent to the cerebral cortex, which can be imaged using fast fluid-attenuated inversion recovery sequences,41,42 may have been missed in the current study and may also have contributed to the MTR histogram findings. On the contrary, the role of pathologic changes in the basal ganglia is likely to be minor, if present at all, due to the low frequency of clinical or MRI involvement of these structures in MS.43,44 Finally, in healthy volunteers, some studies40,45,46 but not all47 showed that the MTR of gray matter is slightly lower than that of white matter. Thus, at least in patients with high T2 lesion loads, the relatively higher proportions of gray matter included in the analysis might have influenced, albeit modestly, the NABT MTR histogram metrics, irrespective of the presence of MS-related abnormalities.
Against this background, we can interpret our results. When compared with control subjects, we found changes of the MTR histograms from NABT both in patients with RRMS and PPMS. This suggests that NABT changes may occur relatively early during the course of the disease and independent from the clinical phenotype at onset. However, the characteristics of the MTR histogram changes we found in the two groups of patients suggest that the nature of the changes in the normal-appearing cerebral tissue may be different. The histogram from patients with RRMS had the lowest average MTR and peak position, and the highest peak height when compared with those from all the other MS phenotypes. This suggests small, discrete lesions beyond the resolution of conventional scanning as the most likely change occurring in a relatively large portion of the normal-appearing cerebral tissue. This is in agreement with the results of previous studies indicating that microscopic lesions do occur in MS.48-52 Barbosa et al.48 showed that T1 and T2 relaxation times are increased in the NABT from MS patients and that these changes, which may involve a large portion of the NABT, are always of the size of one or two pixels. In addition, it is also known that more lesions and higher lesion volumes can be detected when decreasing the MR slice thickness49-51 or when using high-field scanners.52
The situation seems to be different in patients with PPMS. The histogram from these patients has the lowest peak height when compared with those from all the other MS phenotypes, whereas the average histogram MTR and peak position are similar to those from control subjects. This suggests that the amount of residual normal cerebral tissue (i.e., tissue with truly normal MTR) is much lower in PPMS than in the other MS phenotypes and suggests widespread but mild changes as the most likely underlying pathology. This agrees with the results of studies using whole-brain histogram analysis18or ROI analysis.53 That widespread changes might be the MRI hallmark of patients with PPMS is also suggested by the demonstration of diffuse increased signal intensity on PD-weighted images of the spinal cord in such patients.37
In our sample, patients with SPMS had more extensive T1 and T2 abnormalities than those with RRMS. However, the finding that patients with SPMS also had a lower MTR histogram peak height than those with RRMS suggests that, among other factors, a progressive reduction of cerebral tissue with truly normal MTR may be responsible for the evolution from RRMS to SPMS. More difficult to explain is why we found no difference in any of the MTR histogram metrics from RRMS patients and BMS patients. One possible explanation is that the evolution toward BMS requires the pathologic process within T2-visible lesions to be mild. Previous results obtained using different MR techniques seem to support this explanation.14,18,54-56
Although we can only speculate about the nature of the subtle changes potentially responsible for our MTR histogram findings, the demonstration that NABT MTR correlates highly with the brain volume makes tissue loss, rather than transient “inflammatory” changes, the most likely pathologic substrate. Several studies demonstrated that brain atrophy is a frequent finding in patients with MS,57 and it is well correlated with disease severity and its progression over time.32,57 The strongly predictive values of T2 lesion volume and NABT MTR on brain volume indicates that not only T2-visible lesions25 but also subtle NABT changes may be involved in determining tissue loss in patients with MS. This highlights once again the need for accurate estimates of the MS pathology to be obtained.
Another relevant issue is to determine whether the MTR histogram changes we observed are secondary to small focal abnormalities independent of lesions visible on conventional scans48,50,52 or damage to axons traversing macroscopic lesions and resulting in Wallerian degeneration in areas away from such abnormalities.58,59 The results of the current study allow us to propose that these abnormalities are not mutually exclusive and perhaps are different in the different clinical phenotypes of the disease. The association we found between average lesion MTR and NABT MTR, and the large amount of T1 and T2 lesions make secondary axonal degeneration a likely contributor to the changes seen in normal-appearing cerebral tissue of patients with RRMS and SPMS. On the contrary, the paucity36,37,60 and the relatively mild intrinsic damage18 of macroscopic lesions from PPMS favor the role of subtle changes independent of larger abnormalities.
Acknowledgments
Supported by grants from Farmades/Schering (C.T., B.V.) and TEVA (M.P.S.), Italy.
Acknowledgment
The authors thank Mr. Clodoaldo Pereira for his skillful technical assistance in collecting the MR images.
Footnotes
-
Presented at the 51st meeting of the American Academy of Neurology; Toronto, Canada; April 17–24, 1999.
- Received May 5, 1999.
- Accepted July 29, 1999.
References
Disputes & Debates: Rapid online correspondence
NOTE: All authors' disclosures must be entered and current in our database before comments can be posted. Enter and update disclosures at http://submit.neurology.org. Exception: replies to comments concerning an article you originally authored do not require updated disclosures.
- Stay timely. Submit only on articles published within the last 8 weeks.
- Do not be redundant. Read any comments already posted on the article prior to submission.
- 200 words maximum.
- 5 references maximum. Reference 1 must be the article on which you are commenting.
- 5 authors maximum. Exception: replies can include all original authors of the article.
- Submitted comments are subject to editing and editor review prior to posting.