Coefficient D_av Is More Sensitive Than Fractional Anisotropy in Monitoring Progression of Irreversible Tissue Damage in Focal Nonactive Multiple Sclerosis Lesions

Phantom Experiments

As a measure of quality, the DTI sequence was preliminarily tested on a phantom of water doped with NiSO₄ to compare our diffusion and anisotropy measurements with reference values of the literature.

Participants

Twenty-six patients with clinically definite MS (28) were sequentially enrolled in the study between February and August 2000. Of these, 18 (10 with relapsing-remitting disease, eight with secondary progressive disease) fulfilled the inclusion criteria. To be included in the study, patients could not have received immunosuppressive, immunoregulatory, or steroid therapy for at least 6 months before study initiation. They also could not have had a relapse during the 3 months preceding the study. Patients with relapsing-remitting disease (four men and six women; mean age, 26 years; age range, 18–37 years) had had clinically definite MS for at least 2 years with a relapsing-remitting course. The mean age at onset of disease was 23 years (range, 16–32). Expanded Disability Status Scale score was 3 (range, 1.0–5). Patients with secondary progressive MS (four men and four women; mean age, 38 years; age range, 29–46 years) had experienced a slowly progressive increase in disability for at least 6 months, with or without superimposed relapses, after an initial relapsing-remitting course (29). The mean age at disease onset was 26 years (range, 18–37), and mean duration of the relapsing-remitting phase was 11 years (range, 3–25). Expanded Disability Status Scale score was 6.5 (range, 4.5–7.5). Eighteen sex- and age-matched healthy volunteers (eight men and 10 women; mean age, 29.3 years; range, 20–48 years) served as control participants. The study was approved by the local ethics committee, and all participants provided written informed consent before study initiation.

MR Data Acquisition

Both the patients and the volunteers underwent a combined conventional and diffusion tensor MR imaging protocol. A Siemens Vision 1.5-T whole body unit (Erlangen, Germany), equipped with echo-planar capability, and a circularly polarized 270-mm-diameter head coil were used for image acquisition. The MR imaging protocol included the following: 1) axial view T1-weighted spin-echo (650/14 [TR/TE]; flip angle, 70 degrees); 2) axial view dual-echo T2-weighted turbo spin-echo (3800/22, 90[TR/first TE, second TE]; echo train length, 5); 3) axial view DTI single shot echo-planar pulse sequence (5000/98; six b values ranging from 0 to 530 s/mm⁻², with the gradients used for b ≠ 0 applied in six noncollinear directions); and 4) contrast-enhanced (IV administered 0.1 mmol/kg gadopentetate dimeglumine) axial view T1-weighted spin-echo (650/14; flip angle, 70 degrees) imaging. For the sequences described in imaging protocols 1, 2, and 4, 21 contiguous sections were acquired with 5-mm thickness, 256 × 256 matrix, and 240 × 240 field of view (resolution, 0.94 × 0.94 × 5 mm). The sections were positioned to run parallel to the anterior commissure–posterior commissure line and to cover the whole brain. For the DTI sequence, 15 axial view sections with 5-mm thickness, 128 × 128 matrix, and 240 × 240 field of view (resolution, 1.88 × 1.88 × 5 mm) were acquired by positioning the sections to match exactly the 15 central sections of the conventional images. The DTI sequence is implemented with self-correction for eddy current effects by using both compensation gradients and phase correction (30). The total examination time was approximately 20 minutes.

MR Imaging Data and Statistical Analysis

Collected MR imaging studies were transferred via Ethernet to an independent workstation (Silicon Graphics SGI) running a homemade MR image processing software implemented under MATLAB 5.3.1 (Mathworks). After pixel-by-pixel diagonalization of the diffusion tensor matrix, maps of the mean diffusivity coefficient D_av (equal to one-third of the trace) and maps of FA were derived. FA is a robust estimator of diffusion anisotropy that is able to map diffusion anisotropy with the greatest detail and signal-to-noise ratio compared with other anisotropy indices, such as relative anisotropy and volume ratio (31). Values for FA, a unitless ratio of diffusion coefficients, range from 0 to 1, where 0 represents isotropic diffusion and 1 represents diffusion in a single direction. T2-weighted and proton density–weighted images were analyzed for lesion identification by one experienced reader (observer 1). Based on their appearance on contrast-enhanced T1-weighted images, 12 enhancing lesions were identified and excluded from further analysis. With guidance from T2- and proton attenuation-weighted images, regions of interest of uniform shape and size (approximately eight pixels corresponding to 28 mm², given a pixel size of 1.88 × 1.88 mm) were outlined in the plaques on the non-diffusion-weighted b0 (b value=0) images of the DTI dataset and subsequently projected onto the coefficient D_av and FA maps. Plaques <15 mm² (approximately four pixels) were excluded. Likewise, equivocal areas adjacent to ventricular surfaces were not considered. A total of 83 plaques were initially assessed in 18 patients.

Regions of interest recorded on the images with no diffusion weighting (b=0) were transferred to the T2-weighted images and were manually adjusted, considering the lack of spatial correspondence between diffusion- and T2-weighted images, to fit into corresponding lesions. The same regions of interest were also used to generate matching regions of interest on T1-weighted images. Two other experienced readers (A.C.-S., F.F.), blinded to the diffusion and anisotropy data, patient’s identity, and clinical status, assessed by consensus the degree of hypointensity of the plaques selected on unenhanced T1-weighted images in comparison with NAWM and gray matter, as previously described (13). T1 isointense lesions had the same signal intensity (SI) as that of NAWM, T1 mildly hypointense lesions were of SI equal to or higher than that of gray matter, and T1 severely hypointense lesions were of SI lower than that of gray matter (Fig 1). To correct for random variation in overall SI among examinations, analysis of the selected lesions also included calculation of the contrast ratio on T1-weighted images. The contrast ratio (SI_lesion/SI_CSF) was defined as the SI of a lesion divided by the SI of CSF, analyzed in the same section as the investigated lesion.

• Download figure
• Open in new tab
• Download powerpoint

Fig 1.

Three examples of MS lesions (arrows) with different T1 appearance. Top, isointense lesion (FA=0.30 ± 0.06, coefficient D_av=1.10 ± 0.07 × 10⁻³ mm²/s); middle, mildly hypointense lesion (FA=0.27 ± 0.05, coefficient D_av=1.24 ± 0.06 × 10⁻³ mm²/s); bottom, severely hypointense lesion (FA=0.19 ± 0.04, coefficient D_av = 1.76 ± 0.07 × 10⁻³ mm²/s). For each row, from left to right: axial view fast spin-echo T2-weighted image, spin-echo T1-weighted image, coefficient D_av map, and FA map.

For each region of interest in the lesions, a region of interest of identical size was drawn in a symmetric area of NAWM in the contralateral hemisphere. The NAWM region paired with each plaque was at the closest feasible location as the plaque. If the matching NAWM region in the contralateral hemisphere showed abnormal SI, it was discarded together with the corresponding plaque. After removal of seven plaques, a total number of 76 lesions and 76 NAWM regions in the contralateral hemisphere were assessed. To minimize the effect of FA variations due to normal regional differences in the brain architecture, relative FA variation was computed for each plaque by using as reference the contralateral NAWM region. Specifically, where FA_lesion is the FA of the lesion, and FA_{contralateral} is the FA of the NAWM region symmetric to the lesion in the contralateral hemisphere. Finally, for each lesion, a corresponding area of normal white matter was selected in the sex- and age-matched volunteers by choosing the section that most closely matched with the section in the patient, thereby obtaining 76 regions of interest. The mean values ± SD for FA and coefficient D_av were recorded for each region of interest placed in the plaques, NAWM, and normal white matter.

A paired Student’s t test was used to assess differences on the values of absolute FA and coefficient D_av between the following pairs: plaques and NAWM regions, plaques and normal white matter regions, and NAWM regions and normal white matter regions. The paired Student’s t test was used because of the intrinsic region-related FA variation, which required pairing of each value in the plaque with its analogous area in the matched NAWM or normal white matter region. To account for multiple comparisons, data were analyzed by conducting one-way analysis of variance using a model with the T1-based lesion appearance (T1 isointense, T1 mildly hypointense, and T1 severely hypointense) as the grouping factor and either absolute FA, coefficient D_av, or ΔFA (a relative FA measure that minimizes the effect of FA spatial dependence) as the dependent measure. Analysis of variance was applied after verifying that the variables adhered to a normal distribution by using Kolmogorov-Smirnov tests and was followed by Bonferroni post hoc comparisons.

The values of FA, coefficient D_av, and ΔFA were then correlated with the SI_lesion/SI_CSF ratio by calculating the Pearson correlation coefficient. The level for statistical significance was set at P < .05.

Phantom Experiments

In quantitative diffusion tensor measurements of phantom of doped water, coefficient D_av and FA, computed on regions of interest of size comparable with those used for patients, were found to be 2.28 ± 0.03 (× 10⁻³ mm²/s) and 0.05 ± 0.02, respectively. These measured quantities are in close agreement with previously published data (27).

Participants

DTI measurements yielded high quality data for both patients with MS and healthy volunteers. Of the 76 nonenhancing lesions, 25 were isointense (80% on images of patients with relapsing-remitting disease and 20% on images of patients with secondary progressive disease), 26 were mildly hypointense (40% on images of patients with relapsing-remitting disease and 60% on images of patients with secondary progressive disease), and 25 were severely hypointense (30% on images of patients with relapsing-remitting disease and 70% on images of patients with secondary progressive disease) on unenhanced T1-weighted images.

DTI Measures in Various White Matter Regions

Table 1 summarizes the mean value (± SD) for FA and coefficient D_av as measured in MS plaques, NAWM regions, and normal white matter regions. The lowest FA values were measured in plaques, and then, in ascending order, in NAWM and normal white matter. By contrast, coefficient D_av values were highest in plaques, intermediate in NAWM, and lowest in normal white matter. On average, a 13% decrease in FA and an 11% increase in coefficient D_av occurred in NAWM compared with normal white matter, and a 31% decrease in FA and a 35% increase in coefficient D_av occurred in plaques compared with NAWM. The Student’s t tests revealed significant differences in the values of FA and coefficient D_av for all paired comparisons [plaques and NAWM regions (FA, P < .0001; coefficient D_av, P < .0001); plaques and normal white matter regions (FA, P < .0001; coefficient D_av, P < .0001); NAWM and normal white matter regions (FA, P < .001; coefficient D_av, P < .0001)].

Absolute FA, Coefficient D_av, and ΔFA Values in the T1-Based Lesion Groups

Table 2 summarizes the mean value (± SD) for FA, coefficient D_av, and ΔFA as measured in the group of T1 isointense, T1 mildly hypointense, and T1 severely hypointense lesions. The lowest FA values were measured in the T1 severely hypointense lesion group, and then, in ascending order, in the T1 mildly hypointense group and T1 isointense group. By contrast, coefficient D_av values were highest in the T1 severely hypointense lesion group, intermediate in the T1 mildly hypointense group, and lowest in the T1 isointense group. On average, a 16% reduction in FA and a 16% increase in coefficient D_av occurred when moving from the T1 isointense lesion group to the T1 mildly hypointense lesion group and a 9% reduction in FA and a 14% increase in coefficient D_av occurred when moving from the T1 mildly hypointense lesion group to the T1 severely hypointense lesion group. One-way analysis of variance, with appearance of MS lesions on T1-weighted images as the grouping factor, revealed an overall effect of the T1-based lesion type, both for FA (P < .01) and coefficient D_av (P < .0001). However, post hoc contrast analysis disclosed a significant difference in the values of FA only between the T1 isointense group and the T1 severely hypointense lesion group (P < .01) but not between T1 mildly hypointense lesions and either one of these two groups. A significant difference was found in the values of coefficient D_av among all investigated lesion groups (T1 severely hypointense > T1 mildly hypointense > T1 isointense). No significant effects of the T1-based lesion type were found in the values of ΔFA.

Correlations between SI_lesion/SI_CSF Ratio and Either FA, Coefficient D_av, or ΔFA Values

Considering all MS plaques, an inverse correlation was found between coefficient D_av and the T1-weighted SI_lesion/SI_CSF ratio (r=−0.58; P < .001) whereas no correlation was found between either FA or ΔFA and the SI_lesion/SI_CSF ratio (Fig 2).

• Download figure
• Open in new tab
• Download powerpoint

Fig 2.

Scatter plots for the 76 lesions studied.

A, Coefficient D_av versus SI_lesion/SI_CSF ratio. Significant inverse correlation was shown.

B, FA versus SI_lesion/SI_CSF ratio. No correlation was shown.

DTI Measures of Focal Nonactive MS Lesions

In this study, we used the degree of hypointensity on T1-weighted images as a marker of matrix destruction and axonal loss in patients with MS and analyzed the ability of DTI-based parameters to detect differences among three groups of nonactive MS lesions with increasing T1 hypointensity. We found that coefficient D_av values increased and FA values decreased with increasing T1 hypointensity. The mean values of coefficient D_av were significantly different among the three T1-based groups of lesions, thus allowing differentiation among them. More importantly, a correlation was found between coefficient D_av and the SI_lesion/SI_CSF contrast ratio, a semi-quantitative measure of the degree of T1 hypointensity in lesions. This finding suggests that both MR parameters reflect the extent of axonal damage in patients with MS. In contrast to coefficient D_av, the mean values of FA were not significantly different among the three T1-based lesion groups, and a significant difference emerged only when the two extreme lesion categories (T1 isointense and T1 severely hypointense) were considered. Moreover, no correlation was found between FA values and the SI_lesion/SI_CSF contrast ratio in lesions, a finding bringing further evidence to the concept that FA and the degree of T1 hypointensity provide at least partially independent information. One possible explanation is that measured FA values in the brain of patients with MS not only reflect the structural changes determined by the disease process but also the point-by-point variation of diffusion anisotropy within the normal brain (38). Brain white matter areas in which the fibers are incoherently oriented, such as the centrum semiovale or the U fibers, have normally relatively low FA values even though they are highly myelinated (21). Thus, FA may not be significantly reduced by MS lesions involving these regions.

To go some way toward the limitation of the regional dependence of diffusion anisotropy, we also conducted a statistical analysis on the values of ΔFA, a relative FA measure that minimizes the effect of FA spatial dependence. Just like absolute FA, relative FA values did not show correlation with the SI_lesion/SI_CSF contrast ratio nor did they enable discrimination among the three T1-based lesion categories. We therefore conclude that coefficient D_av may be more sensitive than FA in determining the degree of axonal damage in patients with MS.

Comparisons with Previous DTI Studies

Previous studies (17–20, 22, 25, 39, 40) have reported increased coefficient D_av and decreased FA in MS plaques compared with NAWM, with observed values in the ranges of 0.90 to 1.59 (× 10⁻³ mm²/s) and 0.23 to 0.50, respectively. The mean coefficient D_av value of 1.54 (× 10⁻³ mm²/s) and the mean FA value of 0.22 measured in plaques in the current investigation are, with few exceptions (26), similar to those reported in most the above mentioned DTI studies (22–25). Moreover, some of these previous studies (23, 24, 26) have reported lower FA values in T1 hypointense than in T1 isointense lesions, a finding in accordance with our results when considering the T1 isointense and the T1 severely hypointense lesions only. Our mean FA measurements of 0.26 in T1 isointense lesions are very similar to the value of 0.27 reported by Filippi et al (24) but lower than the values of 0.43 and 0.51 reported by Bammer et al (27) and Werring et al (26), respectively. In only one previous study (27) were areas of extensive tissue destruction studied by means of DTI indices, and the reported FA value of 0.23 is very close to the value of 0.19 we found in the group of T1 severely hypointense lesions.

DTI Measures in Widespread Subtle Pathologic Abnormality

In agreement with previous investigations (39, 41, 42), we found significantly increased coefficient D_av and decreased FA values in NAWM compared with normal white matter, a finding supporting the idea that MS is a widespread disease able to cause microscopic changes that extend well beyond the margins of the focal lesions (43). Histopathologic and biochemical studies have reported astrocytosis, demyelination, perivascular inflammation, and increased water content in the NAWM of patients with MS (44, 45). More recently, substantial loss of axons in the NAWM has been suggested by spectroscopic studies (46, 47), MR imaging measures such as T1 and T2 TRs (48), and magnetization transfer imaging (49) and has been confirmed by direct quantitation of axonal attenuation and volume depletion on postmortem MS brains (50). Several possible mechanisms could account for the extensive loss of axons in the NAWM, including denervation by transection in remote lesions (33, 34) and effects of diffusable neurotoxic factors from lesions (51). Because axonal damage in NAWM may have functional consequences for patients with mild disability early in the disease course, in vivo quantification of disease burden has several important implications, both in drug trials and patient treatment.

Limitations of the Study and Conclusions

The rationale for this study design is based on previous studies that have attempted to correlate appearance on T1-weighted images with pathologic findings (13). Based on these correlation studies, the assumption was made that the degree of T1 hypointensity and the extent of axonal loss are highly associated events in focal nonactive MS lesions. Under these circumstances, our data suggest that coefficient D_av may be more sensitive than FA in determining the degree of axonal damage in focal nonactive MS lesions. Thus, in contrast to coefficient D_av, the level of sensitivity of FA may not be sufficient for monitoring disease progression in patients with MS. Unfortunately, the lack of a direct correlation of DTI parameters with pathologic findings precludes firm conclusions regarding the pathologic specificity and sensitivity of DTI measures.

Considering the strong dependence of lesion FA changes on the preexisting normal local architecture of white matter (38), inferences regarding the underlying pathologic substrate should be made cautiously when absolute FA values are used. Although this spatial dependence can be diminished by calculating the relative FA variation, such calculation is not without limitations. Relating lesion FA to corresponding values in supposedly healthy regions of the contralateral hemisphere may lead to underestimation of FA changes because these regions may not be normal, as confirmed by the current study. In one previous investigation (27), lesion FA was related to areas of NAWM located in the vicinity of the lesion, and the resultant region-related FA measurements yielded a markedly reduced overlap between lesion types and normal white matter compared with absolute FA values. However, NAWM in the vicinity of MS lesions may be more profoundly altered than NAWM remote from lesions, as recently shown (25). Ideally, regional comparisons would benefit from spatial normalization and exact interpolation of patient’s anatomic structures onto an anisotropy-based atlas generated from volunteer studies (27).