INTRODUCTION
Patients with primary progressive multiple sclerosis (PP-MS) pose a challenge in monitoring early disease progression (1, 2). PP-MS patients develop progressive disability without relapse or remission, and the lack of diffuse macroscopic damage and lesion formation provides difficulty in guiding treatment options (3). The pathophysiological mechanisms of PP-MS are unknown, yet an interplay of neurodegeneration and demyelination has been suggested (4-7). In addition to the well-recognized role of grey matter (GM) pathology, histopathological studies from post-mortem samples of PP-MS patients have demonstrated that normal appearing white matter (NAWM) shows extensive and diffuse pathology, and the possibility to fully characterize NAWM damage could elucidate the mechanisms of MS progression (8-10).
Quantifying water diffusivities throughout WM tissues may elucidate microstructural changes that characterize NAWM, providing measures that determine clinical outcome over time in diseases such as PP-MS (10,11). Advanced MRI techniques such as diffusion tensor imaging (DTI) have allowed for a greater visualization of microstructural changes of PP-MS cerebral white matter (WM) (11) through its ability to identify water diffusion at the microstructural level (12, 13). DTI has been implemented to detect focal MS lesions, but DTI derived metrics such as fraction anisotropy (FA) and mean diffusivity (MD) lack specificity to demyelination and axonal loss (14). Demyelination and axonal loss have similar impacts on DTI derived metrics, limiting its use to determine small modulatory changes in cerebral tissue rather than characterizing the NAWM of PP-MS (15). Due to its dependence on deriving MD and FA, DTI also only functions through the approximation of low b-values (1000s/mm2 ) that comprise the integrity of tissue contrast and quality of images (16). DTI is only capable of monitoring microstructural diffusivity patterns in strictly anisotropic environments, limiting its use to WM regions rather than extending to the corticospinal tract (CST) where significant inflammation occurs in PP-MS (17, 18). Therefore, DTI-derived metrics can quantify small modulatory changes in MD and FA, but fail in discriminating the pathological processes underlying complex diseases such as MS (19).
More recently, diffusion kurtosis imaging (DKI) has been introduced as an extension of DTI that overcomes its limitations. In complex microtissue such as cerebral WM, water diffusivity is highly anisotropic and non-Gaussian in nature due to axonal fibrosity (20). DKI is able to characterize complex microstructural tissues that deviate from the Gaussian form. This deviation can be regarded as the kurtosis of a system, and non-Gaussian distributions have a positive kurtosis value (k>0). The higher kurtosis values allow for the quantification of diffusion in anisotropic and isotropic environments (21, 22). This allows DKI to be sensitive in its detection of microstructural changes in the WM and CST (23). DKI estimates non-Gaussian distribution in WM, thereby providing derived metrics that accurately reflect cerebral neurodegeneration in heterogeneous tissue, allowing for the full characterization NAWM abnormalities such as intra-axonal damage; axonal loss; extracellular inflammation; gliosis and demyelination (24, 25).
Empirical diffusion metrics derived from DKI provide indirect characterization of microstructure, resulting in ambiguity regarding WM tissue properties (25). Biophysical modeling of the WM has been a recent focus in the field of MRI to interpret the diffusion metrics derived from DKI. A WM tract integrity (WMTI) model has been proposed to elucidate that mechanisms behind WM degeneration and its correlation to decreased clinical function (26, 27). Through DKI’s estimation of non-Gaussian probability of diffusion, it can be incorporated with biophysical models such as WMTI to provide metrics that accurately reflect neural degeneration (28-30). WMTI is a two compartment model that divides the WM into the intra-axonal space and extra-axonal space, providing several metrics that reflect PP-MS disease progression (25). WMTI derived metrics include axonal water fraction (AWF), tortuosity ( T ) intrinsic axonal diffusivity (D axon) , radial extra-axonal diffusivity (D e,radial) , and axial extra-axonal diffusivity (D e,axial) . The AWF and D e,radial, T are sensitive to demyelination and axonal loss, and D axon and D e,axial are sensitive to structural changes along the axon bundle in the intra-axonal space (27).
While DKI derived WMTI-metrics have been applied to investigate neurodegenerative diseases such as Alzheimer’s, there are only a few studies utilizing DKI derived WMTI-metrics to analyze microstructural changes in MS (28-31). In regards to PP-MS, previous studies have only used DTI as a mode of exploring WM integrity, quantifying the increased MD and decreased FA. Still, very little is known regarding the pathological process underlying short term disease progression resulting in worsened clinical outcome over time in PP-MS (32).
Therefore, the aims of this study were to utilize novel DKI-derived WMTI metrics to characterize the presence and extent of WM abnormalities, to assess the impact of WM abnormalities on clinical disability, and to investigate the sensitivity of WMTI metrics to short-term disease progression in PP-MS.
METHODOLOGY
Subjects
Twenty-six patients who met the modified McDonald diagnostic criteria and presented a primary-progressive course of MS were prospectively enrolled. Twenty sex- and age-matched healthy subjects served as controls (11F/9M; mean age, 51.1 years; range, 34–63 years) for the comparison of MRI metrics. Inclusion criteria for PP- MS patients were: (i) age between 25–65 years; (ii) an Expanded Disability Status Scale (EDSS) lower than 6.5 at screening visit; (iii) disease duration up to 15 years. The use of immunomodulatory drugs was allowed but, if treated, patients had to be on current treatment for at least 6 months. Exclusion criteria for all subjects were: (i) neuropsychiatric disorders other than MS, (ii) ophthalmological pathologies (i.e., diabetes mellitus or glaucoma), (iii) history of alcohol or drug abuse, (iv) contraindications to MRI. Twenty patients with baseline and 6 months clinical and MRI examination were included in the longitudinal analysis.
Clinical assessment
All subjects underwent clinical and MRI assessment on the same day. Clinical disability was assessed at baseline, after six and twelve months with the EDSS, timed 25-foot walk (T25FW) test, 9-hole-peg test (9-HPT) and Symbol Digit Modality Test (SDMT). Best-corrected visual acuity (VA) was assessed binocularly at the same time points, using low contrast letter acuity Sloan charts (1.25%, 2.5%, and 100%) at 4 m (Precision Vision, IL, USA). SDMT raw scores were converted to z-scores. Clinical worsening was defined as EDSS score increase of one point if the baseline EDSS score was less than or equal to five, an increase of 0.5 if it was greater than five or as a change of >20% for 25-FWT and 9-HPT scores. Disease activity was defined as presence of new T2 lesions during the assessment period. Disease progression was defined as clinical worsening and/or disease activity (i) at month six compared with baseline, confirmed at month 12 or (ii) at clinical follow-up visit 12 months after study termination compared with month 12 and/or disease activity (i) at month six compared with baseline, confirmed at month 12.
MRI acquisition
MRI was performed using a 3.0 T scanner (Philips Achieva, The Netherlands) with an 8- channel SENSE phased-array head coil (Philips Achieva, The Netherlands). The MRI protocol included the following sequences: a) axial dual echo TSE sequence: TR = 2500 msec, TE1 = 10 msec, TE2 = 80 msec, FOV = 230×230 mm, matrix size = 512×512, 46 contiguous 3 mm-thick slices; b) sagittal 3D T1-weighted turbo field echo sequence: TR = 7.5 msec, TE = 3.5 msec, TI = 900 msec, flip angle = 8°, voxel size = 1x1x1 mm, 172 contiguous slices; c) a twice-refocused spin echo EPI sequence for DKI with b-values of 1000 and 2000 s/mm2 and 30 directions each (repeated twice with opposite phase encoding), in addition to six b=0 s/mm2 images (TR = 8550 msec, TE = 89.5 mssec, flip angle = 90°, spatial resolution = 1.98×1.98x2mm).
Lesion segmentation
For the MS patients, quantification of T2-hyperintense and T1-hypointense lesion volume was performed in each patient by a single experienced observer unaware of subject identity, employing a semiautomated segmentation technique (DISPLAY, Montreal Neurological Institute [MNI]) as previously described (28).
DKI image processing
DKI data were transferred to an offline workstation and processed using in-house developed software in Matlab (R2015a, Math Works, Inc, Natick, MA) to derive the following WM tract
integrity metrics for a coherently aligned single fiber bundle: AWF, Daxon, De,axial, De,radial, and T. In order to register the lesions masks on the DKI maps and extrapolate the NAWM, first T2-weighted and T1-weighted images were co-registered on b=0 images using the automated affine registration tool FLIRT with boundary-based registration, then the resulting transformations were applied to the corresponding lesion masks. For each subject, we used his own T2 lesion mask to identify which voxels of the considered WM tract were affected by a lesion and to compute the lesion volume inside each tract.
Regions of Interest (ROI) analysis was performed to investigate group differences and tissue- specific microstructural damage in WM tracts. ROI analysis was restricted to the corpus callosum (CC), posterior thalamic radiation (PTR) and CST, whose well- ordered axonal structure best corresponds to the WM model used to derive WMTI metrics. All ROI were co-registered on each subject’s diffusion space using the non-linear registration tool FNIRT. Mean values of the different WMTI metrics were then extracted from patients NAWM and healthy controls WM. Since the WM model used to derive WMTI metrics cannot be applied within WM lesions, mean values of FA and MD, which can be derived from DKI images, were extracted from the T2-hyperintense lesions to obtain an estimate of tissue disruption in macroscopic lesions.
Statistical analysis
Statistical analysis was performed using SPSS 23.0 (SPSS, Chicago, IL). A Shapiro-Wilk test was used to test the normality of the data. Mann-Whitney and Fisher exact test were applied to assess differences in terms of age, gender and disease duration between patients and control, progressed and not progressed patients. An analysis of variance model on ranks was applied to investigate differences in MRI metric between patients and control at baseline, taking into account age and gender as covariates. Correlations between MRI metrics and clinical parameters were tested using non-parametric Spearman rank correlation coefficient. Changes in WMTI metrics over 6 month follow-up were assessed using non-parametric Wilcoxon test. A logistic regression analysis followed by a receiver operating characteristics (ROC) curve analysis was performed to assess the performance of WMTI metrics in discriminating between progressed and not progressed patients. Lastly, an analysis of variance model on ranks was applied to investigate differences in MRI metrics between progressed and not progressed PP-MS patients. Statistical significance was set at p < 0.05. Given the exploratory nature of this study, adjustment for multiple comparisons was not performed. Standards protocol approvals, registrations, and patient consents Written informed consent was obtained from all participants before the beginning of the study procedures, according to the Declaration of Helsinki. The protocol was approved by the Institutional Review Board of the Icahn School of Medicine at Mount Sinai.
RESULTS
In order to establish a baseline for control and PP-MS for comparison, EDSS, age, lesion volume, and clinical assessments were implemented.
Table 1. Demographic and clinical characteristics of the study groups at baseline
HC (n=20)
PP-MS baseline (n=26)
Females, n
11
14
Age, yrs
51.1 ±9.80
50.9 ±10.3
DD, yrs
–
8.8 ±4.6
EDSS, median score (range)
–
4.0 (1.5-6.0)
z-SDMT
–
-2.10 ±1.44
9HPT, seconds
–
34.09 ±17.9
T25FWT, seconds
–
7.34 ±2.16
VA 1.25%
–
20 ±12
VA 2.5%
–
31 ±11
VA 100%
–
54 ±5
Total T2 lesion load
–
5.56 ±7.43
T2V CC, mL
–
0.12 ±0.17
T2V CST, mL
–
0.60 ±0.84
T2V PTR, mL
–
0.10 ±0.15
Eight out of the 20 patients who repeated MRI at post month 6 exhibited sustained disease progression based on either EDSS (n = 1), 25-FWT (n = 2), 9-HPT (n = 1), new T2 lesions (n=1); on both 25-FWT and 9-HPT worsening (n = 2), or on both EDSS and new T2 lesions (n = 1).No significant group differences were observed when comparing control and patients with PP-MS with regard to age and sex (p = 0.60 and p = 0.50 respectively). Patients’ and control demographic and clinical features are reported in Table 1. At 6-month follow up, patients were once again screened, and determined values reflect changes within the PP-MS sample over time (Table 2).
Table 2. Demographic and clinical characteristics of the study group at 6-month follow up.
PP-MS baseline (n=20)
PP-MS follow-up (n=20)
p -values
Females, n
12
12
–
Age, yrs
50.15 ±11.08
51.15 ± 11.08
–
DD, yrs
9.1 ±4.9
10.1 ±4.9
–
EDSS, median score (range)
4.0 (1.5-6.0)
4.0 (2.0-6.0)
0.22
z-SDMT
-2.17 ±1.49
-2.27 ±1.38
0.08
9HPT, seconds
33.14 ±14.6
34.98 ±24
0.30
T25FWT, seconds
7.13 ±2.03
7.17 ±2.19
0.30
VA 1.25%
19 ±11
22 ±12
0.12
VA 2.5%
31 ±10
30 ±12
0.24
VA 100%
54 ±5
53 ±7
0.74
Total T2V, mL
6.19 ±8.09
6.90 ±8.77
0.11
T2V CC, mL
0.12 ±0.18
0.15 ±0.21
<0.001
T2V CST, mL
0.67 ±0.94
0.85 ±1.24
<0.001
T2V PTR, mL
0.12 ±0.16
0.16 ±0.48
<0.001
No significant group differences were observed when comparing progressed with not progressed patients with regard to age, sex and disease duration (p=0.816, p=0.468 and p=0.877 respectively).
Between group comparison of WTMI metrics at baseline
To characterize the presence and extent of WM abnormalities, DKI-derived WMTI metrics were derived from the CC, CST, and PTR of control and PP-MS patients. Compared to control, PP-MS showed the presence of widespread changes in all analyzed tracts.
Figure 1. DKI-derived WMTI metrics in the corpus callosum, corticospinal tract, and posterior thalamic radiation.
AWF (A), Tortuosity (B), D a,axial (C), D re,radial (D), and D axon ( E) were measured for selected ROI’s: CC, CST, and PTR. Comparison between control (n=20) and PP-MS (n=26) showed significant decrease in AWF, Tortuosity, and D a,axial, f or all analyzed tracts. DKI images were processed using Matlab software (Mathworks). Statistical analysis includes the use of a Shapiro-Wilk test and Mann-Whitney and Fisher exact test (**p<0.01, ***p<0.001).
AWF values significant decreased in the CC (Figure 1a; p<0.001), and T values also decreased in the CST, PTR, and significantly in the CC as well (Figure 1b; p<0.01). D a,axial were also significantly decreased in the CC and CST (Figure 1c; p<0.01). D e,radial showed a widespread increase in CC and CST (Figure 1d; p<0.001, p<0.01), reflecting demyelination within the NAWM of PP-MS patients. Interestingly D axon values showed no significance between groups, suggesting that D axon is limited in its sensitivity to axonal loss and degeneration.
Longitudinal analysis of WMTI metrics
To investigate the sensitivity of WMTI to short-term disease progression in PP-MS DKI-derived WMTI metrics were analyzed between PP-MS patients at baseline and at 6-month follow up.
Figure 2. Longitudinal analysis of DKI metrics of PP-MS patients at baseline and 6-month follow up.
AWF (A), Tortuosity (B), D a,axial (C), D re,radial (D), and D axon ( E) were measured for selected ROI’s: CC, CST, and PTR. Comparison between PP-MS at baseline (n=26) and PP-MS at 6-month follow up (n=26) showed significant decrease in AWF, Tortuosity, and D a,axial, f or all analyzed tracts. DKI images were processed using Matlab software (Mathworks). DISPLAY, Montreal software was used for lesion segmentation. Statistical analysis includes the use of a nonparametric Wilcoxon test and a analysis of variance model (*p<0.05, **p<0.01, **p<0.001)
Over the 6-month period, progressed patients saw a significant decrease in AWF, T, and D e,axial values. AWF showed its most significant decrease in the CC and also within the CST (Figure 2a; p<0.001, p<0.01). T was decreased significantly within the CC of progressed PP-MS patients after follow up (Figure 2b; p<0.01). Decreased AWF and T values within the CC and CST of PP-MS patients suggests widespread axonal loss. The D e,axial values only showed slight significance with decreasing values in the CST (p<0.05), also suggesting the presence of axonal deterioration present in short term disease progression. D e,radial v alues showed a widespread increased in all analyzed tracts, specifically in the CC, suggesting rapid short term demyelination within the NAWM of PP-MS patients (Figure 2d; p<0.001, p<0.01, p<0.05).
WMTI correlation with motor, visual and cognitive disability at baseline
To assess the impact of WM abnormalities on clinical disability, WMTI values were correlated between clinical parameters using a nonparametrics Spearman rank correlation coefficient. AWF, Ƭ, De,axial and De,radial of the CC were significantly correlated with EDSS (rho=-0.456,
p=0.001; rho=-0.470, p=0.001; rho=-0.301, p=0.042; rho=0.321, p=0.030 respectively). With the
Figure 3. ROC curve for baseline AWF and disease progression.
Results of receiver operator characteristic (ROC) curve analysis with disease progression as outcome variable and baseline AWF as predictor (area under the curve= 0.854; 95% confidence interval [CI] = 0.687-1.000, p=0.021, Sensitivity 75%, Specificity 75%).
exception of D axon and D e,axial, all metrics from the CC correlated with cognitive impairment as measured by z-SDMT score (AWF rho=0.457, p=0.025; Ƭ rho=0.486, p=0.016; De,radial rho=-0.446 p=0.029). AWF of the CST was significantly correlated with EDSS (rho= -0.366, p=0.012) and 9-HPT (rho=-0.393, p=0.047). AWF of the PTR was significantly associated with VA for low contrast 100%, 2.50% and
1.25% (rho=0.476, p=0.014; rho=0.463, p=0.017; rho=0.556, p=0.003 respectively). Over 6 months, a significant decrease in AWF and a significant increase in T2 lesion load were detected in the body of CC, PTR and in the CST. No significant difference was detected for any of the other WTMI metrics in NAWM (p= 0.067-0.794) or for FA and MD within macroscopic lesions (p=0.191 and p=0.135, respectively. To assess the validity of WMTI metrics in discriminating between progressed and non-progressed patients, a ROC curve analysis was performed. Baseline AWF values in CST significantly discriminated clinically progressed patients from not-progressed patients (p=0.021, area under the curve (AUC)=0.854, 95% confidence interval [CI] = 95%: 0.687-1.000, Sensitivity 75%, Specificity 75%) (Fig. 3), and at follow-up progressed patients showed lower AWF values in CST than not-progressed patients (p=0.004, 0.360±0.029 vs 0.406±0.032). This suggests that AWF is the most sensitive marker of axonal loss and neurodegeneration in PP-MS, highlighting its future role in clinical applications to observe early stage biomarkers that will aid in proper detection and treatment of PP-MS.
DISCUSSION
In this study, DKI-derived WMTI metrics established the pathological mechanisms responsible for NAWM damage in PP-MS at baseline and at 6 month follow-up and evaluated its impact on different clinical domains.
Since treatments for PP-MS are now available, there is an increasing need to identify those PP-MS patients who will have a more severe outcome over time. A few relatively recent studies attempted to identify measures able to predict the clinical outcome in patients with progressive MS and found that lower baseline EDSS scores and short-term changes are associated with a higher risk of subsequent clinical regression (33–36). Miller et. al., 2018 investigated the relationship between brain volume and clinical degeneration over 3 years in PP-MS and found that the rate of brain volume loss was most rapid in the severe subgroup who experienced clinical progression and least rapid in the stable subgroup who did not experience confirmed progression. Our results were similar in that AWF values in the CST significantly differentiated between progressed and non-progressed patients at 6-month follow up.
In PP-MS, in addition to the well-known role of GM damage, there is extensive histopathological and MRI evidence that NAWM is affected, albeit to a lesser extent, by the same pathological processes that characterize WM lesions, namely inflammation, demyelination, axonal injury, macrophages infiltration and gliosis (37,38). Even if little is known about the pathological relationship linking WM and GM damage in PP-MS, some evidence (39) suggest that WM changes predict subsequent GM abnormalities, rather than the opposite. Furthermore, abnormalities in NAWM rather than in WM lesions have a greater association with later GM damage developed over 2-years follow-up. Against this background, our analysis of WMTI adds to the current knowledge about prognostic factors of disease progression in PP-MS.
DKI is an extension of DTI that incorporates non-Gaussian diffusion effects (40, 41), potentially unveiling intra- and extra-axonal processes and providing higher specificity than DTI to underlying disease mechanisms. In line with this, the between group comparison of WMTI-derived parameters showed the presence of abnormal values for all WMTI-metrics estimated in PP-MS patients NAWM. We found widespread decreased values of AWF, Ƭ, De,axial and an increase values of De,radial. These results are consistent with those of a recent study conducted on RR-MS patients (42). Specifically, the ROI analysis, restricted to the CC, revealed statistically significant differences in all WMTI-metrics except for Daxon. Mean values of AWF, T, and D e,axial, D e,radial were decreased, which reflected chronic axonal degeneration and demyelination. The similarity of findings across studies in RR- and PP-MS patients are in line with the view, largely recognized in recent years, that the primary- progressive phenotype is part of the MS spectrum, showing mainly quantitative rather than qualitative differences with the other phenotypes (2, 36). In this study, we hypothesized that the increased pathological specificity of DKI-derived metrics to the processes that underlie disease burden and disability would be reflected in clinically meaningful correlations (42) . Indeed, we found statistically significant correlations between WMTI metrics and cognitive, motor and visual scores. At baseline, SDMT- a measure of processing speed, attention and working memory- correlated with all WMTI-metrics except for Daxon and De,axial; 9-HPT- a measure that reflects manual dexterity- and EDSS correlated with AWF in CST; visual acuity (1.25%, 2.5%, 100%) correlated with AWF in PTR. The presence of significant correlations between clinical disability and WMTI metrics specific for demyelination and axonal damage supports the concept that the clinical-radiological paradox in PP-MS is due to the lack of pathological specificity of conventional MRI measures.
Our most interesting finding, however, is related to AWF, which seems to capture the ongoing, progressive axonal loss over time as well as its clinical impact. Progression over time was predicted by AWF values and not by lesion load in CST, suggesting a predominant role of NAWM abnormalities over WM macroscopic damage in PP-MS. In this light, AWF seems to be the most sensitive marker of tissue disruption and disability prediction among WMTI. Finally, the significant AWF reduction in all the examined tracts over 6 month confirms that the prevalence of neurodegeneration over demyelination is the main pathological mechanism sustaining PP-MS evolution. Additionally, unlike in RR-MS (43), diffusion parameters within macroscopic lesions did not change significantly over the 6-month follow-up, confirming once more the prominent role of microscopic tissue damage over focal lesions in this clinical phenotype.
CONCLUSION
Our findings support the role of WMTI metrics such as AWF, T, and de,axial as a specific set of WM pathology markers and suggest that these novel metrics may allow for a better characterization of NAWM in PP-MS. WMTI metrics can help distinguish patients with a faster disease progression from those with a stable course, opening a window on the mechanisms underlying PP-MS progression. In line with the results, Bodin et. al., 2016 stated that in PP-MS, the inflammatory disease activity measured as new lesion formation is not the primary mechanism of disability progression in PP-MS. The authors suggest that a notable component of brain volume loss and disease worsening in PP-MS is independent of concurrent inflammatory activity. We can speculate two different types of population among PP-MS patients and, while all PP-MS patients show significant neurodegeneration over time, only subjects showing prevalent neurodegenerative damage at baseline develop significant clinical worsening over the short-term period, suggesting that “neurodegeneration” progresses faster than “demyelination”. Further longitudinal studies on larger samples are warranted to explore the relationship between WTMI metrics and GM pathology to better understand the sequence of events driving clinical progression (44).
It is important to note that the WM model used to derive our WMTI metrics relies on several, albeit common assumptions regarding the WM microstructure. In particular, it assumes that Daxon< De,axial, and that axonal fibers are organized in a relatively parallel fashion along a single direction, which explains our choice of specific WM tracts for the ROIs analysis. The overparameterization of WM models influences the quantitative estimation and biological accuracy of the system negatively (45, 46). Recent research has suggested that Daxon>De,axial, a parameter that is not met using WMTI. WMTI derived metrics are only valid for voxels with axons that are highly aligned in a single bundle, limiting the ROI’s for deriving metrics. Jesperson et al.,has suggested that applying a Watson Distribution of axons, as shown in the WM model Neurite Orientation Dispersion and Density Imaging (NODDI), would circumvent the strongest assumption of WMTI-derived metrics, thus resulting in more accurate metrics and expanding the range for appropriate ROI’s. Future investigations should focus on implementing parameters that assume that utilize Daxon>De,axial rather than Daxon<De,axial and and combining the Watson Model from NODDI with WMTI to provide more accurate and extensive WM derived metrics (46-48). Doing so will provide the most accurate, cost effective, and timely tools to diagnose MS and predict clinical disability over time, allowing flexibility in treatment options and clinical plans for over
2.5 million people world wide. Despite these limitations, this is the first study that, applying DKI in PP-MS patients, has highlighted the sensitivity of such technique in discriminating disease progression by means of a specific MRI marker of axonal damage.
2018-11-20-1542684595