Evaluation of a connectivity-based imaging metric that reflects functional decline in Multiple Sclerosis.

Cognitive impairment is a common symptom in individuals with Multiple Sclerosis (MS), but meaningful, reliable biomarkers relating to cognitive decline have been elusive, making evaluation of the impact of therapeutics on cognitive function difficult. Here, we combine pathway-based MRI measures of structural and functional connectivity to construct a metric of functional decline in MS. The Structural and Functional Connectivity Index (SFCI) is proposed as a simple, z-scored metric of structural and functional connectivity, where changes in the metric have a simple statistical interpretation and may be suitable for use in clinical trials. Using data collected at six time points from a 2-year longitudinal study of 20 participants with MS and 9 age- and sex-matched healthy controls, we probe two common symptomatic domains, motor and cognitive function, by measuring structural and functional connectivity in the transcallosal motor pathway and posterior cingulum bundle. The SFCI is significantly lower in participants with MS compared to controls (p = 0.009) and shows a significant decrease over time in MS (p = 0.012). The change in SFCI over two years performed favorably compared to measures of brain parenchymal fraction and lesion volume, relating to follow-up measures of processing speed (r = 0.60, p = 0.005), verbal fluency (r = 0.57, p = 0.009), and score on the Multiple Sclerosis Functional Composite (r = 0.67, p = 0.003). These initial results show that the SFCI is a suitable metric for longitudinal evaluation of functional decline in MS.

Introduction

Physical disability and upper extremity function are hallmark symptoms of Multiple Sclerosis (MS), and are among the most important factors affecting quality of life [1, 2]. Another common symptom is cognitive dysfunction, affecting approximately 40% to 65% of patients and impacting employment, daily living skills, and overall quality of life [3-6]. Treatment efficacy is often judged by impact on progression of motor disability [7], with less focus on cognitive function. This is partially due to high variability of cognitive measures and to the relatively weak relationship between cognitive measures and commonly used imaging metrics in MS, such as lesion burden [8]. Here, we introduce and evaluate a composite imaging metric for the assessment of both motor and cognitive function in MS, with the ultimate goal of providing a widely-applicable metric suitable for use in longitudinal studies assessing therapeutic impact.

The two principal measures used to probe the symptomatic domains of MS are the Expanded Disability Status Scale (EDSS) and the Multiple Sclerosis Functional Composite (MSFC). Currently, the EDSS is the primary clinical test accepted by clinical trial monitors, despite the fact that the EDSS does a poor job of assessing upper limb function and cognitive deficits [9]. The MSFC probes three symptomatic domains: lower extremity function (ambulation), upper extremity function, and cognition. The cognitive component of the MSFC is based on the Paced Serial Addition Test (PASAT) [10], or, more recently, the Symbol Digit Modalities Test (SDMT) [11]. Here, we introduce an imaging-based metric that is similarly designed to provide a composite measure of neurologic deficit. This metric was develop based on previous work, with the goal of assessing both disease status and functional decline over time. The data presented here represents the first test of this metric in real-world data.

Our previous work focused on the relationship of imaging measures to motor and cognitive symptoms in MS. We reported reduced functional connectivity, assessed by resting state functional magnetic resonance imaging (rsfMRI), between the bilateral primary motor cortices in patients with MS as compared to controls [12]. A follow-up study focused on both structural and functional connectivity, showing that diffusion tensor imaging (DTI) measures in the transcallosal motor pathway were inversely correlated with rsfMRI of the primary sensorimotor cortices (SMC) in MS [13]. Given the frequency of memory impairment in MS [3, 14], we subsequently extended our study of DTI- and rsfMRI-based connectivity to the posterior cingulum bundle, a pathway connecting the antero-mesial temporal lobe (AMTL) and the posterior cingulate cortex (PCC) and subserving episodic memory [15]. In this pathway, as well as within the transcallosal motor pathway, we found an inverse correlation between radial diffusivity (RD) and rsfMRI in both MS and healthy controls [15]. Because RD is generally accepted to reflect the degree of demyelination and axonal loss in MS [16], this relationship suggests that reduced functional connectivity in monosynaptic pathways in the brain is related to the structural integrity of the white matter along that pathway. Despite this, the correlation between functional and structural connectivity in these studies was moderate (r~0.4). This suggests that, although related, the measures reflect different aspects of the biological processes that affect connectivity, and thus may be complementary.

This work presents the results of a two-year study designed to test the hypothesis that a metric based on connectivity measures along network pathways that are implicated in domains of disability specific to MS can be used as a sensitive marker of disease status and functional decline. We focus on two common domains of disability in MS, motor and cognitive function, and characterize them using measures of both structural (DTI) and functional connectivity (rsfMRI). Our previous work has shown significantly reduced structural and functional connectivity of the transcallosal motor pathway in MS [12, 17], as well as a significant correlation between structural and functional connectivity in this pathway [13, 15]. Although these were cross-sectional studies, they informed our choice for the motor disability domain pathway. We hypothesize that measures of transcallosal motor pathway connectivity will be related to measures of motor function over time, and that this relationship will be reflected in our connectivity metric. Based on work relating the posterior cingulum bundle to memory dysfunction and speed of processing in MS [18, 19], we selected the posterior cingulum bundle as the cognitive pathway. The PCC and AMTL have been implicated in episodic memory by multiple investigators [20, 21], and these regions have been shown to be abnormal in memory-impaired patients with Alzheimer’s disease [22-24]. The direct anatomic connection between these regions has been demonstrated in multiple animal studies [25-28], and by using DTI and functional connectivity measurements in healthy human controls [15]. We hypothesize that measures of PCC-AMTL pathway connectivity will be related to measures of cognitive function over time, and that this relationship will be reflected in our connectivity metric.

Theory

Here, we present an evaluation of this metric, called the Structural and Functional Connectivity Index (SFCI). The SFCI was initially developed using a generated dataset with parameters based on the results of previous work (see S1 Text), and was constructed to measure combined structural and functional connectivity in two commonly-impacted domains of MS–motor and cognitive function. The metric can be normed to either control or patient samples and will decrease with increased connectivity. This provides important flexibility to the metric, as the observed variability of the component measures were higher in MS than in controls. In the current analysis, mean and standard deviations taken from the control group were used to norm Eqs 1 and 2 below. When used in this manner, the SFCI is an evaluation of pathway connectivity in individual MS patients as compared to normative values for healthy controls.

Motor SFCI metric

Using measures from the transcallosal motor pathway, the motor SFCI component is calculated according to the following equation: where fc and sc are functional and structural connectivity, respectively, for an individual, fc and sc are the mean functional and structural connectivity, respectively, of the normative sample, and and are the standard deviations of functional and structural connectivity, respectively, of the normative sample.

Cognitive SFCI metric

Because the PCC-AMTL pathway is intrahemispheric and bilateral, the cognitive SFCI component (Z) was constructed to be sensitive to impairment in either hemisphere. The left and right hemisphere PCC-AMTL measures are z-scored separately according to the following equation:

The notation of Eq 2 is identical to that of Eq 1. L = left, R = right. The results are normalized scores for the left () and right () hemispheres.

SFCI combined metric

The motor (Z) and cognitive ( and ) SFCI metrics are combined in the following manner: where Z is taken to be the minimum Z score of the two hemispheres.

Materials and methods

Data acquisition

Participants

Twenty-five participants with MS and 12 healthy controls were enrolled for longitudinal study. Of this sample, twenty participants with MS and 9 healthy controls were included in the final data analysis (see Results: Sample Description for details). Participants with MS completed six study visits over a two year period. Year 1 included four visits spaced at four month intervals, and year 2 included two visits spaced at two month intervals. Controls were selected to provide an age-, education-, and sex-matched sample and completed two study visits—one at baseline and the other after two years. Fig 1 shows the visit timeline and testing schedule. All data were acquired after providing written informed consent under protocol 13–994, approved by the Cleveland Clinic Institutional Review Board (Federalwide Assurance number 00005367). Anonymized data that support the findings of this study is available at: doi.org/10.6084/m9.figshare.14618265.

Fig 1

Schematic of study visits.

MS = Multiple Sclerosis.

Clinical and cognitive evaluation

The EDSS was measured by an experienced MS neurologist (LS, DO). All participants completed a cognitive battery at baseline and two years, administered by an experienced psychometrist under the supervision of a licensed clinical psychologist (SR). Key tests from the Minimal Assessment of Cognitive Function in MS (MACFIMS) were administered, and raw scores for each measure were corrected using published norms (Table 1). To lessen test-retest effects, alternate test forms were counterbalanced in presentation.

Table 1

Neuropsychological measures.

Cognitive domain	Test	Resulting measures
Verbal episodic memory	California Verbal Learning Test–II (CVLT)	T-score [29]
Visuospatial episodic memory	Brief Visuospatial Memory Test–Revised (BVMT)	T-score; delayed recall T-score [30]
Processing speed	Symbol Digit Modalities Test (SDMT)	corrected score [31]
Executive function	Delis–Kaplan executive function system (D-KEFS)	composite scaled score [32]
Spatial processing	Judgment of Line Orientation Test (JLO)	corrected score [33]
Expressive Language	Controlled Oral Word Association Test (COWAT)	corrected score [34]

MRI acquisition

Data were acquired on a Siemens TIM Trio 3T MRI scanner (Erlangen, Germany) using a 12-channel receive-only head array or a Siemens Prisma 3T MRI scanner using a 20 channel head-neck array. Approximately one quarter of the way into data collection, the Trio scanner was upgraded to the Prisma platform. Statistical analyses described below accounted for possible systematic effects of scanner change. A bite bar was used to minimize participant motion. The bite bar consisted of a molded dental impression (Kerr Dental, Inc., Brea CA) taken of the subject’s teeth and affixed to a removable plastic frame placed over the head coil.

At baseline, participants performed a unilateral patterned finger tapping task during scanning. One scan was performed for each hand, with left/right order randomized across participants. Participants were asked to tap their fingers as fast as possible while maintaining the following pattern: thumb, middle, pinkie, index, ring. During scanning, fiber optic gloves were used to monitor task performance and the verbal commands “start” and”stop” were used to indicate tap/rest periods.

Directly prior to scanning, participants met with the experimenter. The experimenter described the scanning session, including familiarizing the participant with the bite bar apparatus, highlighting the importance of remaining still during scanning, and describing the finger tapping task. Participants practiced tapping both with and without the fiber optic gloves until they felt comfortable with the pattern.

The following scans were performed in the order shown:

Scan 1: Whole brain (MPRAGE): 176 axial slices at 0.94 mm thickness; field-of-view (FOV) 240 mm×240 mm; matrix 256×256; voxel size 0.94 mm3; inversion time (TI)/echo time (TE)/repetition time (TR)/flip angle (FA) = 1100 ms/2.84 ms/1860 ms/10°; bandwidth (BW) 180 Hz/pixel. Acquisition time 4:20 minutes. In addition to providing anatomic detail, this scan was used to provide gray and whiter matter masks using Freesurfer segmentation [35].

Scan 2: SPACE 3D FLAIR: 144 sagittal slices at 1.2 mm thickness; FOV 256 mm×224 mm; matrix 256×224; voxel size 1.2×1×1 mm3; TI/TE/TR/FA = 2000 ms/395 ms/6500 ms/120°; BW 698 Hz/pixel; 6/8 partial Fourier acquisition; GRAPPA factor = 2; 24 reference lines. Acquisition time 5:14 minutes.

Scan 3: Whole brain BOLD resting state scan: 31 axial slices at 4 mm thickness; FOV 256 mm×256mm; matrix 128×128; voxel size 2×2×4mm3; TE/TR = 29 ms/2800 ms; BW 1954 Hz/pixel; 132 repetitions. Data was acquired using a prospective motion-controlled, gradient recalled echo, echoplanar acquisition [36]. Acquisition time 6:10 minutes. Prior to the start of the scan, the participant was instructed to rest with eyes closed and refrain from any voluntary motion.

Scan 4: Whole brain BOLD tapping activation scan (baseline only): Data were acquired using the same parameters as for scan 3, with 160 repetitions. Acquisition time 7:29 minutes, 2 runs. Participants performed a unilateral patterned finger tapping task during four blocks of 16 volume acquisitions (45 seconds), interleaved with rest blocks of the same duration. Prior to the start of the scan, participants were reminded of the tapping pattern and verbal commands used during the task.

Scan 5: High angular resolution diffusion imaging (HARDI): 51 axial slices at 2 mm thickness; FOV 256 mm×256 mm; matrix 128×128; voxel size 2mm3; TE/TR = 92 ms/7800 ms; BW 1628 Hz/pixel; 5/8 partial Fourier acquisition; 71 noncollinear diffusion-weighting gradients with b = 1000 s/mm2, eight b = 0 volumes, NEX = 4. Twice-refocused spin echo was used to minimize eddy current artifact [37]. Acquisition time 10:24, two averages.

Data analysis

MRI data processing

For the fMRI tapping analysis, motion correction was performed using volumetric and slice-based estimators from SLOMOCO [38]. Using in-house code implemented in Matlab R2018b, a 4mm 2D in-plane Gaussian filter was used to improve functional contrast-to-noise ratio and make in-plane and through-plane resolution more similar [39]. The AFNI routine 3dDeconvolve was used to fit a boxcar reference function representing the off/on activation paradigm to the time series data of each voxel [40]. The result was a whole brain Student’s t map, used to determine regions of significant involvement in the uni-manual tapping task.

For the rsfMRI analysis, physiologic signals were regressed using RETROICOR as provided by AFNI [40, 41]. For the majority of studies, a plethysmograph and respiratory bellows were used during scanning to sample cardiac and respiratory signals at 400 Hz. For 15 scans, technical problems prevented acquisition of one or more physiologic signals. In those cases, physiologic signals were estimated with PESTICA [42]. Simultaneously with regression of physiologic noise, data were retrospectively motion-corrected using volumetric and slice-based estimators from SLOMOCO [38]. Mean (TDzmean) and maximum (TDzmax) voxel-level residual displacement were used to characterize motion artifact in each scan for use in summary statistics and quality control. If a participant had motion values of TDzmax > 1mm and TDzmean > 0.2mm at baseline, they would be administratively withdrawn from the study [43]. Follow-up scans over the same threshold were flagged for visual inspection of the rsfMRI time series and resulting correlation maps, detailed below. Using in-house code implemented in Matlab R2018b, spatial filtering with a 2D in-plane Hamming filter was performed to improve functional contrast-to-noise ratio with minimal loss of spatial resolution [44] and the data were temporally filtered to remove all fluctuations above 0.08Hz [45, 46].

For the DTI analysis, image series were concatenated, followed by iterative motion correction [47] that included updating of diffusion gradient directions [48]. The diffusion tensor and diffusivities were calculated on a voxel-by-voxel basis. All analyses were completed using in-house software.

Other MRI measures

In order to compare to MRI measures that are typically used as outcomes in clinical trials of MS therapies, the MPRAGE and FLAIR were used to measure the following in all participants with MS:

At each visit, lesion volume (LV) was calculated using FLAIR images [49]. A detailed description of lesion volume calculation is available in reference [49]. Briefly, optimal thresholding and connected-components analysis were used to generate a starting point for segmentation. A 3D radial search was then performed to locate probable locations of the intra-cranial cavity. The result was total LV for each MS participant at each visit.

At each visit, brain parenchymal fraction (BPF) was calculated using anatomical images. BPF was defined as the ratio of brain parenchymal volume to the total volume within the brain surface contour. The method used here [50] used the segmentation algorithm described above [49] for brain surface detection and brain volume calculation.

Seed region definition

Seeds were defined separately for each subject using baseline data. The AFNI routine align_epi_anat.py was used to perform affine 12 parameter alignment of task-based fMRI and anatomical volumes to rsfMRI and DTI volumes [51]. The same routine was used to align baseline images to the same imaging modality at each visit (e.g. baseline rsfMRI to visit 3 rsfMRI), so that baseline seeds in DTI and rsfMRI space could be propagated to future visits. To transform seeds between volumes, transformation matrices output by align_epi_anat.py were applied using the AFNI routine 3dAllineate with nearest neighbor interpolation. Seeds used in the rsfMRI analysis included a subset of voxels used as seeds for DTI tracking, to ensure that only grey matter was included.

Sensorimotor cortex (SMC). The uni-manual tapping activation maps were used to identify the maximally activated voxel in the M1 motor region of the contralateral hemisphere (Fig 2A). To create the seeds used for DTI tracking, a 6 mm sphere was centered at this voxel in each hemisphere and was transformed from fMRI space to DTI space (Fig 2A, lower left). To create the seeds used for the rsfMRI analysis, a 9-voxel in-plane ROI was centered at the maximally active voxel in each hemisphere. ROIs were visually inspected and, if necessary, manually eroded to ensure all voxels were located in gray matter and not in intragyral CSF (Fig 2A, lower right). Seeds were transformed from the task-activated volume to the rsfMRI volume.

Fig 2

Example of seed placement.

A) Seed placement for the left hemisphere sensorimotor cortex region of interest (ROI). The maximally activated voxel in the primary motor cortex during the right hand tapping functional MRI task is circled in blue. The lower left shows the 6mm sphere centered at that voxel used for the structural connectivity (DTI) analysis. The lower right shows the 9-voxel in plane seed used for the functional connectivity (rsfMRI) analysis. Anatomical ROIs for the B) posterior cingulate cortex and the C) entorhinal cortex. Insets show the cross section of the 9-voxel seed used for the rsfMRI analysis for each ROI.

PCC-AMTL. The PCC-AMTL seeds were defined using a combination of anatomical and functional data. First, the MPRAGE was linearly transformed to Talairach space using the AFNI routine @auto_tlrc. To define the PCC, a 6mm sphere was placed in each hemisphere at [–12 –42 36], according to coordinates reported in Greicius et al., 2003 [23] (Fig 2B). For the AMTL region, ROIs were drawn manually on the MPRAGE in Talairach space and included the entorhinal cortex and medial subiculum from coronal slices 10P-30P (Fig 2C). These ROIs were transformed to original MPRAGE space, checked for accuracy, and transformed to the DTI and rsfMRI volumes. The transformed ROIs were used as the seeds for DTI tracking.

The rsfMRI time series was masked to include only grey matter voxels within the PCC and AMTL ROIs. Cross-correlation was used to identify the two voxels with the highest correlation between the PCC and AMTL [15]. For each ROI, this voxel was taken as the center of a 9-voxel in-plane seed, which functioned as the PCC and AMTL seeds for the functional connectivity analysis (Fig 2B and 2C, inset).

Structural connectivity calculation

The DTI metric, sc, was calculated using probabilistic tracking, performed using an analytic tracking algorithm based on partial differential equations and global and local constraints based on voxel level fiber orientation determined from the HARDI data [52]. The structural connectivity determination for each tracked pathway is described in detail in reference [13], and was calculated using in-house code. Briefly:

FOD determination: For each voxel, the 71-direction diffusion data were used to determine the local fiber orientation distribution (FOD) function [53]. This function determines the local probability for propagation of tracks from the seed (PCC and left SMC) to the target (AMTL and right SMC). Voxel-level probabilities for belonging to the pathway were then determined. Fig 3 shows thresholded tracks in a representative participant.

Fig 3

Example of probabilistic tracking.

Individual subject probabilistic tracking results. A) Blue = transcallosal SMC, red = left hemisphere PCC-AMTL, yellow = right hemisphere PCC-AMTL track. The bottom row shows additional anatomical detail for B) the SMC and C) the left PCC-AMTL tracks. Displayed tracks are count-thresholded surfaces derived from probabilistic maps described in the text. SMC = primary sensorimotor cortices; PCC-AMTL = posterior cingulate cortex-antero-mesial temporal lobe.

DTI determination: Tensor values of water diffusion were calculated for each voxel using a standard log-linear least squares method [54]. The tensor was diagonalized and the scalar measures of tissue microstructure (axial diffusivity (AD), radial diffusivity (RD), mean diffusivity (MD), and fractional anisotropy (FA)) were calculated for each voxel.

Pathway dependent diffusion measures were calculated using the track probability map, the scalar diffusion values, and a white matter mask [35]. Each measure is calculated according to: where the D(v) is the particular tensor-based value of interest (e.g. FA) at voxel v and w(v) is a so-called track probability map, in which the value of each voxel equals probability of track membership generated by the probabilistic tractography algorithm for that voxel. WM is a mask that is set to one for voxels determined to be mostly white matter from the Freesurfer segmentation. The result is 〈FA〉, 〈MD〉, 〈AD〉, and 〈RD〉 for every participant. Based on our prior work, we took pathway averaged RD to be our measure of structural connectivity, or sc.

Functional connectivity calculation

Using in-house code implemented in Matlab R2018b, the rsfMRI metric, fc, was calculated in the following manner:

A reference time series was calculated from the linearly detrended arithmetic average of the nine seed voxels. For the PCC-AMTL pathway, the bilateral PCC seeds were used. For the transcallosal motor pathway, the left SMC seed was used.

A whole-brain voxel-wise cross correlation map was calculated for each pathway of interest using the reference time series, resulting in three whole-brain rsfMRI correlation maps: left hemisphere SMC and right and left hemisphere PCC.

To account for individual differences in global signal, each correlation map was converted to a Student’s t map. For each Student’s t map, the whole-brain distribution was normalized to unit variance and zero mean [46]. The mean and variance from each distribution was used to convert the Student’s t to a z-score.

The metric fc was calculated for each pathway by taking the average z-score of the nine voxels in the target seed. For each hemisphere’s PCC-AMTL pathway, the corresponding AMTL seed was used. For the transcallosal motor pathway, the right SMC ROI was used.

Statistical analysis

To assess group differences in baseline and slope of imaging measures and SFCI and component measures, data from MS participants and controls at baseline and the final visit were entered into a linear mixed-effects (LME) analysis using R [55]. The model tested the impact of group and the effect of time using visit, with education and scanner entered as covariates: [group * visit + scanner + education]. To assess group differences in the rate of change, a group × visit interaction was included. Random effects of subject were included for intercept and slope. A reduced model (omitting group: [visit + scanner + education]) was compared to the full model using the likelihood ratio test (LRT). The false discovery rate adjustment (FDR) was applied to correct for multiple comparisons [56].

Using all six visits, a second LME analysis assessed the impact of time in MS participants. The model assessed the linear effect of time using visit, with education, scanner, disease duration, and sex entered as covariates: [visit + scanner + education + disease duration + sex]. Random effects of subject were included for intercept and slope. A reduced model (omitting visit: [scanner + education + disease duration + sex]) was compared to the full model using the LRT. The FDR was applied.

In participants with MS, the relationship of baseline SFCI and associated imaging measures to baseline and final visit behavioral measures were assessed using Pearson correlations. Measures of motor function (the 9 hole peg test and 25 ft. walk) were log-transformed. To determine if changes in imaging were related to behavioral measures, slopes of SFCI components and imaging measures taken from the LME analysis of time were correlated with behavioral measures taken at the final visit. The FDR was applied for each imaging measure. Potential covariates, including age, education, disease duration, and EDSS, were correlated with imaging behavioral measures, including measures of change. A measure was included as a covariate only if it showed a relationship to either imaging or behavioral measures.

Results

Sample description

A total of 25 MS participants and 12 controls were enrolled. Three participants (2 with MS and 1 control) withdrew from the study prior to completion. Throughout the course of the study, two participants with MS were administratively removed due to a Beck Depression Inventory score > 20, two controls were removed due to very low baseline cognitive performance, and one participant with MS was removed due to extreme motion in all baseline scans. The final sample consisted of 20 participants with MS and 9 controls.

One participant with MS had four rather than six time points in the final sample. This participant withdrew after the fourth visit, but completed the final visit cognitive evaluation at that time. The third time point was removed for a different participant with MS, because technical difficulties prevented the acquisition of DTI data at that visit. The remaining sample was composed of 117 MS and 18 control visits.

Demographics, cognitive performance, and disease characteristics are reported in Table 2. All participants were strongly right handed (Edinburgh score>80) [57]. The groups did not show differences in age or sex, but the MS group had a significantly lower education level (p = 0.046). The MS group showed a decrease in MSFC from baseline to the final visit (p = 0.002). One participant with MS showed a 2-point increase in EDSS, and two showed a 1-point increase. At baseline, the MS group scored lower than controls on the CVLT (p = 0.013), BVMT-DR (p = 0.019) and SDMT (p = 0.039). The SDMT was the only cognitive measure that showed a significant decline from baseline to final visit in the MS group (p = 0.011).

Table 2

Participant demographics, cognitive performance, and clinical characteristics.

	MS		HC		MS v HC	MS v MS
	Baseline	Final Visit	Baseline	Final Visit	Baseline p	Time p
Demographics
N (males)	20 (6)	-	9 (3)	-	0.938	-
Age	50.95 ± 6.8	-	48.11 ± 7.5	-	0.324	-
Education	14.10 ± 2.2	-	16.00 ± 2.4	-	0.046	-
Cognitive performance
CVLT	46.00 ± 7.9	42.65 ± 11.6	54.44 ± 7.7	58.11 ± 7.5	0.013	0.088
BVMT	46.20 ± 13.6	42.45 ± 12.9	53.33 ± 5.9	53.89 ± 7.7	0.145	0.217
BVMT-DR	46.65 ± 14.3	44.65 ± 13.7	59.22 ± 6.8	57.44 ± 9.3	0.019	0.579
SDMT	50.10 ± 11.7	45.45 ± 12.3	59.78 ± 9.5	57.89 ± 9.3	0.039	0.011
D-KEFS	11.53 ± 2.7	11.11 ± 3.4	13.22 ± 3.1	13.33 ± 2.5	0.145	0.282
JLO	24.20 ± 3.8	24.40 ± 4.2	26.67 ± 3.3	26.89 ± 3.7	0.102	0.711
COWAT	38.40 ± 13.0	38.85 ± 14.1	40.78 ± 15.1	44.44 ± 10.3	0.669	0.816
Clinical characteristics
EDSS	3.5 (2–6.5)	4.0 (2.5–6.5)	-	-	-	0.069
MSFC	-0.264 ± 0.58	-0.446 ± 0.58	0.586 ± 0.38	0.383 ± 0.41	7.3×10−4	0.002
DD	20.5 (4–33)	-	-	-	-	-
Lesion vol. (ml)	16.89 ± 13.1	17.24 ± 10.9	-	-	-	0.994
25 ft. walk (sec.)	8.29 ± 5.6	7.57 ± 2.8	4.03 ± 0.5	4.15 ± 0.5	0.043	0.175
9HPd (sec.)	29.00 ± 24.1	29.68 ± 18.6	18.38 ± 2.4	19.95 ± 3.0	0.203	0.703
9HPnd (sec.)	25.06 ± 5.9	26.74 ± 5.7	20.43 ± 3.2	20.42 ± 2.3	0.039	0.028
Disease course	12 RRMS	-	-	-	-	-
	3 PPMS
	5 SPMS

Age, education, cognitive performance measures, MSFC, lesion volume, 25 ft. walk, and 9HP report mean ± standard deviation. DD reports mean (range). EDSS reports median (range). BVMT = Brief Visuospatial Memory Test; BVMT-DR = Brief Visuospatial Memory Test, delayed recall; COWAT = Controlled Oral Word Association Test; CVLT = California Verbal Learning Test; D-KEFS = Delis–Kaplan Executive Function System; DD = disease duration; EDSS = Expanded Disability Status Scale; HC = healthy control; JLO = Judgment of Line Orientation Test; MS = Multiple Sclerosis; MSFC = Multiple Sclerosis Functional Composite; PPMS = primary progressive; RRMS = relapse remitting; SDMT = Symbol Digit Modalities Test; SPMS = secondary progressive; 9HPd = 9 hole peg, dominant hand; 9HPnd = 9 hole peg, non-dominant hand.

Scanner change

The scanner used for this study was upgraded from a Trio to a Prisma during data collection. Three controls were scanned using the Trio at baseline and the Prisma at follow up. The remainder of the controls were scanned using the Prisma at both visits. Of participants with MS, one was scanned using the Trio for visits 1–4 and seven were scanned using the Trio for visits 1–3. For these eight participants, the remaining scans were completed using the Prisma. All other participants with MS were scanned using the Prisma at all visits. Separately for each group, t-tests were used to assess differences in imaging and demographic measures between those scanned at baseline on the Trio or Prisma. Participants with MS with baseline scans on the Prisma showed a significantly longer disease duration than those on the Trio (p = 0.0124), and fc of the left PCC-AMTL was significantly lower at visit 2 (p = 0.0088). No other measures showed differences in either MS or controls.

MRI quality assessment

As described above, one participant with MS failed the baseline motion criteria and was administratively withdrawn from the study. On the basis of the criteria described for the remaining scans, rsfMRI scans from 20 participants with MS (20/117 scans) and 0 controls (0/18 scans) were inspected for evidence of motion-related artifact, including rings of correlation around the outside of the head, correlation in the ventricles, and rapid correlation pattern changes from slice to slice (a consequence of motion in an interleaved style acquisition). Three connectivity scans (a single time point in each of three participants with MS) showed severe motion-related artifacts and were excluded from further analysis. Mean TDzmean and TDzmax were not significantly different between groups or across time in the remaining 114 connectivity scans.

Group differences

Baseline group-averaged rsfMRI maps are shown in Fig 4 (p < 1.0×10−4, cluster size threshold = 60 voxels). The LME assessing the effect of group (MS vs. control) showed no differences in left PCC-AMTL and SMC fc at baseline or over time (Table 3; Fig 5A). Right PCC-AMTL showed effects of group*visit (p = 0.0104), but the model comparison did not survive FDR correction (p = 0.0241). The group analysis did not show a significant effect of education or scanner for any fc measure. The only sc measure to show a group difference was the SMC (p = 0.0054), showing an effect of group (p = 0.0115), but not of group*visit interaction, education, or scanner (Table 3; Fig 5B).

Fig 4

Thresholded group-averaged functional connectivity maps.

Single voxel threshold is 1×10−4 with a cluster requirement of 60 voxels. A. Transcallosal connectivity of the SMC. PCC-AMTL connectivity on the left (B) and right (C). HC = healthy control; LH = left hemisphere; MS = Multiple Sclerosis; PCC-AMTL = posterior cingulate cortex-antero-mesial temporal lobe; RH = right hemisphere; SMC = primary sensorimotor cortices.

Table 3

Results of the LRT for group differences from baseline to the final visit in SFCI components and associated measures.

	Controls vs. MS
	LR	LRT p	group p	group*visit p
SFCI component
SFCI	9.48	0.0088	0.0707	0.0814
Zcog, left	2.98	0.2255	0.4617	0.3075
Zcog, right	7.22	0.0269	0.1992	0.1056
Zmotor	8.72	0.0128	0.0603	0.1632
Imaging measures
fc PCC-AMTL, left	1.40	0.4963	0.3473	0.3252
fc PCC-AMTL, right	7.45	0.0241	0.1200	0.0104
fc SMC	0.97	0.6155	0.4357	0.3846
sc PCC-AMTL, left	3.46	0.1775	0.1489	0.5577
sc PCC-AMTL, right	5.27	0.0718	0.0323	0.7381
sc SMC	10.45	0.0054	0.0115	0.1820

The LRT was performed on the full model vs. the reduced model, which omits group. Values in bold survived FDR correction. fc = functional connectivity; LR = likelihood ratio; LRT = likelihood ratio test; MS = Multiple Sclerosis; PCC-AMTL = posterior cingulate cortex-antero-mesial temporal lobe; sc = structural connectivity; SFCI = Structural and Functional Connectivity Index; SMC = primary sensorimotor cortices; Zcog = SFCI cognitive component; Zmotor = SFCI motor component.

Fig 5

Group-averaged imaging measures.

Group-averaged imaging measures plotted for MS and controls at baseline (1) and final visit (6). Error bars represent standard error. LRT p-values are included in each plot. A. rsfMRI (fc), B. DTI (sc), and C. SFCI component measures in the tracks of interest. * denotes significant result. HC = healthy controls; see Table 3 for additional abbreviations.

Table 3 shows the results of the LME analysis of MS vs. control differences for the SFCI and component measures. The full model (including “group”) was significant for total SFCI (p = 0.0088) and Zmotor (p = 0.0128), although the group and group*visit interaction terms did not reach significance (Fig 5C).

Impact of time

In participants with MS, the LME assessing the impact of time showed a reduction in all fc measures (p < 0.05), but the change in SMC did not survive FDR correction (p = 0.0472; Table 4; Fig 6A). The time analysis did not show an effect of education, disease duration, or sex for any fc measure, although left PCC-AMTL showed an effect of scanner (p = 0.0172). In controls, paired t-tests of baseline and follow-up fc measures showed no significant differences (PCC-AMTL, left p = 0.0809; PCC-AMTL, right p = 0.5907; SMC p = 0.3870). None of the sc measures showed a significant impact of time in the MS participants (Table 4; Fig 6B). Paired t-tests showed no differences between baseline and follow-up sc measures in controls (PCC-AMTL, left p = 0.1874; PCC-AMTL, right p = 0.9430; SMC p = 0.9123). In participants with MS, BPF did not change over time. LV showed a weak increase over time, although this did not survive FDR correction (p = 0.0437; Table 4; Fig 6E).

Table 4

The impact of time (visit) on SFCI components and associated measures in participants with MS.

	MS
	LR	LRT p	visit p
SFCI component
SFCI	6.67	0.0098	0.0121
Zcog, left	2.07	0.1505	0.1609
Zcog, right	0.70	0.4033	0.4192
Zmotor	5.37	0.0205	0.0229
Imaging measures
fc PCC-AMTL, left	4.95	0.0261	0.0288
fc PCC-AMTL, right	6.61	0.0102	0.0084
fc SMC	3.94	0.0472	0.0408
sc PCC-AMTL, left	0.46	0.4976	0.5344
sc PCC-AMTL, right	1.07	0.3020	0.3030
sc SMC	3.66	0.0556	0.0576
Lesion volume	4.07	0.0437	0.0504
BPF	1.44	0.2290	0.2121

The LRT was performed on the full model vs. the reduced model, which omits visit. Values in bold survived FDR correction. BPF = brain parenchymal fraction; see Table 3 for additional abbreviations.

Fig 6

Averaged imaging measures in participants with MS as a function of visit.

Error bars represent standard error. LRT p-values are included in each plot. A. rsfMRI (fc), B. DTI (sc), C. SFCI components, and D. SFCI in the tracks of interest. E. Whole-brain imaging measures. *denotes significant result. BPF = brain parenchymal fraction; see Table 3 for additional abbreviations.

The SFCI and the Zmotor component declined over time in MS (p < 0.03), independently of scanner, education, disease duration, and sex (Table 4; Fig 6C and 6D).

Imaging and behavioral measures

Tables 5 and 6 show Pearson correlation coefficients between imaging and behavioral measures in participants with MS. In the interest of space, only behavioral measures that showed a significant relationship to at least one imaging measure are included in the tables. One participant was an outlier (greater than 3 standard deviations from the mean) on the log-transformed 9 hole peg test score at both baseline and final visit. One participant was an outlier on the log-transformed 25 ft. walk score at baseline. Correlations with these measures excluded the outlier values. Age, education, disease duration, and EDSS were not significantly related to imaging or cognitive measures and were not included as covariates in the following analyses. The three participants that showed a 1-point or greater increase in EDSS did not show differences in imaging measures compared to the rest of the sample.

Table 5

Pearson correlation coefficients between baseline imaging measures and behavioral measures at baseline and the final visit in 20 participants with MS.

	Baseline visit				Final visit
	SDMT	COWAT	MSFC	9HPnd	SDMT	COWAT	MSFC
SFCI component
SFCI	0.645**	0.453	0.590*	-0.519*	0.545*	0.564*	0.504
Zcog, left	0.696^	0.568*	0.670**	-0.379	0.559*	0.675**	0.465
Zcog, right	0.702^	0.510*	0.566*	-0.257	0.774^	0.719^	0.709^
Zmotor	0.513*	0.359	0.489	-0.549*	0.366	0.407	0.340
Imaging measures
fc PCC-AMTL, left	0.390	0.454	0.475	-0.040	0.177	0.383	0.095
fc PCC-AMTL, right	0.112	0.291	0.025	0.379	0.334	0.288	0.231
fc SMC	-0.371	-0.038	-0.356	-0.193	-0.437	-0.164	-0.478
sc PCC-AMTL, left	-0.696^	-0.516*	-0.632**	0.446	-0.612**	-0.674**	-0.514*
sc PCC-AMTL, right	-0.702^	-0.441	-0.588*	0.397	-0.704^	-0.661**	-0.667**
sc SMC	-0.720^	-0.393	-0.687^	0.477	-0.599*	-0.507*	-0.613**
Lesion volume	-0.017	0.085	-0.232	0.330	-0.552	0.179	0.284
BPF	0.090	-0.157	0.212	-0.161	0.550	-0.176	-0.226

Uncorrected p-values

* < 0.03

** < 0.005

^ < 0.001. Values in bold survived FDR correction. BPF = brain parenchymal fraction; COWAT = Controlled Oral Word Association Test; MSFC = Multiple Sclerosis Functional Composite; SDMT = Symbol Digit Modalities Test; 9HPnd = 9 hole peg, non-dominant hand; see Table 3 for additional abbreviations.

Table 6

Pearson correlation coefficients for the slope of imaging measures related to behavioral measures at the final visit in 20 participants with MS.

	Final Visit
	SDMT	COWAT	MSFC	9HPnd
SFCI component
SFCI	0.603**	0.568**	0.665^	-0.518*
Zcog, left	0.534*	0.574**	0.522*	-0.284
Zcog, right	0.635^	0.558*	0.617**	-0.165
Zmotor	0.565**	0.461	0.738^	-0.529*
Imaging measures
fc PCC-AMTL, left	0.180	0.143	0.127	0.172
fc PCC-AMTL, right	0.245	0.285	0.145	0.439
fc SMC	-0.087	0.024	-0.006	-0.397
sc PCC-AMTL, left	-0.544*	-0.598**	-0.550*	0.384
sc PCC-AMTL, right	-0.328	-0.252	-0.261	-0.036
sc SMC	-0.559*	-0.560*	-0.688^	0.378
Lesion volume	-0.332	-0.246	-0.478	0.543*
BPF	0.460	0.278	0.633**	-0.482

Uncorrected p-values

* < 0.03

** < 0.01

^ < 0.005. Values in bold survived FDR correction. BPF = brain parenchymal fraction; COWAT = Controlled Oral Word Association Test; MSFC = Multiple Sclerosis Functional Composite; SDMT = Symbol Digit Modalities Test; 9HPnd = 9 hole peg, non-dominant hand; see Table 3 for additional abbreviations.

Table 5 reports correlation coefficients between baseline imaging measures and behavioral performance at baseline and the final visit. In all tracks of interest, lower baseline sc was associated with higher SDMT (p < 0.0007) and MSFC (p < 0.0007) scores at baseline and with higher SDMT (p < 0.006), MSFC (p < 0.03), and COWAT (p < 0.03) scores at the final visit. Baseline SDMT and non-dominant hand 9 hole peg score were the only measures associated with Zmotor (p < 0.03). Other baseline SFCI measures were positively associated with baseline SDMT (p < 0.005) and MSFC (p < 0.01) scores, with higher SFCI related to higher scores. Similar positive relationships were observed between baseline SFCI measures and performance on the SDMT (p < 0.02) and COWAT (p < 0.01) at the final visit, although only right Zcog score was related to MSFC (p = 0.001) at the final visit.

Table 6 reports correlation coefficients between the change in imaging measures, represented by slopes calculated from the LME analysis of time, and behavioral performance at final visit. In the left PCC-AMTL and SMC, the change in sc was negatively related to SDMT (p < 0.02), COWAT (p < 0.01), and MSFC (p < 0.02), such that an increase in sc was associated with lower behavioral scores at the final visit. All SFCI slopes were positively related to SDMT (p < 0.02), MSFC (p < 0.03), and, with the exception of Zmotor, COWAT (p < 0.02), indicating that an increase in SFCI was related to higher behavioral scores at the final visit. Similarly, the change in BPF was positively related to MSFC (p < 0.005). Both total SFCI and Zmotor were negatively related to the non-dominant hand 9 hole peg score (p < 0.03), so that a larger decrease in SFCI was associated with longer performance times at the final visit.

Discussion

This work demonstrates that a combined structural and functional connectivity metric can be a sensitive tool for identification of clinical impairment in a neurological disease such as MS. The SFCI was constructed to probe neural pathways involved in symptomatic domains of MS, using complementary measures that represent different aspects of disease. For example, the SFCI and the Zmotor component showed a difference between the MS and control groups and a change over time in participants with MS. Our fc measures showed no group differences at baseline, but demonstrated a significant decline over two years in MS participants. Conversely, non-significant baseline differences in sc measures drove group differences in SFCI.

Longitudinal effects on functional connectivity have not been studied widely in neurologic diseases [58], and the longitudinal relationship between functional and structural connectivity has not been well-characterized. Multiple investigators have reported a cross sectional relationship between structural and functional connectivity [59-65]. Some studies report a positive relationship [59-65], while others report a more complicated or inverse relationship [66, 67]. A recent longitudinal study of aging by Fjell et al. studied multiple pathways that were identified by either structural connectivity or functional connectivity [68]. They concluded that age-related changes in functional and structural connectivity are not necessarily strongly correlated, but that the highest positive change relationship is found in regions with a measured functional connection. We interpret this study to imply that the existence of a structural pathway does not necessarily imply a functional connection, but a functional connection does imply a structural connection.

In MS participants, relationships between SFCI and behavioral measures were driven by sc, with all pathway measures related to the MSFC and its component measure, the SDMT. Although baseline SFCI was related to both baseline and longitudinal measures, the change in SFCI was also significant, so that those participants scoring lower on behavioral measures at follow up were more likely to show a decline in SFCI over time. The fact that no significant relationships were found between memory measures and imaging measures of the PCC-AMTL was unexpected, as previous work [19] found that RD was significantly related to both SDMT and BVMT in MS. A possible explanation is that the prior study used a DTI acquisition that had higher spatial resolution and was tailored to target the PCC-AMTL pathway. Although the SDMT has been associated with track-specific changes in MS [69, 70], it has also been associated with whole brain measures such as atrophy and lesion volume [71]. Given the association of the PCC-AMTL to memory function, future work is required to clarify the lack of relationship seen here.

Clinical trials of MS treatments typically do not include structural and functional connectivity measures as outcomes, more commonly reporting LV and brain atrophy. Although LV has been previously associated with episodic memory in MS [72], here LV showed no change over the course of the study and showed no relationship to cognitive measures. Similarly, baseline BPF was not related to cognitive measures. The magnitude of BPF change was related to MSFC at the final visit, although changes in SFCI, Zmotor, and sc SMC showed similar or stronger relationships.

In this study, we show that a combined metric of structural and anatomic connectivity in symptomatic domains of MS is sensitive to the longitudinal progression of disease over two years. The SFCI was constructed as a single, MRI-based metric that can be easily implemented as an outcome measure in clinical trials. The metric is z-scored and changes have, in principle, a straightforward statistical interpretation. A decrease in the metric over time reflects declining pathway connectivity, which may be a surrogate for worsening disease. It is important to note that it was necessary to use a statistical model in this study to demonstrate statistically significant changes in the metric due to the major equipment change that occurred during the study. In a situation with stable equipment, changes in the metric have a simple statistical interpretation.

The proposed SFCI measure performed well in this sample, and the focus on specific pathways makes the interpretation straightforward. The pathway-based imaging measures used in this study have an advantage over other, region-based imaging measures in that it is not necessary to perform any spatial normalization. Although larger, anatomical-based landmarks were used as starting points for seed determination, the final seed regions were identified within-subject by brain function-related techniques, and all metrics were determined in the subject’s native space. This eliminates errors and processing biases that can be introduced by spatial normalization across groups. On the other hand, this methodology is difficult to implement on a large scale, and would benefit from simplification.

The performance of the SFCI in this initial test is encouraging, but given our relatively small sample size, it will require replication in a larger sample and will likely benefit from refinement. As only a few participants showed declines in function as measured by EDSS, we were unable to describe the relationship of SFCI to clinically relevant functional decline. Larger samples will also be required to break down SFCI performance by disease course or treatment. Finally, the methods used here are highly specialized and require considerable data processing effort. Wider applicability of the SFCI will require additional simplification and automation of data processing.

The pathways used here were chosen based on past work showing relationships between connectivity strength and cognitive and functional measures in MS. Although disease mechanisms differ, it is worth noting that the development of a pathway-based, domain-specific composite measure is not restricted to MS. This approach may be applicable in other neurologic diseases that have symptomatic domains that can be related to pathway or network connectivity impairment.

Conclusion

We present evidence that combined structural and functional connectivity measures show significant decrease over the duration of a longitudinal study of Multiple Sclerosis. The combination of structural and functional connectivity along pathways implicated in specific domains of disability in MS resulted in a metric that was related to both disease status and functional decline. The SFCI demonstrated a significant difference between participants with MS and sex- and age- matched HC, as well as a significant two-year decline in participants with MS. Both baseline SFCI and the change over time were related to a measure of speed of processing after two years, and the change in SFCI was related to future MSFC. This indicates that combined structural and functional connectivity measures, tailored to domains of disability, can be sensitive biomarkers for the study of neurologic disease.

ROC curves for measure 1 (M1) for progression scenario 1.

(TIF)

Click here for additional data file.

Progression scenario 1.

Area under the ROC curve for measures 1–3 (M1, M2, and M3), nine time points.

(TIF)

Click here for additional data file.

Progression scenario 2.

Area under the ROC curve for measures 1–3 (M1, M2, and M3), nine time points.

(TIF)

Click here for additional data file.

Development of SFCI.

(DOCX)

Click here for additional data file.

30 Nov 2020

PONE-D-20-32068

Evaluation of a connectivity-based imaging metric that reflects functional decline in Multiple Sclerosis

PLOS ONE

Dear Dr. Koenig,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

In addition to the comments from the Reviewers, please also address the following:

- Carefully check all references, as at least one is improperly formatted/referenced (ref. 34)

Please submit your revised manuscript by Jan 14 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Niels Bergsland

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. We note that you have indicated that data from this study are available upon request. PLOS only allows data to be available upon request if there are legal or ethical restrictions on sharing data publicly. For more information on unacceptable data access restrictions, please see http://journals.plos.org/plosone/s/data-availability#loc-unacceptable-data-access-restrictions.

In your revised cover letter, please address the following prompts:

a) If there are ethical or legal restrictions on sharing a de-identified data set, please explain them in detail (e.g., data contain potentially sensitive information, data are owned by a third-party organization, etc.) and who has imposed them (e.g., an ethics committee). Please also provide contact information for a data access committee, ethics committee, or other institutional body to which data requests may be sent.

b) If there are no restrictions, please upload the minimal anonymized data set necessary to replicate your study findings as either Supporting Information files or to a stable, public repository and provide us with the relevant URLs, DOIs, or accession numbers. For a list of acceptable repositories, please see http://journals.plos.org/plosone/s/data-availability#loc-recommended-repositories.

We will update your Data Availability statement on your behalf to reflect the information you provide.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Partly

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: No

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This very interesting paper looks at pathway-based MRI measures of structural and functional connectivity to construct the motor and cognition dysfunction by a composite imaging metric in MS. The topic is of high interest, given the lack of knowledge on relations between structure and function in relation to clinical deficits. Results are related to disability and cognitive dysfunction. Technically the setup is alright, but the descriptive indicators are too strict, headings and theirs order are confusing. Thresholds for the DTI analysis are missing which effects the correlates between the functioning. Hypothesis is misleading. MS has more complex relationships between pathways and functioning than AD. So, the selected pathway may not represent the memory domain functioning only in MS. Likewise, the current study also showed SDMT correlation as well, which is one the most common cognitive test in MS to indicate the processing speed functioning. The recommendations regarding the correction of these descriptions are attached below.

Abstract:

- Well-written

- Add MS cohort

- Perhaps indicate “2-year longitudinal study at four month intervals” info.

Introduction

- Line 46-48 “We reported reduced resting state functional magnetic resonance imaging (rsfMRI) between the bilateral primary motor cortices in patients with MS as compared to controls” is wrong expression. Perhaps rephrase to “reduced connectivity assessed by rsfMRI”?

- Line 48-50 “A follow-up study focused on both structural and functional connectivity, showing that diffusion tensor imaging (DTI) measures in the transcallosal motor pathway are correlated with rsfMRI of the primary sensorimotor cortices (SMC) in MS” How was this correlation? Perhaps indicate “The change of DTI and rs-fMRI measures were positively correlated over …”

- Line 69-72 “The cognitive component of the MSFC is based on the Paced Serial Addition Test (PASAT) [14], or, more recently, the Symbol Digit Modalities Test (SDMT). The metric proposed here was constructed as an analog to these composite measures of neurologic deficit.” Is this indicating the current study? This is very confusing. In the method section, different kind of tests are existed, but no PASAT.

- Line 73-75 “We focus on two common domains of disability in MS, motor function and memory,..” Why memory only? Several cognitive tests are given in the method and results sections. Conclusion also does not indicate this info. Perhaps the term was going to be “cognition” or “cognitive function”? In addition, even if the main idea was to track the memory, PASAT, BVMT and etc would have been chosen to specify, but not SDMT. SDMT more indicates processing speed.

On the other hand, PCC-AMLT is a well-known memory pathway in AD. However, MS has more complicated pathways that are presented by several cognitive dysfunction in it. Based on the results of the current study, it seems that this pathway is related with memory (COWAT) and processing speed (SDMT) domains. Therefore, instead of memory only, representing as “motor and cognitive functions” would be better.

- Line 73-75: The hypothesis is poorly introduced. Referring the previous work does not help to understand the aim. The citation for the ROI selection for the specific domain should be explained in previous paragraph. Than the exact functions and pathways that are going to be investigated in this study should be described in the hypothesis-paragraph. Such as “therefore, we investigated this this this pathway and its relationship with this this functions using SFCI…. over 2 years..”??

Theory

- Line 89-90. Why the metric will decrease with increased disability? It is a fact that both adaptive and maladaptive responses can occur in MS.

- Line 99. Is the “sample means” indicates the mean of the global connectivity for each patient? The equation should be explained better. Such as for Eq.2: fcpop indicates global or individual functional connectivity of healthy controls, �cpop represents the standard deviation of global or individual functional connectivity of healthy controls and Zmotor indicates the effect size of the motor SFC…

- Line 106. Please introduce the terms in the Eq. 3. What kind of notation is the L,R etc.

- Line 109. Again please explain the notation.

Materials and methods:

- Line 118. The term “Data acquisition” sounds like collecting imaging data etc... Perhaps reword as “Participants”?

In this section, it is written that the study includes 25 MS and 12 HC but in the abstract it is written 20 MS and 9 HC. Indication of the only final numbers of the participant (after withdraw) would be better to follow the article. Another way to describe better would be adding the inclusion/exclusion criteria’s and writing the final number of participants was determined based on these criteria’s following neuropsychological examination given in results.

Ethical committee number and chair should be written.

- Sample size is too strict. Is the Power analysis done? If it is not within the acceptable range the calculations should be repeated by applying something like Bootstrapping.

- Line 139. Table 1. SDMT is rather processing speed. Processing speed and working memory functioning should not be combined under the same section titled as “memory”. SDMT and D-KEFS do not represent the components of the memory domains and rather PASAT would have been involved in the neuropsychological tests to show the working memory functioning as described in the introduction.

- Line 148: Numerical indication of the slices would be easier to read: 176 axial slices.

- Line 173. Only “MRI post- processing” heading misleads the readers. This part includes both pre and post processing for the current version. Perhaps the heading would be “Processing of the MRI data”?

Other MRI Measures

- Line 200-203. Please specify the lesion filling procedures/tolls etc. more detailed.

- Line 249: What are the exact values for the thresholding of the structural connectivity construction? FOD cut off, number of fibers etc.?

Results

- Line 335: Table 2 should include the lesion volume information.

- Descriptive statistics that are represented in the Table3&4 are hard to follow in the text. Combining the functional and structural connectivity sections would be better for following the results in the tables that belong to both sections.

- Line 417-418: “In the interest of space, only behavioral measures that showed significant relationships are included.” Which relationship? Did you mean that only those that show a significant correlation between at least one of the SFCI components and one of the neuropsychological tests are given in the table?

- Line 431-446: Instead of the terms “positively/negatively”, explaining like “lower performance by SDMT or etc related with the lower/stronger/higher connectivity” may be easier to read.

Discussion

-Line 448-449: “This work demonstrates that a combined structural and functional connectivity metric can be a sensitive tool for identification of neurological disease” there is no other type of neurological disease that shows different patterns etc than MS or identify MS. So this sentence may be confusing. Rather it can be “This work demonstrates that a combined structural and functional connectivity metric can be a sensitive tool for identification of clinical impairment in neurological disease such as MS.”

-Methodological limitations must be discussed

-Line 497-500: I respectfully disagree with the authors that normalization is not necessary to compensate bias field effect (i.e noise) either it is region based or not. Before the calculation of the ROIs, registration to a standard atlas is mandatory which the current study also involved registration. This issue should be discussed more. Perhaps, exemplify the order of the analysis for this study compared to other studies.

Reviewer #2: The authors propose a composite MRI index, made out from MRI connectivity metrics relative to motor and cognitive domains, to relate to functional status and its decline in MS, tested in a 2-year longitudinal dataset.They focus on the transcallosal motor pathway connectivity and the connectivity mediated by the posterior cingulum bundle.

The authors theorize three different models to obtain the index and they test them both on simulated and real data.

line 150 : 256 x 128 might be a matrix error? ( siemens has this 50% distance factor - or gap- of the voxel on MPRAGE sequences in the direction of acquisition , but this does not change the matrix). I suppose also axial acquisition has to be verified, because 176 slices is typical of sagittal acquisitions.

liines 148-171 : please keep same order in the parameter description of each sequence (e.g. slice number,, thickness, FOV, matrix, voxel size etc). Also please report the acquisition duration, confirm that this is the order of acquisition and describe experimental setting , paradigm , pre-scan instructions and training.

line 172 MRI data analysis:: please provide all tools that you use in every step i.e. to apply Gaussian filter, fitting the boxcar model, the LV and on. I suppose that this can be provided as supplementary info. Also more info is needed for the experimental set, e.g. bite bar, physiological parameter acquisition. This is to ameliorate reproducibility.

line 219-220: How did you manually modify ROIs, e.g. by erosion of the voxels found in CSF?

line 230-242 : please add details for registration and transformation of images and refer to different spaces in a more explicit way: individual space can be functional , DTI, T13D or FLAIR. What kind of interpolation did you use to apply a transformation matrix to a flat ROI?

Fig3 : the 3D glass brain image appears blurred and does not permit a good evaluation of the localization.

line 265: by "cluster size 60" do you mean cluster size threshold of 60 voxels?please rephrase.

line 475-476: some grammar error, please rephrase

line 479 : rather than finding it is a missed relationship or no finding.

Conclusion:

This study can be included among the efforts in the advanced neuroimaging in neurological disease to identify connectivity biomarkers of disease status and progression. The small sample size may determine type I statistical error, as might be the case of the discussed missing PCC-AMTL association to memory. It is comprehensible that producing data for a power analysis at this point is too much to ask, therefore the value of this study can only be considered as in an exploratory study. It is valuable however for the longitudinal design, the right choice to focus on specific function-structure target, the sane methodology and for the correct admission and description of the limits. It is a valid scientific contribution in the search of viable imaging targets to obtain biomarkers that can translate into clinic, after validation in an adequately large study, so it should be considered as such.

It would be desired the availability, when published, of anonymized metadata of the group analysis, for example z-score maps and demographic, clinical and the other metrics that have been used, upon which to make a power analysis by anyone interested: itwould be an important addition for the usefullness of the paper and would fulfill the journal's policy and mission.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Nikolaos Petsas

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

25 Feb 2021

Thank you to the reviewers for your thorough and insightful comments on our manuscript. We think the manuscript is stronger and more readable with your suggestions. Below, each reviewer comment is numbered, with our response following (indented).

Reviewer #1:

Abstract

1. Add MS cohort

The abstract was modified so that the description of the cohort is earlier in the paragraph.

2. Perhaps indicate “2-year longitudinal study at four month intervals” info.

This information has been added to the abstract. Please note that we corrected an error in the description of the study visits – the first year included four visits at 4 month intervals, and the second year included two visits spaced at 6 month intervals. This has been corrected in the text.

Introduction

3. Line 46-48 “We reported reduced resting state functional magnetic resonance imaging (rsfMRI) between the bilateral primary motor cortices in patients with MS as compared to controls” is wrong expression. Perhaps rephrase to “reduced connectivity assessed by rsfMRI”?

This sentence was rephrased as suggested.

4. Line 48-50 “A follow-up study focused on both structural and functional connectivity, showing that diffusion tensor imaging (DTI) measures in the transcallosal motor pathway are correlated with rsfMRI of the primary sensorimotor cortices (SMC) in MS” How was this correlation? Perhaps indicate “The change of DTI and rs-fMRI measures were positively correlated over …”

We clarified that this was an inverse relationship.

5. Line 69-72 “The cognitive component of the MSFC is based on the Paced Serial Addition Test (PASAT) [14], or, more recently, the Symbol Digit Modalities Test (SDMT). The metric proposed here was constructed as an analog to these composite measures of neurologic deficit.” Is this indicating the current study? This is very confusing. In the method section, different kind of tests are existed, but no PASAT.

This sentence was modified to clarify that the SFCI is designed to be used in a similar manner to the MSFC – as a composite measure of deficit.

6. Line 73-75 “We focus on two common domains of disability in MS, motor function and memory,..” Why memory only? Several cognitive tests are given in the method and results sections. Conclusion also does not indicate this info. Perhaps the term was going to be “cognition” or “cognitive function”? In addition, even if the main idea was to track the memory, PASAT, BVMT and etc would have been chosen to specify, but not SDMT. SDMT more indicates processing speed.

On the other hand, PCC-AMLT is a well-known memory pathway in AD. However, MS has more complicated pathways that are presented by several cognitive dysfunction in it. Based on the results of the current study, it seems that this pathway is related with memory (COWAT) and processing speed (SDMT) domains. Therefore, instead of memory only, representing as “motor and cognitive functions” would be better.

We used the term “memory” based on the primary function of the PCC-AMTL pathway. However, the reviewer makes a good point – in our previous work we found that PCC-AMTL connectivity was related to both BVMT and SDMT in MS, so is not exclusively memory-related. We use the more general term “cognitive dysfunction” repeatedly throughout the manuscript, and as a metric we hope the SFCI would capture multiple domains of cognitive dysfunction. We have modified the manuscript to use more general language, including renaming the memory component of the SFCI to the cognitive component of the SFCI.

7. Line 73-75: The hypothesis is poorly introduced. Referring the previous work does not help to understand the aim. The citation for the ROI selection for the specific domain should be explained in previous paragraph. Than the exact functions and pathways that are going to be investigated in this study should be described in the hypothesis-paragraph. Such as “therefore, we investigated this this this pathway and its relationship with this this functions using SFCI…. over 2 years..”??

We reconfigured the Introduction to, hopefully, flow more naturally to our hypothesis. We moved the EDSS/MSFC discussion to the second paragraph, to give examples of other composite metrics as we explain the development of the SFCI, an imaging-based composite metric. We also added verbiage to motivate the transcallosal motor pathway and explicitly stated the expected relationship between pathway and behavioral measures.

Theory

8. Line 89-90. Why the metric will decrease with increased disability? It is a fact that both adaptive and maladaptive responses can occur in MS.

The reviewer is correct that compensatory mechanisms can result in increased connectivity between some regions involved in a task with measured deficits, and there is substantial literature on these results. However, the pathways probed here are monosynaptic and the relationship between connectivity to these regions and clinical disability has been reported as inversely related (i.e. increased disability corresponds to decreased connectivity in these pathways). Based on prior findings, we expect this measure, as constructed, to decrease with disability. That said, as constructed, the metric will decrease with decreased connectivity, which is what we meant to say. We have modified the text accordingly and are grateful to the reviewer for pointing this out.

9. Line 99. Is the “sample means” indicates the mean of the global connectivity for each patient? The equation should be explained better. Such as for Eq.2: fcpop indicates global or individual functional connectivity of healthy controls, �cpop represents the standard deviation of global or individual functional connectivity of healthy controls and Zmotor indicates the effect size of the motor SFC…

Notation for the equation has been added.

10. Line 106. Please introduce the terms in the Eq. 3. What kind of notation is the L,R etc.

Notation for the equation has been added.

11. Line 109. Again please explain the notation.

Notation for the equation has been added.

Materials and methods

12. Line 118. The term “Data acquisition” sounds like collecting imaging data etc... Perhaps reword as “Participants”?

We added a section labeled “Participants” under the “Data acquisition” heading. We intended the “Data acquisition” heading to include the subsections of “Clinical and cognitive evaluation” and “MRI acquisition,” but agree with the reviewer that without a section labeled “Participants,” the intention of the “Data Acquisition” heading is not clear.

13. In this section, it is written that the study includes 25 MS and 12 HC but in the abstract it is written 20 MS and 9 HC. Indication of the only final numbers of the participant (after withdraw) would be better to follow the article. Another way to describe better would be adding the inclusion/exclusion criteria’s and writing the final number of participants was determined based on these criteria’s following neuropsychological examination given in results.

To clarify the sample size, we added the following sentence directly after stating the total enrollment: “Of this sample, twenty participants with MS and 9 healthy controls were included in the final data analysis (see Results: Sample Description for details).”

14. Ethical committee number and chair should be written.

The protocol IRB number is now included. The chair of the Cleveland Clinic IRB is a rotating position that may be different upon publication. Therefore, rather than including a specific name, we elected to include the Federalwide Assurance number for the Cleveland Clinic IRB. This is now included.

15. Sample size is too strict. Is the Power analysis done? If it is not within the acceptable range the calculations should be repeated by applying something like Bootstrapping.

Yes, a power analysis was completed during the project development phase of this study. This project was funded by the NMSS, and the sample sizes proposed in that grant application were based on a power analysis derived from preliminary data from prior studies that are discussed in the manuscript. The results we present are based on the data collected for that study.

16. Line 139. Table 1. SDMT is rather processing speed. Processing speed and working memory functioning should not be combined under the same section titled as “memory”. SDMT and D-KEFS do not represent the components of the memory domains and rather PASAT would have been involved in the neuropsychological tests to show the working memory functioning as described in the introduction.

We modified the manuscript to focus on cognitive function rather than memory, as suggested by the reviewer in point 6.

17. Line 148: Numerical indication of the slices would be easier to read: 176 axial slices.

This has been modified.

18. Line 173. Only “MRI post- processing” heading misleads the readers. This part includes both pre and post processing for the current version. Perhaps the heading would be “Processing of the MRI data”?

The heading has been changed to “MRI data processing.”

Other MRI Measures

19. Line 200-203. Please specify the lesion filling procedures/tolls etc. more detailed.

No lesion filling was done. The automatic algorithm simultaneously segments brain parenchyma, total brain contour, and lesions. As described in the reference [48], the core of algorithm is iterative common mode (ICM) algorithm with classes for background, brain, and MS lesions. We clarified that a more detailed description of the lesion analysis can be found in reference [48].

20. Line 249: What are the exact values for the thresholding of the structural connectivity construction? FOD cut off, number of fibers etc.?

For the structural connectivity analysis, there is no threshold applied. A probability density map is produced from the tracking methodology, and the mean diffusivities, as described in equation 1, are produced with no thresholding.

Results

21. Line 335: Table 2 should include the lesion volume information.

Lesion volume information was added to Table 2.

22. Descriptive statistics that are represented in the Table3&4 are hard to follow in the text. Combining the functional and structural connectivity sections would be better for following the results in the tables that belong to both sections.

We appreciate the feedback on the structure of the manuscript. To make the tables and text follow more naturally, we elected to categorize the results by analysis. There are now separate sections for “Group differences,” “Impact of time,” and “Imaging and behavioral measures.” Results for all imaging measures are presented in each section, mirroring the presentation in the tables and figures. We hope that this makes the manuscript more readable.

23. Line 417-418: “In the interest of space, only behavioral measures that showed significant relationships are included.” Which relationship? Did you mean that only those that show a significant correlation between at least one of the SFCI components and one of the neuropsychological tests are given in the table?

The reviewer is correct. We clarified this in the text as follows: “Tables 5 and 6 show Pearson correlation coefficients between imaging and behavioral measures in participants with MS. In the interest of space, only behavioral measures that showed a significant relationship to at least one imaging measure are included in the tables.”

24. Line 431-446: Instead of the terms “positively/negatively”, explaining like “lower performance by SDMT or etc related with the lower/stronger/higher connectivity” may be easier to read.

This section has been revised to more clearly state the relationships between the variables.

Discussion

25. Line 448-449: “This work demonstrates that a combined structural and functional connectivity metric can be a sensitive tool for identification of neurological disease” there is no other type of neurological disease that shows different patterns etc than MS or identify MS. So this sentence may be confusing. Rather it can be “This work demonstrates that a combined structural and functional connectivity metric can be a sensitive tool for identification of clinical impairment in neurological disease such as MS.”

This sentence has been modified as suggested by the reviewer.

26. Methodological limitations must be discussed.

A paragraph discussing methodological limitations has been added near the end of the discussion section.

27. Line 497-500: I respectfully disagree with the authors that normalization is not necessary to compensate bias field effect (i.e noise) either it is region based or not. Before the calculation of the ROIs, registration to a standard atlas is mandatory which the current study also involved registration. This issue should be discussed more. Perhaps, exemplify the order of the analysis for this study compared to other studies.

With all due respect to the reviewer, the authors stand by the statement that the methodologies employed here don’t require registration to an atlas or common space. All ROI’s and resultant pathways are, in the final determination, determined within subject and in the subject’s native space. For the PCC/AMTL ROI’s, an anatomic landmark is used as a starting point. However, the final ROI, which is very small in relation to the starting region, is determined functionally, within subject in native space. We have clarified that in the paragraph referenced.

Reviewer #2:

1. line 150 : 256 x 128 might be a matrix error? ( siemens has this 50% distance factor - or gap- of the voxel on MPRAGE sequences in the direction of acquisition , but this does not change the matrix). I suppose also axial acquisition has to be verified, because 176 slices is typical of sagittal acquisitions.

The authors apologize that it was not made clear that this was an axial 3D MPRAGE. The text has been modified to reflect that. The acquisition matrix was 256 in frequency, 128 in phase and 176 in the slice direction. It was a single slab, 3D acquisition, so there is no distance factor issue. The images were reconstructed to 256x256, but the acquired matrix and slices were as stated in the manuscript.

2. liines 148-171 : please keep same order in the parameter description of each sequence (e.g. slice number,, thickness, FOV, matrix, voxel size etc). Also please report the acquisition duration, confirm that this is the order of acquisition and describe experimental setting , paradigm , pre-scan instructions and training.

The scan descriptions have been made more uniform, acquisition durations have been added to each scan, and the order of acquisition has been confirmed. Two paragraphs were added prior to the scan descriptions that include information on the tapping paradigm and participant training.

3. line 172 MRI data analysis:: please provide all tools that you use in every step i.e. to apply Gaussian filter, fitting the boxcar model, the LV and on. I suppose that this can be provided as supplementary info. Also more info is needed for the experimental set, e.g. bite bar, physiological parameter acquisition. This is to ameliorate reproducibility.

We added more information on tools throughout the “MRI data processing” section. We also added more information on the finger tapping task and bite bar under “MRI acquisition.” Finally, additional information on physiological monitoring was added under “MRI data processing.”

4. line 219-220: How did you manually modify ROIs, e.g. by erosion of the voxels found in CSF?

This sentence has been modified to clarify the procedure for ROI modification: “ROIs were visually inspected and, if necessary, manually eroded to ensure all voxels were located in gray matter and not in intragyral CSF (Fig. 2a, lower right).”

5. line 230-242 : please add details for registration and transformation of images and refer to different spaces in a more explicit way: individual space can be functional , DTI, T13D or FLAIR. What kind of interpolation did you use to apply a transformation matrix to a flat ROI?

Additional details of image registration and transformation have been added to the first paragraph under “Seed region definition.” We used nearest neighbor interpolation for the ROI transforms, and this information is now included. References to specific image spaces have been clarified.

6. Fig3 : the 3D glass brain image appears blurred and does not permit a good evaluation of the localization.

Figure 3 was redone to provide a sharper image and allow clearer localization.

7. line 265: by "cluster size 60" do you mean cluster size threshold of 60 voxels? please rephrase.

This was rephrased to state ”(p < 1.0×10-4, cluster size threshold = 60 voxels).”

8. line 475-476: some grammar error, please rephrase

This sentence has been rephrased as: “A possible explanation is that the prior study used a DTI acquisition that had higher spatial resolution and was tailored to target the PCC-AMTL pathway.”

9. line 479 : rather than finding it is a missed relationship or no finding.

This sentence has been revised to say “…future work is required to clarify the lack of relationship seen here.”

10. It would be desired the availability, when published, of anonymized metadata of the group analysis, for example z-score maps and demographic, clinical and the other metrics that have been used, upon which to make a power analysis by anyone interested: it would be an important addition for the usefullness of the paper and would fulfill the journal's policy and mission.

The Cleveland Clinic Institutional Review board has strict requirements for sharing individual-level data, particularly demographic and clinical information. We are currently seeking approval from the Cleveland Clinic IRB to share a de-identified dataset. If accepted, the authors agree to share group analysis results and other metadata, as permitted by our institution and within ethical and legal guidelines. We propose to share these data through figshare.

Submitted filename: Response_to_reviewers.docx

Click here for additional data file.

10 Mar 2021

PONE-D-20-32068R1

Evaluation of a connectivity-based imaging metric that reflects functional decline in Multiple Sclerosis

PLOS ONE

Dear Dr. Koenig,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

As you will see below, most of the remaining comments raised from Reviewer are matters of preference. As such, I leave you some leeway in deciding which to specifically address. However, please justify your responses, accordingly. I do recommend, though, that you update some of the references, as suggested.

Please submit your revised manuscript by Apr 24 2021 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: http://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols

We look forward to receiving your revised manuscript.

Kind regards,

Niels Bergsland

Academic Editor

PLOS ONE

Journal Requirements:

Please review your reference list to ensure that it is complete and correct. If you have cited papers that have been retracted, please include the rationale for doing so in the manuscript text, or remove these references and replace them with relevant current references. Any changes to the reference list should be mentioned in the rebuttal letter that accompanies your revised manuscript. If you need to cite a retracted article, indicate the article’s retracted status in the References list and also include a citation and full reference for the retraction notice.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: No

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: Overall the review was performed well, the paper has certainly improved. Some remaining points:

Although the authors improved readability of the paper a lot, the references are too old. Cognition in MS is highly topic in recent years and I believe there are more updated relevant works in the literature.

Abstract:

- 3 digits after decimal point is enough for the p value description and 2 digits for r values.

Materials and methods:

- Line 176: Please rephrased as “176 axial slices and 0.94 mm slice thickness”

-“Acquisition time (TA)” would be better instead of “Scan length”

-Please provide abbreviations such as inversion time (TI), TE, TR ….

Results:

In general each table and figure caption should include the abbreviations.

-Figure 4: Please provide all abbreviations MS, SMC… and descriptions like (A) shows sensorimotor cortex etc. in the figure caption.

Reviewer #2: All my questions have been successfully addressed. The only concern of mine remaining is that full data availability has not been definitely addressed

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: Yes: Arzu Ceylan Has Silemek

Reviewer #2: Yes: Nikolaos Petsas

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

22 Apr 2021

Thank you again to the reviewers for the time and effort applied to our manuscript. Please find our responses to your comments below.

Reviewer #1:

1. Although the authors improved readability of the paper a lot, the references are too old. Cognition in MS is highly topic in recent years and I believe there are more updated relevant works in the literature.

Updated references have been added. Please note that using the “track changes” function to note reference changes made the manuscript very difficult to read. Rather, the new references are in red type.

Abstract

2. 3 digits after decimal point is enough for the p value description and 2 digits for r values.

This was modified.

Materials and methods

3. Line 176: Please rephrased as “176 axial slices and 0.94 mm slice thickness”

This was modified for each scan description.

4. “Acquisition time (TA)” would be better instead of “Scan length”

This was modified for each scan description.

5. Please provide abbreviations such as inversion time (TI), TE, TR ….

These definitions were added under Scan 1.

Results

6. In general each table and figure caption should include the abbreviations.

Abbreviations have been added to the figures and tables. For the sake of space, a number of captions refer the reader to abbreviations under Table 3.

7. Figure 4: Please provide all abbreviations MS, SMC… and descriptions like (A) shows sensorimotor cortex etc. in the figure caption.

Abbreviations and additional description have been added for this figure.

Reviewer #2:

1. The only concern of mine remaining is that full data availability has not been definitely addressed.

As previously mentioned, the Cleveland Clinic has strict requirements for data sharing, including review by the Institutional Review Board, legal department, and cybersecurity. We hoped to receive full approval by the resubmission deadline, but still await approval from the Cleveland Clinic legal department. We expect full approval by the end of April 2021. We assure the reviewers and editors that we have moved as quickly as possible to gain approval, and apologize for the delay. The data supporting this work is uploaded to Figshare. As soon as approval is received, the data will be published publically and we will notify the Editor. In the meantime, we have made a private link available to the reviewers and editors:

https://figshare.com/s/db2f3272926c015412fb

Submitted filename: Response_to_reviewers.docx

Click here for additional data file.

26 Apr 2021

Evaluation of a connectivity-based imaging metric that reflects functional decline in Multiple Sclerosis

PONE-D-20-32068R2

Dear Dr. Koenig,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Niels Bergsland

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

31 May 2021

PONE-D-20-32068R2

Evaluation of a connectivity-based imaging metric that reflects functional decline in Multiple Sclerosis

Dear Dr. Koenig:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Niels Bergsland

Academic Editor

PLOS ONE