1 Supplementary Methods
Inclusion and exclusion criteria for participants
Diagnosis of PP or SP MS was made according to the revised McDonald’s criteria 2010 [1].
Inclusion criteria were the following: i) age between 25 and 65 years, ii) expanded disability
status scale (EDSS) score below 6.5 at screening, iii) if treated, patients had to be on a stable
treatment for the previous year, iv) no history of relapses in the previous year. Exclusion
criteria included: current or past history of major hematological, renal, hepatic, psychiatric or
neurological disease (other than MS for the patient group) and contraindications to MRI. From
the initial sample of 74 subjects, 4 patients were excluded because of failure in MRI processing
(i.e. functional-structural co-registration) and 13 (1 healthy control, 12 patients) were excluded
after motion scrubbing because more than 25% of volumes in their scans had a framewise
displacement above the threshold of 0.5 mm (see ”Motion” section), resulting in a final sample
of 57 subjects.
MRI sequences parameters
T2-weighted sequence: voxel size 0.5x0.5x3 mm, repetition time (TR) 8000 ms, echo time
(TE) 95ms. 3D T1-weighted sequence: voxel size:0.8x0.8x0.8 mm, TR 3000 ms, TE 2.47 ms,
flip angle 7°, inversion time 1000 ms). Single shot gradient echo planar imaging sequence:
voxel size: 2.1x2.1x2.1 mm, TR 1000 ms, TE 35 ms, multi-band acceleration factor 7, flip
angle 60°, 400 volumes. During rs-fMRI, subjects were asked to rest with their eyes closed.
Motion
We identified subjects that displayed too much motion characterized by their framewise
displacement values [2]. Those with FDPower > 0.5 mm in more than 25% of the volumes [3]
were excluded from the analysis. In the final sample, the average FD did not differ between
patients with PMS and HC at the baseline (t(55)= -1.37, p=0.18) and at FU (t(33)= -1.52,
p=0.14, see Table 1). In all other subjects, scrubbing was applied with contaminated volumes
(FDPower > 0.5mm) censored and replaced by the spline interpolation of the previous and
following volumes. Motion frames were excluded from the computation of iCAPs temporal
properties.
Total activation and iCAPs extraction
The iCAPs analysis was performed using a publicly available implementation (https:
//c4science.ch/source/iCAPs/). First, a subsequent denoising and hemodynamic
deconvolution was applied using total activation (TA) to the single subject’s native space rsfMRI preprocessed time-courses, both at baseline and FU. This method allows to analyze the
fMRI signal independently from the hemodynamic response, employing the changes of signal
induced by neuronal activity. [4]. For each voxel v at each moment t, the time course results
as y[v, t] = x[v, t] + ε, where the BOLD activity-related signal x can bee described by x[v, t] =
Σtτ =0 s[v, t]h(t − τ ), and s describes the changes induced by neural activity. The activity related
1
signal x can be retrieved through a spatio-temporal regularization as follows:
x̂ = argmin
x
n1
2
ky − xk22 + RT (x) + RS (x)
o
.
The temporal regularization can be expressed as:
RT (x) =
Nx
X
λT (v)
v=1
Nt
X
|∆L x[v, t]|
t=1
where ∆L is the discrete version of the differential operator L and λT (v) is the regularization parameter for voxel (v).
The spatial regularization is given by:
RS (x) =
Nt
X
t=1
λS (t)
Nx X
X
v=1
u∈N (v)
x[v, t] − x[u, t]
2
1/2
where N (v) is a set of neighor voxels around v and λS (v) is the spatial regularization
parameter at a given timepoint t. For more details about the implementation of the TA paradigm
we refer to [5] and [4].
After running TA, the sparse innovation signal (or transients) are extracted by computing
the temporal derivative of the activity-inducing signals. Afterwards, the significant transients
are extracted by taking only a fraction across brain volumes following a two-step thresholding
procedure. The temporal thresholding includes a phase-randomization, from which the cut-off
thresholds using the lowest 5th and highest 95th percentile of its corresponding innovation signal are extracted. Following the temporal thresholding, the frames with at least 5% of the total
number of gray matter voxels are selected to undergo temporal clustering using k-means. For
more details of the iCAPs extraction and the optimization of the thresholding, we refer to [6].
Thereafter, rs-fMRI frames corresponding to transients were normalized to the MNI coordinate space and concatenated across all subjects. The resulting matrix of dimension “number
of voxels x transients” underwent temporal k-means clustering, to obtain brain patterns that
are simultaneously transitioning, i.e. iCAPs. Consensus clustering was used to determine the
optimum number of clusters. The time-courses “assigned” to each iCAP were Z-scored within
each subject and thresholded at |z| > 1 to obtain binary data and be able to define the “active”
time-points. The choice of threshold was motivated by previous works that implemented TA
and iCAPs framework [6]–[8]. Lastly, the activity inducing time-courses for each subject were
obtained for all iCAPs using spatio-temporal transient-informed regression[7]. ICAPs duration is defined as the percentage of the active time-points with respect to the total time-points
for each iCAP. Couplings and anti-couplings for each pair of iCAPs were derived using the
Jaccard index and defined as follow: i) the coupling: number of time-points during which the
two iCAPs were both co-(de)active divided by the number of time points during which at least
one of them was (de)active, and ii) the anti-coupling: number of time-points during which the
two iCAPs presented opposite sign of activation, divided by the number of time points during
which at least one of them was (de)active. For the computation of duration and couplings, only
activation blocks of two or more time-points have been considered.
Choice of the optimal number of clusters
2
Consensus clustering [9] was used to find the optimal number of clusters in the concatenated
significant innovation frames. This method includes a subsampling of the data and multiple
runs of the clustering algorithm. The consistency of each frame to be grouped in a similar
cluster is monitored through a consensus metric. The consensus clustering matrices for each
evaluated K are represented in Figure SI1 for the baseline sample. The distribution of each
cluster consensus is plotted in Figure SI2. Based on the highest median consensus we chose a
K= 13 [9]. ICAPs determined by less than 5 subjects were discarded from further analysis.
Figure SI 1: Consensus clustering matrices for different cluster values K = 11 to 21 for baseline
(57 subjects) data. Axes represents frames number. Values in the matrix range from 0 to 1,
indicating the reproducibility of the sampling across multiple runs, with 1 being perfectly resampled at all times. Diagonal values are expected to be equal to 1 (the same frame indices will
always be clustered into the same group).
3
Figure SI 2: Clustering consensus as a measure of the stability of the observed iCAPs with
respect to different K (x axis) over multiple runs of the clustering
Analysis at follow-up
The TA analysis was applied to the 35 subjects at follow-up (FU). Then, the iCAPs retrieved from the baseline sample were fitted into the FU sample. This was done by matching
each significant innovation frame of subjects at FU to the closest iCAP obtained from baseline
using cosine distance, as a metric. After obtaining the cluster index for the innovation frames
at FU and their distance to each cluster center, the transient-informed regression step was repeated on FU data to obtain each subject’s iCAP time-courses and their corresponding temporal
properties.
Partial least squares correlation analysis
Partial least squares correlation (PLSC) analysis [10] procedure included: first, the [subjects x temporal properties] matrix X for brain variables (iCAPs temporal properties) and the
[subjects x behavioral variables] matrix Y for clinical ones were built, concatenating HC and
PMS patients. Then, their relationship was assessed by means of singular value decomposition
of the correlation matrix between X and Y, which resulted in latent components that indicate
multivariate patterns of brain-behavior correlation. Lastly, the significance and stability of the
components was assessed by permutation (1000 permutations) and bootstrap (800 bootstrap
samples) testing, respectively.
4
2 Supplementary Results
Contribution of each group to the number of frames selected for each iCAP
To evaluate whether the 2 groups (HC and MS pts) significantly differed in the contribution
of frames to each iCAP, we retrieved, for each subject, the number of frames with which they
contributed to each iCAP and compared the two groups using the Wilcoxon rank sum test,
given the non normal distribution of the data. All p values resulted ¿ 0.05. Mean, median and
SD values for number of frames with which HC or MS pts groups contributed to each iCAP
are reported in the Table SI 4.
Frequency of cognitive impairment
Twenty-four (75%), 13 (68%) and 8 (42%) patients were classified as CI at baseline, in the
sub-sample (BS-Sub) of 35 patients at baseline and at FU, respectively. A chi-squared test did
not find significant differences in the number of CI patients between the BS-Sub and FU groups
(Figure SI 3 A). Performance at BS-Sub and FU timepoints were significantly correlated for
the three tests (Pearson correlation coefficients range: 0.46-0.56). (Figure SI 3 B-D).
Table SI 1: Clinical and demographic data at baseline of the drop-out subjects.
Age
Gender (females, males)
Education
Disease duration
EDSS (median; range)
SDMT raw score
SDMT Z-score
BVMT-R raw score
BVMT-R Z-score
CVLT-II raw score
CVLT-II Z-score
T25FW
9HPT
T2 LV (mL)
T1 LV (mL)
Grey Matter Volume (cm3 )
White Matter Volume (cm3 )
Brain Volume (cm3 )
aDMN duration
aDMN-ECN anti-coupling
HC (n=9)
46.2 ± 11.4
3; 6
15.6 ± 0.9
790.3 ± 31.4
720.4 ± 37.7
1510.7 ± 65.5
31.8 ± 5.4
0.23 ± 0.1
Drop-out subjects (n=22)
p-value‡
0.66
0.61
0.25
0.38
0.74
0.47
0.39
0.96
PMS patients (n=13) p-value†
49.5 ± 10.7
0.55
7; 6
0.40
16.2 ± 2.7
0.49
17.2 ± 11.8
0.50
5.5; 3.0-6.0
0.82
42.6 ± 15.0
0.27
−1.8 ± 1.5
16.8 ± 7.0
0.92
−2.0 ± 1.3
51.1 ± 15.0
0.19
−0.19 ± 1.5
7.3 ± 2.4
0.11
33.2 ± 9.1
0.64
10.0 ± 11.0
0.28
8.0 ± 7.1
0.22
747.2 ± 51.7
0.35
655.6 ± 40.1
0.80
1402.9 ± 86.2
0.69
24.8 ± 7.2
0.28
0.18 ± 0.1
0.14
All variables are expressed as mean ± standard deviation if not otherwise specified. ‡
unpaired t-test between HC who undergo FU and HC who dropped out. † unpaired t-test
between PMS patients who undergo FU and PMS patients who dropped out. EDSS: expanded
disability status scale; SDMT: symbol digit modalities test; BVMT-R: Brief Visuospatial
Memory Test–Revised; CVLT-II: CVLT-II California Verbal Learning Test Second Edition;
T25FW: timed 25-foot walk test; 9HPT: 9-hole peg test; T2 LV: lesion volume in T2-weighted
sequence; T1 LV: lesion volume in T1-weighted sequence.
5
Table SI 2: Clinical data of the subsample of subjects who underwent the follow-up analysis at
baseline (BS-Sub) and Follow-up.
BS-Sub (n=19) Follow-up (n= 19)
EDSS (median; range)
6; 1.0-6.5
5; 2.5-6.5
SDMT raw score
48.8 ± 15.7
51.3 ± 13.3
BVMT-R raw score
16.5 ± 9.9
19.8 ± 8.9
CVLT-II raw score
57.3 ± 11.6
58.7 ± 13.3
T25FW
9.8 ± 5.2
12.1 ± 15.8
9HPT
31.4 ± 11.4
34.2 ± 20.3
T2 LV (mL)
6.8 ± 9.8
7.0 ± 10.0
T1 LV (mL)
4.7 ± 6.9
5.5 ± 8.1
p-value‡
0.59
0.28
0.06
0.38
0.47
0.41
0.13
0.02
All variables are expressed as mean ± standard deviation if not otherwise specified. ‡ paired
t-test between BS-Sub and FU PMS patients. BS-Sub: subsample of 35 subjects at baseline;
EDSS: expanded disability status scale; SDMT: symbol digit modalities test; BVMT-R: Brief
Visuospatial Memory Test–Revised; CVLT-II: CVLT-II California Verbal Learning Test
Second Edition; T25FW: timed 25-foot walk test; 9HPT: 9-hole peg test; T1 LV: lesion
volume in T1-weighted sequence; T2 LV: lesion volume in T2-weighted sequence.
Validation analyses
To evaluate the robustness and reliability of our results, we performed three validation analyses, performing the k-means clustering using the optimal k=13 (as resulted from the Consensus Clustering) and the transient-informed regression steps:
1. on the baseline sample of 57 subjects to assess the robustness of the Clustering step.
The same iCAPs as in the original analysis were retrieved. Among iCAPs durations, the
aDMN duration resulted significantly reduced in patients with PMS, compared to HC
(F(1,52)=12.00, p= .001).
2. on the baseline sub-sample of 35 subjects (16 HC and 19 patients with PMS) who underwent FU. Nine out of the 11 iCAPs were retrieved from these analysis (the last 4 were
discarded as determined by less than 5 subjects) and represented in Figure SI 8. Among
iCAPs durations, the aDMN duration resulted significantly different between HC and
patients with PMS (F(1,30)=6.77, p= .014, reduced in PMS patients).
3. on a sub-sample of 40 subjects, balancing the number of HC (n=20) and patients with
PMS (n=20). Ten out of the 11 iCAPs were retrieved from these analysis (the last 3
were discarded as determined by less than 5 subjects). Among iCAPs durations, the
aDMN duration resulted significantly reduced in patients with PMS compared to HC
(F(1,35)=15.50, p¡ .001).
6
Figure SI 3: Number of patients performing below the cut-off scores of 1.5 SD using the
regression-based norms as in [11] at baseline (BS-Sub, n=19) and at the 1 year follow-up
(n=19), and correlations between the test performances at the two time points.
7
Table SI 3: Baseline iCAPs corresponding to the 17 networks of Yeo atlas [12], the Greicius
atlas [13] and the regions in the automated anatomical labeling atlas ([14]). Percentiles indicate
the fraction of voxels belonging to a network or region that has a z-score > 1.5. Only networks
and regions that have at least 25% surviving percentile are included in the list.
iCAP
1
2
3
Network Schaefer(%)
DefaultC (43,12)
Limbic2 (31,46)
Network Shirer (%)
Lobe
Basal Ganglia (56,31) Limbic
Limbic
Limbic
Limbic
Subcortical
Subcortical
Subcortical
Subcortical
Subcortical
Limbic
Subcortical
Limbic
Frontal
Frontal
Occipital
Temporal
Temporal
Occipital
SomMotB (77,87)
Auditory (89,05)
Central
SalVentAttnA (53,05) post Salience (42,88) Central
SomMotA (27,66)
Temporal
Temporal
Parietal
Temporal
Parietal
Temporal
Limbic
Parietal
Parietal
Limbic
Frontal
Frontal
Limbic
Limbic
Frontal
Frontal
VisPeri (87,77)
prim Visual (97,41)
Occipital
VisCent (60,94)
high Visual (69,93)
Occipital
DefaultC (28,1)
Occipital
Occipital
Occipital
Occipital
Occipital
Occipital
Occipital
Occipital
Occipital
Occipital
8
Region AAL (%)
Amygdala L (91,57)
Amygdala R (86,3)
Hippocampus L (86,14)
Hippocampus R (84,05)
Thalamus R (80,42)
Caudate L (73,7)
Caudate R (72,85)
Thalamus L (72,82)
Putamen L (71,82)
ParaHippocampal L (70,28)
Putamen R (63,4)
ParaHippocampal R (61,8)
Olfactory R (54,84)
Olfactory L (49,73)
Fusiform L (41,9)
Temporal Pole Sup L (35,5)
Temporal Pole Sup R (34,88)
Fusiform R (32,58)
Rolandic Oper L (89,44)
Rolandic Oper R (89,19)
Heschl R (85,55)
Heschl L (74,65)
SupraMarginal L (67,71)
Temporal Sup L (66,25)
SupraMarginal R (55,09)
Temporal Sup R (52,31)
Insula R (50)
Postcentral R (46,78)
Postcentral L (44,82)
Insula L (42,61)
Supp Motor Area R (37,58)
Frontal Inf Oper R (36,73)
Cingulum Mid L (32,88)
Cingulum Mid R (30,88)
Precentral R (30,41)
Supp Motor Area L (29,03)
Cuneus R (95,62)
Calcarine R (94,5)
Cuneus L (91,34)
Calcarine L (87,44)
Occipital Sup L (86,42)
Occipital Mid R (80,69)
Occipital Sup R (78,62)
Occipital Mid L (69,3)
Lingual L (65,19)
Lingual R (64,42)
Occipital Inf L (41,67)
Fusiform R (29,56)
mean z-score voxels
3,03
152
3,07
126
2,45
491
2,66
432
2,65
193
2,12
339
2,14
314
2,48
217
2,39
209
2,51
395
2,42
220
2,51
419
1,96
85
2,02
92
2,09
538
2,28
197
2,22
195
1,98
463
2,56
466
2,87
561
2,42
148
2,24
106
2,68
239
2,65
642
2,88
341
2,68
633
2,14
543
2,34
269
2,36
277
2,13
548
2,03
177
1,94
119
1,96
316
1,98
348
2,03
111
2,02
126
2,69
371
3,04
687
2,55
369
2,78
787
2,43
140
2,42
418
2,18
125
2,33
526
2,54
472
2,62
603
1,83
105
1,88
420
iCAP
4
5
6
7
8
Network Schaefer (%) Network Shirer (%)
Lobe
ContA (70,45)
RECN (83,27)
Frontal
SalVentAttnB (53,83) antSalience (45,45)
Parietal
ContB (46,41)
LECN (40,02)
Frontal
Visuospatial (33,2)
Frontal
post Salience (30,77) Frontal
Frontal
Frontal
Frontal
Parietal
Frontal
Frontal
Parietal
Frontal
Frontal
Frontal
DefaultA (47,3)
dorsalDMN (65,09)
Limbic
SalVentAttnB (44,43) antSalience (34,36)
Frontal
Frontal
Limbic
Frontal
Frontal
Frontal
Frontal
Frontal
SalVentAttnA (50,53) antSalience (42,34)
Limbic
SomMotB (27,47)
Auditory (33,06)
Limbic
SalVentAttnB (25,23)
Frontal
Frontal
Subcortical
Subcortical
Central
Frontal
Temporal
Central
Frontal
Temporal
Limbic
Frontal
Temporal
TempPar (75,35)
Language (78,83)
Parietal
DefaultB (38,16)
Temporal
Temporal
Parietal
Temporal
Temporal
Parietal
VisCent (67,65)
high Visual (82,47)
Occipital
DorsAttnA (52,55)
Occipital
Temporal
Occipital
Occipital
Occipital
Occipital
Temporal
Occipital
Occipital
9
Region AAL (%)
Frontal Mid R (87,95)
Parietal Inf R (78,72)
Frontal Inf Tri R (76,47)
Frontal Mid L (73,99)
Frontal Mid Orb R (66,19)
Frontal Inf Tri L (58,52)
Frontal Sup R (55,87)
Frontal Mid Orb L (51,27)
Parietal Inf L (46,61)
Frontal Inf Oper R (41,)
Frontal Sup L (39,85)
SupraMarginal R (38,45)
Frontal Sup Medial R (34,72)
Frontal Inf Oper L (34,58)
Frontal Sup Medial L (25,57)
Cingulum Ant R (89,64)
Frontal Sup Medial R (89,63)
Frontal Sup Medial L (89,02)
Cingulum Ant L (85,54)
Frontal Sup Orb Medial R (81,63)
Frontal Sup Orb Medial L (79,14)
Frontal Sup L (64,7)
Frontal Sup R (34,82)
Rectus L (26,77)
Insula L (87,33)
Insula R (85,36)
Frontal Inf Oper R (54,63)
Frontal Inf Tri R (53,33)
Putamen R (52,45)
Putamen L (51,89)
Rolandic Oper L (48,56)
Frontal Inf Oper L (46,78)
Heschl L (42,25)
Rolandic Oper R (37,68)
Frontal Inf Orb R (33,15)
Temporal Pole Sup L (31,71)
Cingulum Ant R (31,07)
Frontal Inf Tri L (28,94)
Temporal Pole Sup R (28,44)
Angular L (81,14)
Temporal Mid L (76,58)
Temporal Mid R (67,69)
Angular R (55,02)
Temporal Sup R (39,01)
Temporal Pole Sup L (30,63)
Parietal Inf R (25,07)
Occipital Inf R (98,47)
Occipital Inf L (91,67)
Temporal Inf R (47,23)
Occipital Mid R (46,53)
Occipital Mid L (45,45)
Fusiform R (37,37)
Lingual R (32,16)
Temporal Inf L (29,87)
Fusiform L (29,36)
Calcarine L (29)
mean z-score
2,85
2,48
2,44
2,55
2,44
2,46
2,27
2,45
2,05
1,9
1,99
2,34
2,21
1,85
2,1
3,45
2,82
3,15
3,46
2,64
2,57
2,15
1,98
2
2,82
2,94
2,43
2,24
1,76
1,79
2,03
2,01
1,81
2,05
2,17
1,99
1,82
2,26
1,91
2,63
2,82
2,84
2,51
2,66
1,98
2,36
3,59
3,01
2,82
2,34
2,21
2,41
2,77
2,55
2,6
2,25
voxels
1212
270
390
879
327
364
414
222
213
136
263
238
251
102
212
554
648
738
704
360
330
427
258
140
1123
927
177
272
182
151
253
138
60
237
241
176
192
180
159
241
1383
1280
252
472
170
86
258
231
894
241
345
531
301
463
377
261
iCAP Network Schaefer (%) Network Shirer (%) Lobe
9
ContC (70,11)
Precuneus (66,46)
Limbic
DefaultA (50,89)
ventralDMN (48,38) Limbic
dorsalDMN (37,96) Parietal
Parietal
Parietal
Parietal
Frontal
Frontal
Occipital
Limbic
Limbic
Occipital
Parietal
10
Limbic2 (74,67)
Temporal
Temporal
Limbic
Limbic
Temporal
Limbic
Limbic
Limbic
Temporal
Limbic
Temporal
11
Limbic1 (88,89)
Frontal
ContB (31,29)
Frontal
Frontal
Frontal
Frontal
Frontal
Frontal
Frontal
Frontal
Frontal
Frontal
Subcortical
Frontal
Subcortical
Region AAL (%)
Cingulum Post L (99,43)
Cingulum Post R (99,07)
Angular L (90,57)
Angular R (89,52)
Precuneus R (73,42)
Precuneus L (68,91)
Frontal Sup Orb Medial R (42,63)
Frontal Sup Orb Medial L (42,45)
Cuneus R (30,67)
Cingulum Mid R (27,15)
Cingulum Mid L (26,74)
Cuneus L (26,49)
Parietal Inf R (26,24)
Temporal Pole Mid R (91,36)
Temporal Pole Mid L (86,98)
Amygdala R (67,81)
Amygdala L (63,25)
Temporal Pole Sup R (53,67)
ParaHippocampal R (47,2)
ParaHippocampal L (42,35)
Hippocampus R (41,25)
Temporal Inf L (36,45)
Hippocampus L (32,98)
Temporal Inf R (32,17)
Frontal Sup Orb R (95,1)
Frontal Sup Orb L (88,67)
Rectus L (86,62)
Rectus R (82,84)
Frontal Mid Orb L (78,98)
Frontal Mid Orb R (75,3)
Olfactory L (69,19)
Olfactory R (68,39)
Frontal Sup Orb Medial L (56,35)
Frontal Inf Orb L (53,96)
Frontal Sup Orb Medial R (49,21)
Caudate L (42,83)
Frontal Inf Orb R (36,86)
Caudate R (35,5)
mean z-score voxels
4,02
174
4,25
106
2,47
269
2,75
410
3,26
721
3,15
614
1,87
188
1,85
177
2,11
119
3,15
306
2,82
257
2,27
107
2,29
90
3,11
476
2,77
354
2,08
99
1,98
105
2,09
300
2,69
320
2,72
238
2,27
212
3,25
565
1,98
188
3,32
609
3,25
408
3,18
407
2,76
453
2,94
367
3,24
342
3,16
372
2,19
128
2,14
106
2,23
235
2,41
436
2,33
217
1,91
197
2,24
268
1,92
153
Table SI 4: Median, mean and SD of the number of frames each group contributed to each
iCAP and p values of the resulting Wilcoxon rank sum test
1. Deep GM
2. Auditory/Sensory-Motor
3. Primary Visual
4. ECN
5. aDMN
6. SAL
7. TempPar/LAN
8. Secondary Visual
9. pCun/pDMN
10. Amy/TempPole
11. OFC
Median HC Median MS pts
47.00
57.50
52.00
48.50
48.00
51.50
46.00
49.50
45.00
43.00
42.00
42.50
38.00
44.00
41.00
40.50
36.00
35.00
19.00
31.50
27.00
33.00
10
Mean HC Mean MS pts
57.12
60.78
57.28
58.97
51.88
54.56
50.00
51.38
51.56
45.84
49.36
45.16
42.08
46.41
42.52
42.38
36.64
42.09
34.64
40.50
30.24
35.72
SD HC SD MS pts
31.68
36.64
30.96
32.10
21.94
21.13
22.60
24.19
23.22
20.68
22.77
21.86
23.32
26.66
23.49
22.96
17.57
20.63
33.28
31.62
14.70
19.34
p value
.60
.74
.53
.65
.43
.61
.64
.92
.49
.45
.27
Table SI 5: Results of the ANCOVA analysis to assess differences in iCAPs duration between
HC and patients with PMS. Effect size is reported as Cohen’s d.
iCAP
1. Deep GM
2. Auditory/Sensory-Motor
3. Primary Visual
4. ECN
5. aDMN
6. SAL
7. TempPar/LAN
8. Secondary Visual
9. pCun/pDMN
10. Amy/TempPole
11. OFC
p value F(1,52) effect size (Cohen’s d)
0.49
0.48
-0.24
0.33
0.99
0.25
0.86
0.03
-0.22
0.59
0.30
0.43
0.002
10.99
1.07
0.01
6.98
0.47
0.65
0.20
-0.20
0.73
0.12
-0.03
0.55
0.37
-0.27
0.67
0.19
-0.23
0.36
0.85
0.24
11
Table SI 6: Results of the ANCOVA analysis to assess differences in iCAPs Couplings between
HC and patients with PMS. Effect size is reported as Cohen’s d.
iCAPs pair
Couplings
1-2
1-3
2-3
1-4
2-4
3-4
1-5
2-5
3-5
4-5
1-6
2-6
3-6
4-6
5-6
1-7
2-7
3-7
4-7
5-7
6-7
1-8
2-8
3-8
4-8
5-8
6-8
7-8
1-9
2-9
3-9
4-9
5-9
6-9
7-9
8-9
1-10
2-10
3-10
4-10
5-10
6-10
7-10
8-10
9-10
1-11
2-11
3-11
4-11
5-11
6-11
7-11
8-11
9-11
10-11
p value
F(1,52)
effect size (Cohen’s d)
0.31
0.24
0.06
0.68
0.59
0.98
0.37
0.78
0.09
0.27
0.38
0.10
0.70
0.56
0.07
0.10
0.77
0.07
0.60
0.10
0.06
0.64
0.53
0.13
0.66
0.98
0.56
0.76
0.96
0.12
0.54
0.19
0.43
0.11
0.02
0.71
0.48
0.89
0.99
0.78
0.97
0.22
0.72
0.20
0.56
0.15
0.97
0.92
0.97
0.75
0.32
0.76
1.00
0.66
0.88
1.06
1.39
3.77
0.17
0.30
0.00
0.83
0.08
3.07
1.26
0.79
2.74
0.15
0.35
3.52
2.78
0.09
3.40
0.27
2.80
3.79
0.23
0.39
2.39
0.20
0.00
0.34
0.09
0.00
2.50
0.38
1.77
0.63
2.62
5.37
0.14
0.50
0.02
0.00
0.08
0.00
1.58
0.13
1.68
0.34
2.19
0.00
0.01
0.00
0.11
0.99
0.09
0.00
0.19
0.02
-0.17
-0.36
0.46
0.07
0.30
-0.16
0.26
0.21
0.53
-0.47
0.28
0.38
-0.01
0.20
0.34
-0.49
-0.01
0.46
0.06
0.46
0.40
0.08
-0.01
0.25
0.13
-0.03
0.10
0.05
0.00
0.25
-0.03
-0.41
0.29
0.39
0.36
0.08
-0.29
-0.15
0.01
-0.18
-0.06
-0.37
-0.02
-0.27
0.10
0.42
-0.11
-0.03
0.01
0.08
-0.37
-0.07
-0.04
0.06
-0.03
12
Table SI 7: Results of the ANCOVA analysis to assess differences in iCAPs Anti-couplings
between HC and patients with PMS. Effect size is reported as Cohen’s d.
iCAPs pair
Anti-couplings
1-2
1-3
2-3
1-4
2-4
3-4
1-5
2-5
3-5
4-5
1-6
2-6
3-6
4-6
5-6
1-7
2-7
3-7
4-7
5-7
6-7
1-8
2-8
3-8
4-8
5-8
6-8
7-8
1-9
2-9
3-9
4-9
5-9
6-9
7-9
8-9
1-10
2-10
3-10
4-10
5-10
6-10
7-10
8-10
9-10
1-11
2-11
3-11
4-11
5-11
6-11
7-11
8-11
9-11
10-11
p value
F(1,52)
effect size (Cohen’s d)
0.20
0.43
0.08
0.22
0.48
0.32
0.96
0.34
0.87
0.01
0.47
0.66
0.07
0.82
0.28
0.95
0.25
0.16
0.70
0.79
0.83
0.60
0.23
0.76
0.82
0.71
0.22
0.83
0.64
0.94
0.57
0.08
0.73
0.77
0.36
0.65
0.96
0.44
0.57
0.82
0.38
0.87
0.44
0.50
0.41
0.13
0.41
0.25
0.23
0.74
0.55
0.62
0.11
0.88
0.43
1.69
0.64
3.20
1.56
0.50
1.01
0.00
0.93
0.03
7.20
0.53
0.19
3.33
0.05
1.18
0.00
1.37
2.05
0.15
0.07
0.05
0.27
1.48
0.10
0.05
0.14
1.55
0.05
0.22
0.01
0.33
3.27
0.12
0.09
0.86
0.21
0.00
0.61
0.33
0.06
0.80
0.03
0.62
0.46
0.70
2.40
0.70
1.33
1.45
0.12
0.37
0.25
2.65
0.02
0.62
0.34
0.18
-0.59
-0.24
-0.12
0.31
0.02
0.34
0.00
0.90
0.02
-0.18
0.41
-0.17
0.24
0.02
0.22
-0.50
0.11
0.13
-0.04
-0.14
-0.35
-0.14
0.01
0.07
-0.37
-0.12
0.05
-0.08
-0.16
0.48
-0.07
-0.01
-0.19
0.07
-0.04
-0.22
-0.24
-0.06
-0.31
0.05
-0.35
-0.34
-0.28
-0.55
0.20
0.23
0.39
-0.06
0.23
-0.10
0.38
-0.08
-0.30
13
Table SI 8: Results of the ANCOVA analysis to assess differences in iCAPs duration between
CI and CP patients. Effect size is reported as Cohen’s d.
iCAP
1. Deep GM
2. Auditory/Sensory-Motor
3. Primary Visual
4. ECN
5. aDMN
6. SAL
7. TempPar/LAN
8. Secondary Visual
9. pCun/pDMN
10. Amy/TempPole
11. OFC
p value F(1,27) effect size (Cohen’s d)
0.59
0.30
-0.32
0.90
0.02
0.13
0.75
0.10
0.08
0.68
0.18
0.07
0.23
1.54
-0.83
0.80
0.07
-0.08
0.93
0.01
0.24
0.11
2.76
-0.32
0.69
0.17
0.35
0.86
0.03
0.12
0.22
1.57
-0.45
14
Table SI 9: Results of the ANCOVA analysis to assess differences in iCAPs Couplings between
CI and CP patients with PMS. Effect size is reported as Cohen’s d.
iCAPs pair
Couplings
1-2
1-3
2-3
1-4
2-4
3-4
1-5
2-5
3-5
4-5
1-6
2-6
3-6
4-6
5-6
1-7
2-7
3-7
4-7
5-7
6-7
1-8
2-8
3-8
4-8
5-8
6-8
7-8
1-9
2-9
3-9
4-9
5-9
6-9
7-9
8-9
1-10
2-10
3-10
4-10
5-10
6-10
7-10
8-10
9-10
1-11
2-11
3-11
4-11
5-11
6-11
7-11
8-11
9-11
10-11
p value
F(1,27)
effect size (Cohen’s d)
0.71
0.39
0.42
0.001
0.94
0.13
0.35
0.98
0.27
0.76
0.63
0.91
0.89
0.57
0.54
0.47
0.78
0.16
0.26
0.54
0.18
0.83
0.57
0.42
0.72
0.81
0.64
0.27
0.36
0.65
0.70
0.58
0.64
0.15
0.48
0.82
0.65
0.55
0.62
0.55
0.11
0.52
0.22
0.90
0.16
0.55
0.01
0.19
0.53
0.62
0.10
0.47
0.29
0.65
0.76
0.14
0.78
0.67
13.10
0.005
2.46
0.91
0.0004
1.26
0.09
0.24
0.01
0.02
0.32
0.38
0.53
0.08
2.05
1.3
0.39
1.87
0.04
0.34
0.67
0.14
0.06
0.22
1.25
0.85
0.21
0.15
0.31
0.22
2.19
0.51
0.05
0.21
0.36
0.25
0.37
2.75
0.42
1.6
0.02
2.12
0.37
8.35
1.78
0.41
0.25
2.84
0.54
1.16
0.21
0.10
-0.11
0.17
0.47
1.53
0.11
-0.39
-0.35
-0.38
-0.49
-0.10
-0.10
-0.18
-0.13
0.17
-0.005
-0.29
0.06
0.58
-0.34
-0.09
-0.06
-0.20
0.68
-0.19
0.11
-0.2
0.16
0.02
-0.13
0.32
-0.33
0.40
-0.11
0.64
0.60
0.01
-0.10
-0.01
-0.17
-0.22
0.50
-0.45
0.24
-0.29
-0.41
0.08
-0.86
-0.43
0.24
-0.28
-0.56
-0.33
-0.32
-0.08
0.05
15
Table SI 10: Results of the ANCOVA analysis to assess differences in iCAPs Anti-Couplings
between CI and CP patients with PMS. Effect size is reported as Cohen’s d.
iCAPs pair
Anti-Couplings
1-2
1-3
2-3
1-4
2-4
3-4
1-5
2-5
3-5
4-5
1-6
2-6
3-6
4-6
5-6
1-7
2-7
3-7
4-7
5-7
6-7
1-8
2-8
3-8
4-8
5-8
6-8
7-8
1-9
2-9
3-9
4-9
5-9
6-9
7-9
8-9
1-10
2-10
3-10
4-10
5-10
6-10
7-10
8-10
9-10
1-11
2-11
3-11
4-11
5-11
6-11
7-11
8-11
9-11
10-11
p value
F(1,27)
effect size (Cohen’s d)
0.80
0.44
0.11
0.03
0.47
0.01
0.65
0.38
0.98
0.58
0.12
0.74
0.87
0.21
0.79
0.39
0.47
0.24
0.12
0.64
0.27
0.51
0.23
0.34
0.62
0.83
0.08
0.48
0.15
0.63
0.74
0.43
0.19
0.27
0.72
0.28
0.64
0.37
0.87
0.29
0.15
0.28
0.39
0.83
0.16
0.76
0.64
0.63
0.92
0.19
0.81
0.99
0.30
0.40
0.51
0.06
0.62
2.75
5.22
0.54
6.93
0.21
0.81
0.00
0.31
2.59
0.12
0.03
1.63
0.08
0.77
0.53
1.47
2.56
0.22
1.24
0.45
1.50
0.96
0.26
0.05
3.31
0.52
2.20
0.23
0.11
0.65
1.80
1.29
0.13
1.24
0.23
0.84
0.03
1.15
2.19
1.20
0.78
0.05
2.08
0.09
0.22
0.23
0.01
1.78
0.06
0.00
1.10
0.73
0.44
0.04
-0.25
-0.53
-1.11
0.41
1.12
-0.17
-0.22
-0.07
0.04
0.57
-0.06
0.02
0.17
-0.04
0.39
0.46
-0.41
0.78
-0.45
0.74
-0.16
-0.59
-0.06
-0.09
0.18
-0.59
0.14
0.21
-0.03
0.28
0.47
0.38
-0.54
0.09
-0.14
-0.18
0.13
0.09
0.42
-0.50
0.40
0.02
0.19
0.56
-0.05
0.12
-0.16
0.13
-0.48
-0.17
0.07
-0.16
-0.09
0.11
16
Figure SI 4: Bar plot representing the 11 iCAPs duration (A) and the percentage of opposite
(anti-couplings. B) or same (couplings. C) sign overlap between iCAP 5 and the other iCAPs
for HC and MS patients at baseline (subsample of 35 subjects) and FU. ∗p < 0.05, ∗∗p < 0.01.
17
Figure SI 5: Grouped PLSC between aDMN anti-couplings and cognitive test scores. The anticouplings explain the behavioral variance in the CI group. Specifically, the SDMT and BVMTR were inversely associated to the aDMN-Aud/SM and aDMN-pCun/pDMN anti-coupling; the
SDMT additionally correlated positively with aDMN-SecVis and negatively with aDMN-ECN
anti-couplings. Instead, CVLT-II was inversely associated to the aDMN-Amy/TP anti-coupling
but positively to the aDMN-Aud/SM and aDMN/SAL anti-couplings
Figure SI 6: PLSC between all iCAPs durations and clinical test scores. In particualr, the EDSS
and T25FW test positively correlated to the DGM and inversely to Aud/SM and pCun/pDMN
durations. Moreover, T25FW inversely correlated with the TempPar/LAN duration. It emerges
again the association between aDMN and cognitive test scores.
18
Figure SI 7: Spatial patterns of the 11 iCAPs retrieved from the analysis on the subsample of 35 HC and patients with PMS. Under each iCAP the average consensus and the
number of innovation frames assigned to it. Coordinates refer to the Montreal Neurological Institute space. Each iCAP is numbered as the correspondent resulted from the original analysis on 57 subjects. DGM: deep grey matter; Aud/SM: the auditory/sensory-motor;
PrimVis: primary visual; ECN: executive control network; aDMN: anterior DMN; SAL:
salience; TempPar/Lan: temporo-parietal/language; SecVis: secondary visual; pCun/pDMN:
precuneus/posterior DMN; Amy/TP: amygdala/temporal pole; OFC:orbito-frontal cortex.
19
Table SI 11: Appendix for MRI methodology at baseline and FU.
Field strength
3.0 T
Manufacturer
Siemens
Model
MAGNETOM Skyra syngo MR D13
Coil type
head coil
Number of coil channels 32 channel
Type
resting-state fMRI, 3D T1 MPRAGE, T2–weighted TSE sequence
Acquisition time
resting-state fMRI: 6.51 min
3D T1 MPRAGE: 8.45 min
T2–weighted TSE: 2.42 min
Orientation
resting-state fMRI: T ¿ C-20.0
3D T1 MPRAGE: Sagittal
T2–weighted TSE: Transversal
Voxel size
resting-state fMRI: 2.1×2.1×2.1 mm
3D T1 MPRAGE: 0.8×0.8×0.8 mm
T2–weighted TSE: 0.5×0.5×3.0 mm
TR
resting-state fMRI: 1000 ms
3D T1 MPRAGE: 3000.0 ms
T2–weighted TSE: 8000.0 ms
TE
resting-state fMRI: 35.00 ms
3D T1 MPRAGE: 2.47 ms
T2–weighted TSE: 95.00 ms
TI
3D T1 MPRAGE: 1000 ms
Flip angle
resting-state fMRI: 60 deg
3D T1 MPRAGE: 7 deg
T2–weighted TSE: 160 deg
Field of view
resting-state fMRI: 228x228 mm2
3D T1 MPRAGE: 256x256 mm2
T2–weighted TSE: 256x256 mm2
Parallel imaging
resting-state fMRI: none
3D T1 MPRAGE: GRAPPA
T2–weighted TSE:none
Contrast enhancement
No
Lesions
Type
T2-hyperintense, T1-hypointense
Analysis method
Semiautomatic segmentation technique
Analysis software
Jim software package (Version 7, , Xinapse Systems, Northants, UK)
Output measure
Lesion volume [ml]
Tissue volumes
Type
Whole brain, gray matter and white matter normalized volumes, percentage of brain volume changes at follow-up
Analysis method
Non-brain tissue stripping, brain segmentation and skull based normalization
Analysis software
SIENAX SIENA
Output measure
Normalized brain volume (NBV), normalized grey matter volume, normalized white matter volume.
Percentage brain of volume change (PBVC)
Other MRI measures
Type
RS fMRI dynamic functional activity
Analysis method
RS fMRI pre-processing iCAPs pipeline (https://c4science.ch/source/iCAPs/) Calculation of the temporal
properties of iCAPs (overall duration and couplings/anti-couplings between each pair of iCAPs
Analysis software
SPM12, MATLAB, DPARSF, IBASPM
Output measure
Overall duration of each iCAP (or brain state, retrieved from innovation signals, which encode onsets of
activations/de-activations by positive/ negative spikes,) and interactions between each pair of iCAPs.
20
References
[1] C. H. Polman, S. C. Reingold, B. Banwell, M. Clanet, J. A. Cohen, M. Filippi, K. Fujihara, E. Havrdova, M. Hutchinson, L. Kappos, F. D. Lublin, X. Montalban, P. O’Connor,
M. Sandberg-Wollheim, A. J. Thompson, E. Waubant, B. Weinshenker, and J. S. Wolinsky, “Diagnostic criteria for multiple sclerosis: 2010 revisions to the McDonald criteria.,” eng, Annals of neurology, vol. 69, no. 2, pp. 292–302, Feb. 2011, ISSN: 1531-8249
(Electronic). DOI: 10.1002/ana.22366.
[2] J. D. Power, K. A. Barnes, A. Z. Snyder, B. L. Schlaggar, and S. E. Petersen, “Spurious
but systematic correlations in functional connectivity MRI networks arise from subject
motion,” NeuroImage, vol. 59, no. 3, pp. 2142–2154, Feb. 2012, ISSN: 1053-8119. DOI:
https : / / doi . org / 10 . 1016 / j . neuroimage . 2011 . 10 . 018. [Online].
Available: http://www.sciencedirect.com/science/article/pii/
S1053811911011815.
[3] L. Parkes, B. Fulcher, M. Yucel, and A. Fornito, “An evaluation of the efficacy, reliability, and sensitivity of motion correction strategies for resting-state functional MRI.,”
eng, NeuroImage, vol. 171, pp. 415–436, May 2018, ISSN: 1095-9572 (Electronic). DOI:
10.1016/j.neuroimage.2017.12.073.
[4] Y. Farouj, F. I. Karahanoglu, and D. Van De Ville, “Regularized spatiotemporal deconvolution of fMRI data using gray-matter constrained total variation,” Proceedings - International Symposium on Biomedical Imaging, pp. 472–475, 2017, ISSN: 19458452. DOI:
10.1109/ISBI.2017.7950563.
[5] F. I. Karahanoğlu, C. Caballero-Gaudes, F. Lazeyras, and D. Van De Ville, “Total activation: fMRI deconvolution through spatio-temporal regularization,” NeuroImage, vol. 73,
pp. 121–134, 2013, ISSN: 1053-8119. DOI: https : / / doi . org / 10 . 1016 / j .
neuroimage.2013.01.067. [Online]. Available: http://www.sciencedirect.
com/science/article/pii/S1053811913001146.
[6] F. I. Karahanoglu and D. Van De Ville, “Transient brain activity disentangles fMRI
resting-state dynamics in terms of spatially and temporally overlapping networks.,” eng,
Nature communications, vol. 6, p. 7751, 2015, ISSN: 2041-1723 (Electronic). DOI: 10.
1038/ncomms8751.
[7] D. M. Zoller, T. A. Bolton, F. I. Karahanoglu, S. Eliez, M. Schaer, and D. Van De
Ville, “Robust recovery of temporal overlap between network activity using transientinformed spatio-temporal regression,” IEEE Transactions on Medical Imaging, vol. 38,
no. 1, pp. 291–302, 2019, ISSN: 1558254X. DOI: 10.1109/TMI.2018.2863944.
[8] A. Tarun, D. Wainstein-Andriano, V. Sterpenich, L. Bayer, L. Perogamvros, M. Solms,
N. Axmacher, S. Schwartz, and D. Van De Ville, “NREM sleep stages specifically alter
dynamical integration of large-scale brain networks,” iScience, vol. 24, no. 1, p. 101 923,
2021, ISSN: 25890042. DOI: 10.1016/j.isci.2020.101923. [Online]. Available: https://doi.org/10.1016/j.isci.2020.101923.
[9] S. Monti, P. Tamayo, J. Mesirov, and T. Golub, “Consensus clustering: A resamplingbased method for class discovery and visualization of gene expression microarray data,”
Machine Learning, vol. 52, no. 1-2, pp. 91–118, 2003, ISSN: 08856125. DOI: 10 .
1023/A:1023949509487.
21
[10] A. Krishnan, L. J. Williams, A. R. McIntosh, and H. Abdi, “Partial Least Squares (PLS)
methods for neuroimaging: a tutorial and review.,” eng, NeuroImage, vol. 56, no. 2,
pp. 455–475, May 2011, ISSN: 1095-9572 (Electronic). DOI: 10.1016/j.neuroimage.
2010.07.034.
[11] B. A. Parmenter, S. M. Testa, D. J. Schretlen, B. Weinstock-Guttman, and R. H. Benedict, “The utility of regression-based norms in interpreting the minimal assessment of
cognitive function in multiple sclerosis (MACFIMS),” Journal of the International Neuropsychological Society, vol. 16, no. 1, pp. 6–16, 2010, ISSN: 13556177. DOI: 10 .
1017/S1355617709990750.
[12] A. Schaefer, R. Kong, E. M. Gordon, T. O. Laumann, X.-N. Zuo, A. J. Holmes, S. B.
Eickhoff, and B. T. T. Yeo, “Local-Global Parcellation of the Human Cerebral Cortex
from Intrinsic Functional Connectivity MRI,” Cerebral Cortex, vol. 28, no. 9, pp. 3095–
3114, 2018, ISSN: 1047-3211. DOI: 10.1093/cercor/bhx179. [Online]. Available:
https://academic.oup.com/cercor/article/28/9/3095/3978804.
[13] W. R. Shirer, S. Ryali, E. Rykhlevskaia, V. Menon, and M. D. Greicius, “Decoding
subject-driven cognitive states with whole-brain connectivity patterns,” Cerebral Cortex,
vol. 22, no. 1, pp. 158–165, 2012, ISSN: 10473211. DOI: 10.1093/cercor/bhr099.
[14] N. Tzourio-Mazoyer, B. Landeau, D. Papathanassiou, F. Crivello, O. Etard, N. Delcroix,
B. Mazoyer, and M. Joliot, “Automated anatomical labeling of activations in SPM using
a macroscopic anatomical parcellation of the MNI MRI single-subject brain,” NeuroImage, vol. 15, no. 1, pp. 273–289, 2002, ISSN: 10538119. DOI: 10.1006/nimg.2001.
0978.
22