Advertisement

Evaluating Magnetic Resonance Diffusion Properties Together with Brain Volumetry May Predict Progression to Multiple Sclerosis

Open AccessPublished:January 20, 2022DOI:https://doi.org/10.1016/j.acra.2021.12.015

      Rationale and Objectives

      Although the gold standard in predicting future progression from clinically isolated syndrome (CIS) to clinically definite multiple sclerosis (CDMS) consists in the McDonald criteria, efforts are being made to employ various advanced MRI techniques for predicting clinical progression. This study's main aim was to evaluate the predictive power of diffusion tensor imaging (DTI) of the brain and brain volumetry to distinguish between patients having CIS with future progression to CDMS from those without progression during the following 2 years and to compare those parameters with conventional MRI evaluation.

      Materials and Methods

      All participants underwent an MRI scan of the brain. DTI and volumetric data were processed and various parameters were compared between the study groups.

      Results

      We found significant differences between the subgroups of patients differing by future progression to CDMS in most of those DTI and volumetric parameters measured. Fractional anisotropy of water diffusion proved to be the strongest predictor of clinical conversion among all parameters evaluated, demonstrating also higher specificity compared to evaluation of conventional MRI images according to McDonald criteria.

      Conclusion

      Conclusion: Our results provide evidence that the evaluation of DTI parameters together with brain volumetry in patients with early-stage CIS may be useful in predicting conversion to CDMS within the following 2 years of the disease course.

      Key Words

      INTRODUCTION

      Multiple sclerosis (MS) is a chronic inflammatory disease of the central nervous system arising from inflammatory infiltration and causing demyelination of axons and their subsequent destruction in various areas of the central nervous system (
      • Confavreux C
      • Vukusic S
      • Moreau T
      • Adeleine P
      Relapses and progression of disability in multiple sclerosis.
      ). Clinically isolated syndrome (CIS) is a monophasic clinical episode with patient-reported symptoms and objective findings reflecting a focal or multifocal inflammatory demyelinating event in the central nervous system, developing acutely or subacutely and lasting at least 24 hour. There may be one or multiple affected areas (
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ). CIS patients may or may not progress into clinically definite multiple sclerosis (CDMS).
      The gold standard in predicting future progression from CIS to multiple sclerosis consists is the McDonald criteria (
      • Thompson AJ
      • Banwell BL
      • Barkhof F
      • et al.
      Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
      ), which combine clinical findings, laboratory results, and findings on MRI examinations of the brain and spinal cord. The MRI criteria evaluate dissemination in space, which means to assess the number and distribution of hyperintense lesions in the white matter visible in T2-weighted or fluid attenuated inversion recovery (FLAIR) images, and dissemination in time, which means the presence of new lesions or lesions enhancing after gadolinium contrast administration. Due to the evolving capabilities of advanced MRI techniques, an effort is being made to use them to refine the prediction of clinical progression in patients with CIS to definitive multiple sclerosis. The main aim of this prospective single-centre study was to evaluate the predictive power of diffusion MRI of the brain and brain volumetry to distinguish patients with CIS with future progression to CDMS from those without progression and compare those parameters with evaluation of conventional MRI.

      MATERIALS AND METHODS

      The study was approved by the ethics committee of the university hospital, and all subjects signed informed consent forms in order to participate in the study. Two groups of subjects were included in the prospective study. The first group comprised 72 patients with CIS, who were further divided into two subgroups: 50 patients without clinical progression to CDMS and 22 patients with progression during the observed 2-year period since the initial MRI examination. The second group consisted of 62 community-recruited healthy controls. More detailed information about the study participants is shown in (Tables 1 and 2).
      Table 1Characteristics of Patient and Control Groups
      Healthy Volunteers (n = 62)Patients (n = 72)p
      SexWomen45 (72.6%)48 (66.7%)0.573
      Men17 (27.4%)24 (33.3%)
      Age (ys)Mean ± SD33.3 ± 7.434.1 ± 8.40.779
      Median (min–max)32.5 (19.9–51.3)33.1 (19.9–61.3)
      MRI deviceMR147 (75.8%)64 (88.9%)0.065
      MR215 (24.2%)8 (11.1%)
      Demographic characteristics of study participants and number of examinations performed on two different MRI devices (MR1 and MR2). Statistical significance of differences between groups is given by p-values calculated using Fisher's exact test or Mann–Whitney U test, as applicable.
      Table 2Characteristics of Patients Subgroups
      Patients without progression (n = 50)Patients with progression (n = 22)p
      SexWomen32 (64.0%)16 (72.7%)0.591
      Men18 (36.0%)6 (27.3%)
      Age (ys)Mean ± SD33.7 ± 7.834.9 ± 9.90.774
      Median (min–max)32.9 (21.3–50.8)34.4 (19.9–61.3)
      TreatmentYes24 (48.0%)16 (72.7%)0.072
      No26 (52.0%)6 (27.3%)
      EDSSMean ± SD1,7 ± 1,22,2 ± 0,80,063
      Median (min–max)2 (0-5)2 (1-3,5)
      MRI deviceMR148 (96%)16 (72.7%)0.008
      MR22 (4%)6 (27.3%)
      Demographic characteristics of patients classified according to their later clinical progression within 2 years of observation. Numbers of examinations performed on two different MRI devices (MR1 and MR2) are also shown as well as numbers of treated patients within both subgroups. Expanded disability status scale (EDSS) at the first MR examination. Statistical significance of differences between groups is given by p-values calculated using Fisher's exact test or Mann–Whitney U test, as applicable.
      All patients underwent a clinical neurological examination including evaluation of expanded disability status scale (EDSS), laboratory testing of blood and cerebrospinal fluid, and such electrophysiological or other testing as designed to exclude other causes of particular episodes. CIS was diagnosed on the basis of typical clinical findings, such as sensitivity or motor disorders, vertigo, optic neuritis, etc. The progression to CDMS was defined in accordance with the Poser criteria indicating a further clinical demyelinating disease attack (
      • Poser CM
      • Paty DW
      • Scheinberg L
      • et al.
      New diagnostic criteria for multiple sclerosis: guidelines for research protocols.
      ). The patients were clinically monitored at 3-month intervals to register any further clinical attacks. The 2 patient subgroups were subsequently classified according to progression to CDMS emerging during a 2-year period of clinical follow-up after their initial MRI scan. Treatment with first-line disease-modifying drugs (interferon beta or glatiramer acetate) was initiated during the follow-up period in 24 of the non-progressed patients and in 16 of the progressed patients (Table 2). The exclusion criteria for healthy controls, as verified by a questionnaire, included any history of symptoms suggestive of MS. Also excluded were subjects with history or suspicion of meningoencephalitis, stroke, transitory ischemic attacks, epilepsy, and systemic inflammatory diseases, as well as subjects with known blood relatives suffering from MS and subjects with significant load of incidentally found T2 hyperintensities fulfilling the dissemination in space (DIS) criteria according to the 2017 McDonald criteria.
      All participants underwent an MRI scan of the brain using one of two 1.5T MR devices (an older Philips Achieva device or a newer Philips Ingenia device) using a 16-channel head and neck coil. Detailed proportions of examinations performed on each scanner among all study groups are shown in (Tables 1 and 2). Patient examinations were performed within the first couple of months after the first recorded clinical attack (median [minimum–maximum] 2 [0 –8] months). In patients treated with corticosteroids, the examinations were performed after an interval of at least 3 weeks from the last such dose. The protocol comprised sequences for structural imaging and subsequent volumetric analyses (T2, FLAIR 3D, and T1 3D) and diffusion tensor imaging (DTI) sequence. Details about the imaging protocol are shown in (Table 3). DTI was acquired with b factor 0 and 1000s/mm2 using 32 directions of the magnetic gradient. Two experienced radiologists (MK and JS) evaluated by consensus the images for the presence of T2/FLAIR hyperintense lesions and classified the findings in terms of the DIS in accordance with the 2017 McDonald criteria. Final decisions in doubtful cases were made by group consensus while including the other co-authors (AŠ and MM).
      Table 3Parameters of Magnetic Resonance Imaging Protocol
      SequenceOrientationTR (ms)TE (ms)Acquisition Voxel Size (mm)
      T2 TSEtransverse48511100.9 × 1.12 × 5
      FLAIR 3Dsagittal80002751.2 × 1.2 × 1.4
      T1 3D FFEtransverse254.10.9 × 0.9 × 1.6
      DTItransverse21000622 × 2 × 2
      Parameters of magnetic resonance imaging protocol. TSE, turbo spin echo; TR, repetition time; TE, echo time; FLAIR, fluid attenuation inversion recovery; FFE, fast field echo; DTI, diffusion tensor imaging.
      DTI data processing was done using FMRIB's Software Library (FSL) (
      • Jenkinson M
      • Beckmann CF
      • Behrens TE
      • Woolrich .W
      • Smith SM
      Fsl.
      ), starting with brain extraction, eddy current, and movement correction and calculation of maps of the scalar diffusion parameters (fractional anisotropy [FA], mean diffusivity [MD], axial diffusivity [AD], and radial diffusivity [RD]). The data were subsequently analyzed by voxel-wise tract-based spatial statistics (TBSS) (
      • Smith SM
      • Jenkinson M
      • Johansen-Berg H
      • et al.
      Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data.
      ) and its statistical module (Randomize) to compare the diffusion parameters between the study groups with multiple comparisons correction and subjects’ age, sex, and MR device set as covariates. Mean values of the aforementioned scalar parameters were extracted from the sum of white matter voxels differing significantly (p < 0.05) between those groups being compared, and those values were then entered into further statistical analyses.
      The total brain volume and volumes of white matter (WM) and grey matter (GM), normalized for subject head size, were estimated with SienaX (
      • Smith SM
      • Rao A
      • De Stefano N
      • et al.
      Longitudinal and cross-sectional analysis of atrophy in Alzheimer's disease: cross-validation of BSI, SIENA and SIENAX.
      ), which is part of FSL (
      • Smith SM
      • Jenkinson M
      • Woolrich MW
      • et al.
      Advances in functional and structural MR image analysis and implementation as FSL.
      ), using T1-weighted 3D images. Normalization was done by SienaX scaling factor. Pre-processing included also automatic lesion segmentation and lesion filling procedure using the lesion growth algorithm (
      • Battaglini M
      • Jenkinson M
      • De Stefano N
      Evaluating and reducing the impact of white matter lesions on brain volume measurements.
      ,
      • Schmidt P
      • Gaser C
      • Arsic M
      • et al.
      An automated tool for detection of FLAIR-hyperintense white-matter lesions in multiple sclerosis.
      ) as implemented in the LST toolbox version 3.0.0 (www.statistical-modelling.de/lst.html) for statistical parametric mapping.
      For overall evaluation of brain abnormalities in patients with CIS, all DTI and volumetric parameters were compared between the whole group of patients and healthy controls. To evaluate the predictive value of those parameters, we mutually compared the data of individual subgroups of patients defined according to progression to CDMS using Mann–Whitney U test. The subjects’ age was described as mean with SD and/or median with minimum–maximum values and was compared between the groups using Mann–Whitney U test. The subjects’ categorical characteristics (such as sex, treatment, or scanner type) were tested between groups using Fisher's exact test. Effects of potential confounding factors (age and sex of the subjects and MRI scanner type) were eliminated from the data prior to statistical testing using linear regression in order to avoid biased results. Similarly, these parameters were set as covariates for voxel-based analysis of diffusion parameters using Randomize. Furthermore, a receiver operating characteristic (ROC) analysis was conducted in order to evaluate the sensitivity and specificity of significantly differing diffusion and volumetric parameters between the patient subgroups.
      To evaluate the possible influence of the two different MRI scanners on the power of DTI and volumetric parameters to predict clinical progression, we included also separate analysis of those modalities evaluating only the group of 64 patients (41 women, 23 men, mean age 34.7 years), who were examined solely on the MR1 device (more details about the group are shown in Table 2S in the supplementary material).
      In addition, sensitivity and specificity of McDonald DIS criteria were calculated on the basis of predicted progression to CDMS. The significance level for all statistical tests was set at p < 0.05. Statistical analyses were performed using IBM SPSS Statistics 25, R 3.4.1, and Statistica 12 (StatSoft).

      RESULTS

      There were no statistically significant differences in the age or sex of the subjects between patients and controls or between the patient subgroups. Employment of the two MRI scanners yielded significantly different outcomes between the two patient subgroups (Tables 1 and 2). The differences in number of treated patients were not statistically significant between the subgroups defined by progression to CDMS (Table 2). The median (min–max) interval between the MRI scan and clinical progression to CDMS in the subgroup of 22 CIS patients who developed CDMS was 11.5 (0 –19) months.
      TBSS analysis of the diffusion data identified extensive areas within the brain WM differing significantly in FA, MD, and RD between the subgroups of CIS patients as defined by their later progression to CDMS (Fig 1). Several areas within the brain also differed significantly in FA, MD, and AD values when all patients were compared with controls. The differences in FA and MD were apparently more widespread across the brain tissues in patients with progression compared to the progression-free subgroup than in all patients compared to healthy controls (Fig 2). The number of voxels (percentage of all significant voxels) with significantly different FA and MD values overlapping between the two TBSS analyses came to 9872 (19.63%) and 5202 (9.59%) voxels, respectively. The median values of the diffusion parameters extracted from the significant voxels (based on TBSS analyses) differed significantly between the groups also according to Mann–Whitney U test (Table 4). ROC analysis of the diffusion parameters in terms of predicting progression from CIS to CDMS revealed FA as the strongest predictor (sensitivity 77.3%, specificity 90%) against MD (sensitivity 63.6%, specificity 78%) and RD (sensitivity 63.6%, specificity 86%) (Table 5, Fig 3).
      Figure 1
      Figure 1Graphic representation of the results of tract-based spatial statistics (TBSS) analyses. The selected brain images show red–yellow areas where significant differences were found between subgroups of CIS patients with and without progression to CDMS in the parameters of fractional anisotropy (FA), mean diffusivity (MD), and radial diffusivity (RA). The skeleton of white matter tracts used in data processing is marked in green.(Color version of figure is available online).
      Figure 2
      Figure 2Graphic presentation of differences in distribution of diffusivity changes between the two tract-based spatial statistics (TBSS) analyses comparing patients and heathy controls and both subgroups of patients. Distribution of voxels with significantly different fractional anisotropy (FA) (a–c) and mean diffusivity (MD) (d–f) values between the groups compared in axial (a, d), coronal (b, e), and sagittal (c, f) projection. Blue colour represents voxels with significant differences between all patients and healthy controls. Green colour represents voxels with significant differences between the subgroups of patients defined by their later clinical progression, and red colour marks the significant voxels common for both analyses.(Color version of figure is available online).
      Table 4Diffusion Tensor Imaging Parameters in Patients and Healthy Controls
      ParameterPatients Median (MM)Controls Median (MM)p VoxelsPatients with Progression Median (MM)Patients without Progression Median(MM)p Voxels
      FA0.532 (0.398–0.574)0.545 (0.472–0.59)<0.001 193030.394 (0.343–0.429)0.426 (0.391–0.455)<0.001 50278
      MD [10−6]730 (562–800)736 (699–773)<0.001 15271835 (702–969)811 (647–864)<0.001 54227
      AD [10−6]1168 (881–1235)1175 (1117–1234)<0.001 49301---
      RD [10−6]---659 (587–780)624 (515–683)<0.001 75755
      FA, fractional anisotropy; MD, mean diffusivity; AD, axial diffusivity; RD, radial diffusivity; MM, minimum–maximum. p-values representing statistical significance of differences between patients and controls (p†) and between the patient subgroups (p‡) were calculated using Mann–Whitney U test with correction for age, sex,and scanner. Number of voxels identified as significantly (p<0.05) different between patients and controls and between both subgroups of patients as revealed by tract-based spatial statistics (TBSS) (voxels†and voxels‡respectively).ROIs used for the two analyses († and ‡) only partly overlap;therefore, the values entering those analyses are not fully comparable.
      Table 5Receiver Operating Characteristic (ROC) Analysis of Significant Diffusion Tensor Imaging and Volumetric Parameters
      ParameterROC area (CI)pCut-OffSensitivitySpecificity
      FA0.890 (0.806–0.974)<0.0010.40869877.390.0
      MD0.707 (0.562–0.852)<0.0050.00082663.678.0
      RD0.750 (0.615–0.885)<0.0010.00064563.686.0
      White matter0.753 (0.643 – 0.863)<0.001725609.490.958.0
      Whole brain0.628 (0.488 – 0.768)0.0851480222.850.070.0
      FA, fractional anisotropy; MD, mean diffusivity; RD, radial diffusivity.
      Figure 3
      Figure 3Receiver operating characteristic (ROC) curve of the fractional anisotropy (FA) parameter as a predictor of progression to clinically definite multiple sclerosis in patients with clinically isolated syndrome.
      The analyses of the DTI data of the group of 64 patients excluding those examined on the MR2 device revealed similar results in terms of prediction of the clinical conversion (FA: sensitivity 68.8% and specificity 93.7%, MD: sensitivity 68.8% and specificity 87.5%, RD: sensitivity 68.8% and specificity 85.4%). More detailed results are shown in tables 4S and 5S in the supplementary material.
      By conventional evaluation of T2-w and FLAIR images, we identified 18 (81.8%) subjects among the progressed patients who met DIS criteria, while in the group of non-progressed patients DIS criteria were met in 26 (52%) subjects. Thus, the sensitivity of DIS in terms of predicting clinical progression was 81.8% and the specificity 48%.
      All measured volumetric parameters (whole brain, WM, and GM) differed significantly between patients and controls, revealing generally lower volumes in patients (Table 6). Similarly, the volume of brain WM and whole brain volume were significantly reduced in patients with progression compared to non-progressed patients, but the volume of GM did not differ significantly between these subgroups (Table 6). In the subsequent ROC analysis, the volume of WM was able to predict the conversion of CIS to CDMS with sensitivity 90.9% and specificity 58.0% (Table 5). Very similar results were obtained when only the 64 patients examined on MR1 were included in the analysis of the WM volume parameter to differentiate between CIS and CDMS subgroups (revealing sensitivity 87.5% and specificity 56.2%). More detailed results are shown in tables 5S and 6S in the supplementary material.
      Table 6Volumetric Parameters in Patients and Healthy Controls
      VolumePatients Median (MM)Controls Median (MM)p (value)Patients with Progression Median (MM)Patients without Progression Median (MM)p (value)
      GM [cm3]787 (674.2–900.6)812.5 (686.1–939.3)<0.001780 (705.4–897.6)787 (674.2–900.6)0.932
      WM [cm3]717.5 (644.5–799.8)731.4 (624.3–810.6)<0.05694.2 (644.5–749.2)735 (655.4–799.8)<0.001
      WB [cm3]1511.9 (1348–1627)1537.6 (1379–1722)<0.0011491.3 (1365–1623)1515.1 (1348–1627)<0.05
      GM, grey matte; WM, white matter; WB, whole brain; MM, minimum–maximum; p-values representing statistical significance of differences between patients and controls (p†) and between patient subgroups (p‡) were calculated using Mann–Whitney U test with correction for age, sex,and scanner.

      DISCUSSION

      This study's aim was to explore the potential of diffusion scalar parameters and brain volumetry analysis to predict the clinical conversion to CDMS in patients with early-stage CIS. From a clinical point of view, identifying patients with CIS having high-risk of conversion to CDMS is a matter of great importance. In recent years, a number of other groups of researchers have endeavoured to introduce methods for predicting the progression of CIS to CDMS based on clinical, electrophysiological, laboratory, or imaging findings (
      • Cinar BP
      • Özakbaş S
      Prediction of conversion from clinically isolated syndrome to multiple sclerosis according to baseline characteristics: a prospective study.
      ,
      • Kolčava J
      • Kočica J
      • Hulová M
      • et al.
      Conversion of clinically isolated syndrome to multiple sclerosis: a prospective study.
      ).
      If we focus on those studies using diagnostic imaging methods, efforts are being made either to find new possibilities for evaluating and quantifying findings on conventional MRI sequences or to find new advanced methods of MRI and possibly introduce them into routine clinical practice. An example of a study employing advanced techniques of structural MRI data analysis is that of Bendfeldt et al. (
      • Bendfeldt K
      • Taschler B
      • Gaetano L
      • et al.
      MRI-based prediction of conversion from clinically isolated syndrome to clinically definite multiple sclerosis using SVM and lesion geometry.
      ), which evaluates combination of clinical and demographic data with evaluation of image-based lesion-specific geometry and brain volume. The highest prediction accuracy of 70.4% was achieved by a combination of lesion-specific geometric (image-based) and demographic and/or clinical features (
      • Bendfeldt K
      • Taschler B
      • Gaetano L
      • et al.
      MRI-based prediction of conversion from clinically isolated syndrome to clinically definite multiple sclerosis using SVM and lesion geometry.
      ). Similarly, Wotschell et al. (
      • Wottschel V.
      • Alexander D.C.
      • Kwok P.P.
      • et al.
      Predicting outcome in clinically isolated syndrome using machine learning.
      ) report the potential of machine-learning algorithms for predicting conversion to CDMS by analysis of conventional proton density and T2-weighted images, revealing sensitivity of 77% and specificity of 66% during 1 year of observation.
      We have chosen TBSS, a module of FSL, for the analysis of DTI data. This tool represents a widely accepted approach to diffusion data analysis using nonlinear registration of FA maps, reconstruction of main WM tracts, and projection of FA values of individual subjects onto this skeleton (
      • Smith SM
      • Jenkinson M
      • Johansen-Berg H
      • et al.
      Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data.
      ). This fully automatic method is virtually operator independent and also time efficient, which are the major benefits over manual techniques based upon region-of-interest that may be prone to subjective error and may provide poorer reproducibility compared to TBSS (
      • Lilja Y
      • Gustafsson O
      • Ljungberg M
      • Nilsson D
      • Starck G
      Impact of region-of-interest method on quantitative analysis of DTI data in the optic tracts.
      ).
      Our results generally correspond with previous findings of changes in DTI parameters of brain WM in patients with multiple sclerosis (

      Metcalf M. Advanced MR technique development for improved characterization of multiple sclerosis (Doctoral dissertation, UCSF). 2008.

      ) or CIS (
      • Cappellani R
      • Bergsland N
      • Weinstock-Guttman B
      • et al.
      Diffusion tensor MRI alterations of subcortical deep gray matter in clinically isolated syndrome.
      ) in comparison to healthy controls. In the normal, healthy population, brain WM, unlike the GM, has a comparatively high FA value and low MD values (
      • Kodiweera C
      • Alexander AL
      • Harezlak J
      • McAllister TW
      • Wu YC
      Age effects and sex differences in human brain white matter of young to middle-aged adults: A DTI, NODDI, and q-space study.
      ). As the brain ages, gradual physiological degeneration has been observed, and thus there is gradual decrease in FA and increase in MD in the WM (
      • Draganski B
      • Ashburner J
      • Hutton C
      • et al.
      Regional specificity of MRI contrast parameter changes in normal ageing revealed by voxel-based quantification (VBQ).
      ). Similarly, in our study, mean values of MD were higher and values of FA lower in patients with progression compared to those without progression, which, from the perspective of diffusion properties of the brain WM, may resemble an accelerated process of aging. In contrast, we observed moderately lower MD values in some WM areas (mostly parietal and frontal lobe WM, Fig 2d–f) within the group of all patients compared to the control group. It is necessary to realize that the reported values of scalar parameters are not fully comparable between the two analyses, because they were measured within different areas of WM as given by TBSS analysis. Physiological structural heterogeneity probably plays some role here. Furthermore, according to previous studies, early demyelinating changes may demonstrate some degree of restricted diffusion (
      • Eisele P
      • Szabo K
      • Griebe M
      • et al.
      Reduced diffusion in a subset of acute MS lesions: a serial multiparametric MRI study.
      ) in contrast to chronic changes in CDMS patients, within whom higher MD values were found (
      • Sbardella E
      • Tona F
      • Petsas N
      • Pantano P
      DTI measurements in multiple sclerosis: evaluation of brain damage and clinical implications.
      ). Inasmuch as the MRI examinations were performed quite early after the initial clinical attack (approximately 42% of patients examined within the first month, median 2 months), we may speculate that lower MD values in the whole group of early-stage CIS patients compared to controls may generally be related to early demyelination, but comparatively higher MD values in selected patients with further clinical progression compared to non-progressed patients may be due to underlying (possibly subclinical) chronic ultrastructural abnormalities that are associated with risk of future clinical progression.
      The differences in FA and MD were generally more widespread across the brain (eg, cerebellum) when the subgroups of patients defined by later progression to CDMS were mutually compared than in the case of comparing patients to healthy controls (Fig 2). Moreover, patients differed from the control group in AD, and, conversely, we proved significant differences in RD in patients with progression compared to those without progression. Those two parameters may be attributed to different ultrastructural abnormalities of WM, where RD is recognized as a marker of demyelination while changes in AD may more likely reflect axonal disintegration (
      • Song SK
      • Sun SW
      • Ju WK
      • Lin SJ
      • Cross AH
      • Neufeld AH
      Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia.
      ). From the perspective of diffusion properties, therefore, the changes in patients with later clinical progression appear moderately specific and show different characteristics compared to general abnormalities in CIS patients found in comparison to healthy subjects.
      The FA scalar parameter quantifying anisotropy of diffusion within brain tissue may be perceived as a general marker of nerve fibre integrity disruption (
      • Seewann A
      • Vrenken H
      • van der Valk P
      • et al.
      Diffusely abnormal white matter in chronic multiple sclerosis: imaging and histopathologic analysis.
      ). In our study, FA of brain WM appeared as the strongest predictor of conversion to CDMS, with sensitivity of 77.3% and specificity of 90%. There is not much data in the literature about the potential of MRI diffusion techniques for predicting CIS to CDMS conversion. One previous study from Gallo et al. (
      • Gallo A
      • Rovaris M
      • Riva R
      • et al.
      Diffusion-tensor magnetic resonance imaging detects normal-appearing white matter damage unrelated to short-term disease activity in patients at the earliest clinical stage of multiple sclerosis.
      ) found significant DTI abnormalities within normal appearing white matter (NAWM) in the brains of patients with CIS, but it did not find significant differences between a subgroup of patients with CIS initially fulfilling McDonald criteria for dissemination in space with later progression to CDMS and a subgroup without progression. Conversely, Kugler et al. (
      • Kugler AV
      • Deppe M
      Non-lesional cerebellar damage in patients with clinically isolated syndrome: DTI measures predict early conversion into clinically definite multiple sclerosis.
      ), in their study (and similarly to our study), proved alterations of FA in cerebellar tissues as a predictor of conversion to CDMS. Moreover, histogram analysis of cervical cord's diffusion parameters also has been used in a recent study to predict CIS to CDMS progression with sensitivity and specificity of FA kurtosis of both WM and NAWM of 93% and 72%, respectively (
      • Dostál M
      • Keřkovský M
      • Stulík J
      • et al.
      MR diffusion properties of cervical spinal cord as a predictor of progression to multiple sclerosis in patients with clinically isolated syndrome.
      ). Conventional evaluation of structural MRI data according to McDonald DIS criteria revealed sensitivity of 81.8% in terms of prediction to CDMS. That was comparable to the predictive power of FA, but the specificity of the conventional criteria and evaluation was substantially lower, at 48%. From this perspective, it may appear that the analysis of diffusion data is more accurate than is conventional MRI.
      Several studies have been published confirming brain volume reduction in patients with CDMS. Moreover, the rate of volume reduction correlates with progression of the disease's clinical symptoms (
      • Giorgio A
      • De Stefano N
      Clinical use of brain volumetry.
      ). Some studies have also evaluated cerebral atrophy in CIS patients (
      • Rojas JI
      • Patrucco L
      • Besada C
      • Bengolea L
      • Cristiano E
      Brain atrophy in clinically isolated syndrome.
      ) or even considered using brain volume measurements as a predictor of progression from CIS to CDMS while taking into consideration separately GM and WM volumes (
      • Giorgio A
      • De Stefano N
      Clinical use of brain volumetry.
      ). One of the studies of a nature similar to that of ours is the study of Dalton et al. (
      • Dalton CM
      • Chard DT
      • Davies GR
      • et al.
      Early development of multiple sclerosis is associated with progressive grey matter atrophy in patients presenting with clinically isolated syndromes.
      ). In this work, the only statistically significant predictor of clinical progression was reduction in the volume of GM. The volume of WM was not significantly different in the two investigated groups. These findings contradict the results of our study, which evidence statistically significant reduction of whole brain volume and of WM to be most pronounced in patients with progression to CDMS within the next 2 years. The reason for these discrepancies may relate to a smaller sample of patients in the case of the first study and especially differences in methodology, as the cited older study uses not a T1 3D sequence for segmentation but only a 2D T2 sequence. In such case, a poorer WM / GM contrast can be expected and the resulting weaker spatial resolution may provide less precise volumetric data. We believe that our results are logical, given that demyelination generally affects WM more than GM (
      • Lassmann H
      Multiple sclerosis pathology.
      ) and WM atrophy correlates with the clinical state of the patients (
      • Sbardella E
      • Petsas N
      • Tona F
      • et al.
      Assessing the correlation between grey and white matter damage with motor and cognitive impairment in multiple sclerosis patients.
      ). WM volume demonstrated comparatively low specificity (58.0%) with respect to predicting clinical conversion in patients with CIS, but the sensitivity was comparatively higher (90.9%) and the discrimination power of volumetry was generally weaker compared to those of DTI parameters.
      This study had several limitations. Some may consider as a limitation the use of a 1.5T MRI device that provides images with generally lower signal-to-noise ratio compared to 3T systems. On the other hand, this shortcoming is partially offset by the fact that a lower magnetic field, by its physical nature, produces smaller numbers of susceptibility artefacts compared to machines with higher induction that may become important especially in anatomical areas near the skull base. Furthermore, one of the recent multicentre studies has shown that most of the diffusion MRI-derived parameters are robust even across 1.5T and 3T scanners (
      • Grech-Sollars M
      • Hales PW
      • Miyazaki K
      • et al.
      Multi-centre reproducibility of diffusion MRI parameters for clinical sequences in the brain.
      ).
      Another limitation is the use of two different MRI devices, as the hardware had been replaced during the study. Although both scanners were 1.5T devices from the same manufacturer and the examinations on the newer MRI device were performed using exactly the same acquisition parameter settings, the influence of different MRI hardware, especially on the diffusion scalar parameters, may be significant (
      • Landman BA
      • Farrell JA
      • Jones CK
      • Smith SA
      • Prince JL
      • Mori S
      Effects of diffusion weighting schemes on the reproducibility of DTI-derived fractional anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5 T.
      ). Because the proportions of examinations performed on the two MRI devices were not equal among the study groups, the MRI device was included as a covariate into all statistical analyses to correct for possible influence of this factor; such approach has already been reported in the literature (
      • Takao H
      • Hayashi N
      • Ohtomo K
      Sex dimorphism in the white matter: fractional anisotropy and brain size.
      ). Moreover, to further validate the results, we provide also the key analysis comparing progressed and non-progressed patients restricted merely to the group of 64 patients examined on a single MRI device. Inasmuch as these data do not differ substantially from the whole-group analysis, we believe that the influence of the different MRI hardware is not crucial. In any case, the reproducibility among different MRI devices with different field strengths should be studied in relation to this topic before the techniques investigated here can be used in routine diagnostics.
      Another limitation is the relatively small size of the study group due to recruitment of patients from only one multiple sclerosis centre. Although the number of patients is sufficient to provide statistically significant results, it would be appropriate to verify the results on a larger number of patients. In addition, the 2-year follow-up period is a relatively short time frame within which to evaluate the potential for CDMS conversion. The time to conversion to CDMS in CIS patients reported in the literature is somewhat variable. In a large study with more than 1,000 patients, for example, the median time to conversion was 1,096 days (
      • Kuhle J
      • Disanto G
      • Dobson R
      • et al.
      Conversion from clinically isolated syndrome to multiple sclerosis: a large multicentre study.
      ). Another study investigating a smaller group of patients indicated mean time to conversion of 11 months (
      • Cinar BP
      • Özakbaş S
      Prediction of conversion from clinically isolated syndrome to multiple sclerosis according to baseline characteristics: a prospective study.
      ). With this in mind, a 2-year monitoring period was considered acceptable, especially given that relapsing activity during the initial 1 –2 years of the disease's course in patients with MS is of crucial prognostic importance in anticipating the severity level of future damage (
      • Weinshenker BG
      • Bass B
      • Rice GPA
      • et al.
      The natural history of multiple sclerosis: a geographically based study. 2. Predictive value of the early clinical course.
      ). It is important to note, however, that the conversion rate in CIS patients in the next 20 years reported in long-term studies reaches up to 50% –60% (
      • Fisniku LK
      • Brex PA
      • Altmann DR
      • et al.
      Disability and T2 MRI lesions: a 20-year follow-up of patients with relapse onset of multiple sclerosis.
      ,
      • Miller D
      • Barkhof F
      • Montalban X
      • Thompson A
      • Filippi M
      Clinically isolated syndromes suggestive of multiple sclerosis, part 2: non-conventional MRI, recovery processes, and management.
      ). Considering the lower conversion rate established in our study (30.6%), we may expect that some of the patients who remained clinically stable for 2 years may develop further clinical attack in future. Thus, the predictive power and longitudinal evolution of DTI and volumetric parameters need to be further investigated by long-term studies.
      A certain bias may also have arisen from the effects of treatments initiated during the follow-up period in patients belonging to both subgroups. Although the numbers of treated patients were not significantly different between the patient subgroups as defined by their progression to CDMS, this fact could to some extent influence the measured diffusion parameters. Nevertheless, any intentional observation of this disease's natural progression, while a theoretically optimal methodology, would be wholly unacceptable from an ethical standpoint.

      CONCLUSION

      This study provides evidence that the evaluation of DTI parameters together with brain volumetry in patients with early-stage CIS may be useful in predicting CIS conversion to CDMS within the following 2 years of the disease.

      ACKNOWLEDGEMENTS

      Contract grant sponsor: Czech Health Research Council; contract number: AZV-15 32133A + Supported by Ministry of Health, Czech Republic - conceptual development of research organization (FNBr, 65269705)

      Appendix. Supplementary materials

      REFERENCE

        • Confavreux C
        • Vukusic S
        • Moreau T
        • Adeleine P
        Relapses and progression of disability in multiple sclerosis.
        N Engl J Med. 2000; 343: 1430-1438
        • Thompson AJ
        • Banwell BL
        • Barkhof F
        • et al.
        Diagnosis of multiple sclerosis: 2017 revisions of the McDonald criteria.
        Lancet Neurol. 2018; 17: 162-173
        • Poser CM
        • Paty DW
        • Scheinberg L
        • et al.
        New diagnostic criteria for multiple sclerosis: guidelines for research protocols.
        Ann Neurol. 1983; 13: 227-231
        • Jenkinson M
        • Beckmann CF
        • Behrens TE
        • Woolrich .W
        • Smith SM
        Fsl.
        Neuroimage. 2012; 62: 782-790
        • Battaglini M
        • Jenkinson M
        • De Stefano N
        Evaluating and reducing the impact of white matter lesions on brain volume measurements.
        Hum brain mapp. 2012; 33: 2062-2071
        • Smith SM
        • Rao A
        • De Stefano N
        • et al.
        Longitudinal and cross-sectional analysis of atrophy in Alzheimer's disease: cross-validation of BSI, SIENA and SIENAX.
        Neuroimage. 2007; 36: 1200-1206
        • Smith SM
        • Jenkinson M
        • Woolrich MW
        • et al.
        Advances in functional and structural MR image analysis and implementation as FSL.
        Neuroimage. 2004; 23: S208-S219
        • Schmidt P
        • Gaser C
        • Arsic M
        • et al.
        An automated tool for detection of FLAIR-hyperintense white-matter lesions in multiple sclerosis.
        Neuroimage. 2012; 59: 3774-3783
        • Cinar BP
        • Özakbaş S
        Prediction of conversion from clinically isolated syndrome to multiple sclerosis according to baseline characteristics: a prospective study.
        Arch Neuropsychiatry. 2018; 55: 15
        • Kolčava J
        • Kočica J
        • Hulová M
        • et al.
        Conversion of clinically isolated syndrome to multiple sclerosis: a prospective study.
        Mult Scler Relat Disord. 2020; 44102262
        • Bendfeldt K
        • Taschler B
        • Gaetano L
        • et al.
        MRI-based prediction of conversion from clinically isolated syndrome to clinically definite multiple sclerosis using SVM and lesion geometry.
        Brain Imaging Behav. 2019; 13: 1361-1374
        • Wottschel V.
        • Alexander D.C.
        • Kwok P.P.
        • et al.
        Predicting outcome in clinically isolated syndrome using machine learning.
        NeuroImage Clin. 2015; 7: 281-287
        • Smith SM
        • Jenkinson M
        • Johansen-Berg H
        • et al.
        Tract-based spatial statistics: voxelwise analysis of multi-subject diffusion data.
        Neuroimage. 2006; 31: 1487-1505
        • Lilja Y
        • Gustafsson O
        • Ljungberg M
        • Nilsson D
        • Starck G
        Impact of region-of-interest method on quantitative analysis of DTI data in the optic tracts.
        BMC Med Imaging. 2016; 16: 1-10
      1. Metcalf M. Advanced MR technique development for improved characterization of multiple sclerosis (Doctoral dissertation, UCSF). 2008.

        • Cappellani R
        • Bergsland N
        • Weinstock-Guttman B
        • et al.
        Diffusion tensor MRI alterations of subcortical deep gray matter in clinically isolated syndrome.
        J Neurol Scienc. 2014; 338: 128-134
        • Kodiweera C
        • Alexander AL
        • Harezlak J
        • McAllister TW
        • Wu YC
        Age effects and sex differences in human brain white matter of young to middle-aged adults: A DTI, NODDI, and q-space study.
        Neuroimage. 2016; 128: 180-192
        • Draganski B
        • Ashburner J
        • Hutton C
        • et al.
        Regional specificity of MRI contrast parameter changes in normal ageing revealed by voxel-based quantification (VBQ).
        Neuroimage. 2011; 55: 1423-1434
        • Eisele P
        • Szabo K
        • Griebe M
        • et al.
        Reduced diffusion in a subset of acute MS lesions: a serial multiparametric MRI study.
        Am J Neuroradiol. 2012; 33: 1369-1373
        • Sbardella E
        • Tona F
        • Petsas N
        • Pantano P
        DTI measurements in multiple sclerosis: evaluation of brain damage and clinical implications.
        Mult Scler Int. 2013; 2013671730
        • Song SK
        • Sun SW
        • Ju WK
        • Lin SJ
        • Cross AH
        • Neufeld AH
        Diffusion tensor imaging detects and differentiates axon and myelin degeneration in mouse optic nerve after retinal ischemia.
        Neuroimage. 2003; 20: 1714-1722
        • Seewann A
        • Vrenken H
        • van der Valk P
        • et al.
        Diffusely abnormal white matter in chronic multiple sclerosis: imaging and histopathologic analysis.
        Arch Neurol. 2009; 66: 601-609
        • Gallo A
        • Rovaris M
        • Riva R
        • et al.
        Diffusion-tensor magnetic resonance imaging detects normal-appearing white matter damage unrelated to short-term disease activity in patients at the earliest clinical stage of multiple sclerosis.
        Arch Neurol. 2005; 62: 803-808
        • Kugler AV
        • Deppe M
        Non-lesional cerebellar damage in patients with clinically isolated syndrome: DTI measures predict early conversion into clinically definite multiple sclerosis.
        NeuroImage Clin. 2018; 19: 633-639
        • Dostál M
        • Keřkovský M
        • Stulík J
        • et al.
        MR diffusion properties of cervical spinal cord as a predictor of progression to multiple sclerosis in patients with clinically isolated syndrome.
        J Neuroimaging. 2021; 31: 108-114
        • Giorgio A
        • De Stefano N
        Clinical use of brain volumetry.
        J Magn Reson Imaging. 2013; 37: 1-14
        • Rojas JI
        • Patrucco L
        • Besada C
        • Bengolea L
        • Cristiano E
        Brain atrophy in clinically isolated syndrome.
        Neurología (English Edition). 2010; 25: 430-434
        • Dalton CM
        • Chard DT
        • Davies GR
        • et al.
        Early development of multiple sclerosis is associated with progressive grey matter atrophy in patients presenting with clinically isolated syndromes.
        Brain. 2004; 127: 1101-1107
        • Lassmann H
        Multiple sclerosis pathology.
        Cold Spring Harb Perspect Med. 2018; 8a028936
        • Sbardella E
        • Petsas N
        • Tona F
        • et al.
        Assessing the correlation between grey and white matter damage with motor and cognitive impairment in multiple sclerosis patients.
        PloS One. 2013; 8: e63250
        • Grech-Sollars M
        • Hales PW
        • Miyazaki K
        • et al.
        Multi-centre reproducibility of diffusion MRI parameters for clinical sequences in the brain.
        NMR Biomed. 2015; 28: 468-485
        • Landman BA
        • Farrell JA
        • Jones CK
        • Smith SA
        • Prince JL
        • Mori S
        Effects of diffusion weighting schemes on the reproducibility of DTI-derived fractional anisotropy, mean diffusivity, and principal eigenvector measurements at 1.5 T.
        Neuroimage. 2007; 36: 1123-1138
        • Takao H
        • Hayashi N
        • Ohtomo K
        Sex dimorphism in the white matter: fractional anisotropy and brain size.
        J Magn Reson Imaging. 2014; 39: 917-923
        • Kuhle J
        • Disanto G
        • Dobson R
        • et al.
        Conversion from clinically isolated syndrome to multiple sclerosis: a large multicentre study.
        Mult Scler J. 2015; 21: 1013-1024
        • Weinshenker BG
        • Bass B
        • Rice GPA
        • et al.
        The natural history of multiple sclerosis: a geographically based study. 2. Predictive value of the early clinical course.
        Brain. 1989; 112: 1419-1428
        • Fisniku LK
        • Brex PA
        • Altmann DR
        • et al.
        Disability and T2 MRI lesions: a 20-year follow-up of patients with relapse onset of multiple sclerosis.
        Brain. 2008; 131: 808-817
        • Miller D
        • Barkhof F
        • Montalban X
        • Thompson A
        • Filippi M
        Clinically isolated syndromes suggestive of multiple sclerosis, part 2: non-conventional MRI, recovery processes, and management.
        Lancet Neurol. 2005; 4: 341-348