Tracking whole-brain connectivity dynamics in the resting-state

Similar documents
Supplementary Information. Brain networks involved in tactile speed classification of moving dot patterns: the. effects of speed and dot periodicity

SUPPLEMENTARY INFORMATION

SUPPLEMENTARY MATERIAL. Patient Group Hem. Lesion Description Etiology FRONTAL LOBE GROUP 1 FL-LAT right SFG (pos), SFS, Central sulcus

One patient and one control were unable to complete the TOMM. An additional HC had abnormally low scores

- Supplementary information -

Nature Neuroscience: doi: /nn Supplementary Figure 1. Localization of responses

Supplementary Material & Data. Younger vs. Older Subjects. For further analysis, subjects were split into a younger adult or older adult group.

Morphometrics with SPM12

REVEALING THE INFLUENCE OF DEFINITE BRAIN REGIONS UPON THE EMERGENCE OF SPATIAL AND TEMPORAL PATTERNS IN THE RESTING-STATE BRAIN ACTIVITY

Integrative Methods for Functional and Structural Connectivity

Hierarchical Clustering Identifies Hub Nodes in a Model of Resting-State Brain Activity

RS-fMRI analysis in healthy subjects confirms gender based differences

A Multivariate Time-Frequency Based Phase Synchrony Measure for Quantifying Functional Connectivity in the Brain

Modelling Time- varying effec3ve Brain Connec3vity using Mul3regression Dynamic Models. Thomas Nichols University of Warwick

Signal Modeling Techniques in Speech Recognition. Hassan A. Kingravi

Undirected graphs of frequency-dependent functional connectivity in whole brain networks

A Study of Brain Networks Associated with Swallowing Using Graph-Theoretical Approaches

Self-Evaluation Across Adolescence: A Longitudinal fmri Study. Dani Cosme

Figure 1-figure supplement 1

NeuroImage. Diffeomorphic registration using geodesic shooting and Gauss Newton optimisation. John Ashburner, Karl J. Friston.

Resting-state fmri and DTI

Resampling-Based Information Criteria for Adaptive Linear Model Selection

Internally generated preactivation of single neurons in human medial frontal cortex predicts volition

Longitudinal growth analysis of early childhood brain using deformation based morphometry

Will Penny. 21st April The Macroscopic Brain. Will Penny. Cortical Unit. Spectral Responses. Macroscopic Models. Steady-State Responses

Neural Network. Eung Je Woo Department of Biomedical Engineering Impedance Imaging Research Center (IIRC) Kyung Hee University Korea

Statistical Analysis Aspects of Resting State Functional Connectivity

Outlines: (June 11, 1996) Instructor:

New Procedures for False Discovery Control

Supplementary Figure 1. Structural MRIs.

Electroencephalogram Based Causality Graph Analysis in Behavior Tasks of Parkinson s Disease Patients

New Machine Learning Methods for Neuroimaging

Unravelling the Intrinsic Functional Organization of the Human Lateral Frontal Cortex: A Parcellation Scheme Based on Resting State fmri

Kernel Granger Causality Mapping Effective Connectivity on fmri Data Wei Liao, Daniele Marinazzo, Zhengyong Pan, Qiyong Gong, and Huafu Chen*

腦網路連結分析 A Course of MRI

Nature Neuroscience: doi: /nn Supplementary Figure 1. Expression of GCaMP6s in LGN and their axons in V1.

Part I: SEM & Introduction to Time Series Models A brief introduction to SEM with an eating disorder example A fmri time series study + AR and MV

Dynamic Causal Modelling for fmri

Discovering the Human Connectome

Modelling temporal structure (in noise and signal)

Finding a Basis for the Neural State

From Pixels to Brain Networks: Modeling Brain Connectivity and Its Changes in Disease. Polina Golland

A MULTIVARIATE MODEL FOR COMPARISON OF TWO DATASETS AND ITS APPLICATION TO FMRI ANALYSIS

Recipes for the Linear Analysis of EEG and applications

UvA-DARE (Digital Academic Repository)

DYNAMIC FUNCTIONAL CONNECTIVITY USING HEAT KERNEL

Bayesian Regression with Undirected Network Predictors with an Application to Brain Connectome Data

Neuroimage Processing

Multiple Orderings of Events in Disease Progression

Group Study of Simulated Driving fmri Data by Multiset Canonical Correlation Analysis

Contextual connectivity: A framework for understanding the intrinsic dynamic architecture of large-scale functional brain networks

Fundamentals of Computational Neuroscience 2e

Cortical Shape Analysis using the Anisotropic Global Point Signature

Consider the following spike trains from two different neurons N1 and N2:

Functional Connectivity and Network Methods

to appear in Frank Eeckman, ed., Proceedings of the Computational Neurosciences Conference CNS*93, Kluwer Academic Press.

Experimental design of fmri studies

Searching for Nested Oscillations in Frequency and Sensor Space. Will Penny. Wellcome Trust Centre for Neuroimaging. University College London.

Dynamical Constraints on Computing with Spike Timing in the Cortex

Evidence for Hubs in Human Functional Brain Networks

Jan 16: The Visual System

3D Moment Invariant Based Morphometry

King s Research Portal

Fundamentals of Computational Neuroscience 2e

ROI analysis of pharmafmri data: an adaptive approach for global testing

HST 583 FUNCTIONAL MAGNETIC RESONANCE IMAGING DATA ANALYSIS AND ACQUISITION A REVIEW OF STATISTICS FOR FMRI DATA ANALYSIS

Experimental design of fmri studies & Resting-State fmri

MultiDimensional Signal Processing Master Degree in Ingegneria delle Telecomunicazioni A.A

Journal of Neuroscience Methods

CORRELATION TRANSFER FROM BASAL GANGLIA TO THALAMUS IN PARKINSON S DISEASE. by Pamela Reitsma. B.S., University of Maine, 2007

Effective Connectivity & Dynamic Causal Modelling

Uncertainty quantification and visualization for functional random variables

Neuroimage Processing

Clicker Question. Discussion Question

Intro and Homeostasis

Learning Structural Equation Models for fmri

Non-parametric Bayesian graph models reveal community structure in resting state fmri

Testing Stationarity of Brain Functional Connectivity Using Change-Point Detection in fmri Data

Machine learning comes from Bayesian decision theory in statistics. There we want to minimize the expected value of the loss function.

For more information about how to cite these materials visit

ENVIRONMENTAL DATA ANALYSIS WILLIAM MENKE JOSHUA MENKE WITH MATLAB COPYRIGHT 2011 BY ELSEVIER, INC. ALL RIGHTS RESERVED.

A framework for systematic promoter motif discovery and expression profiling from high dimensional brain transcriptome data. Jeremy A.

Induced Magnetic Force in Human Heads Exposed to 4 T MRI

Optimal Aggregation of Classifiers and Boosting Maps in Functional Magnetic Resonance Imaging

Dreem Challenge report (team Bussanati)

Temporal context calibrates interval timing

arxiv: v2 [q-bio.nc] 27 Sep 2018

Event-related fmri. Christian Ruff. Laboratory for Social and Neural Systems Research Department of Economics University of Zurich

COMP 546. Lecture 21. Cochlea to brain, Source Localization. Tues. April 3, 2018

Revealing Interactions Among Brain Systems With Nonlinear PCA

Higher Order Cartesian Tensor Representation of Orientation Distribution Functions (ODFs)

Measuring Brain Variability by Extrapolating Sparse Tensor Fields Measured on Sulcal Lines

Quantifying and mitigating correlated noise between formed beams on the ASKAP Phased Array Feeds

EEG- Signal Processing

Transfer Learning in Brain-Computer Interfaces

Supplementary Figure 1. Characterization of the single-photon quantum light source based on spontaneous parametric down-conversion (SPDC).

NeuroImage xxx (2009) xxx xxx. Contents lists available at ScienceDirect. NeuroImage. journal homepage:

Bagging During Markov Chain Monte Carlo for Smoother Predictions

NCNC FAU. Modeling the Network Architecture of the Human Brain

Single-Trial Neural Correlates. of Arm Movement Preparation. Neuron, Volume 71. Supplemental Information

Transcription:

Tracking whole-brain connectivity dynamics in the resting-state Supplementary Table. Peak Coordinates of ICNs ICN regions BA t max Peak (mm) (continued) BA t max Peak (mm) X Y Z X Y Z Subcortical networks ITG (93) Caudate (57) L inferior temporal gyrus 9 42.9 5 63 9 R caudate nucleus 7.4 2 5 R inferior temporal gyrus 9 39.2 5 6 9 L caudate nucleus 66.7 9 5 PreCG (56) Putamen (8) L precentral gyrus 4 36.4 48 6 48 R putamen 69.6 2 2 6 R precentral gyrus 4 36.7 5 6 48 L putamen 7.4 2 9 3 L supplementary motor area 6 3.4 66 Putamen (3) MCC (9) R putamen 76.4 27 6 3 Bi middle cingulate cortex 32 37.5 3 8 39 L putamen 69.7 3 3 pinsula (7) Thalamus (5) R insula 3 5.5 45 3 3 R thalamus 49.2 9 5 6 L insula 3 4.4 45 6 3 L thalamus 52. 9 8 6 L IPL (76) Auditory networks L inferior parietal lobule 4 5.6 42 5 5 STG (99) L middle frontal gyrus 46 29.8 48 33 2 L superior temporal gyrus 3 38. 5 33 8 L inferior temporal gyrus 37 27.8 57 54 2 R superior temporal gyrus 3 36.6 5 3 2 MiFG (47) STG (98) R middle frontal gyrus 9 42.3 45 5 33 R superior temporal gyrus 2 47.8 63 24 L middle frontal gyrus 9 36. 42 9 33 L superior temporal gyrus 4 44.8 57 2 3 IFG (88) Somatomotor networks R inferior frontal gyrus 46 37.3 48 39 PreCG (2) L inferior frontal gyrus 46 39. 48 33 9 R postcentral gyrus 6 74.8 57 6 27 ainsula (2) L postcentral gyrus 6 7. 54 9 3 R insula 47 56.7 33 24 6 R cerebellum (VI) 7.7 8 66 8 L insula 47 5.7 3 24 3 L cerebellum (VI) 9.3 8 63 2 R medial frontal gyrus 6 27.6 6 3 39 R PoCG () IPL (59) R postcentral gyrus 3 57.6 42 24 6 R inferior parietal lobule 4 49. 57 45 42 L cerebellum (VI) 7.7 24 5 24 L inferior parietal lobule 4 45.6 57 45 39 L PoCG (4) MiFG (36) L postcentral gyrus 3 57.5 42 24 57 R middle frontal gyrus 44.2 3 54 6 R cerebellum (VI) 23.2 2 5 24 L middle frontal gyrus 45.6 3 54 2 L supplementary motor area 6 9.6 3 2 54 SMA (29) SMA (58) Bi supplementary motor area 6 45. 3 5 6 Bi supplementary motor area 24 46.3 3 3 48 R STG+IFG (66) ParaCL (37) R superior temporal gyrus 22 43.9 57 48 5 Bi paracentral lobule 6 65.8 3 24 66 R inferior frontal gyrus 47 22.2 5 3 3 PoCG (45) PHG (8) L postcentral gyrus 2 52.4 57 24 36 R ParaHippocampal gyrus 28 37.2 24 8 2 R postcentral gyrus 4 4.4 57 2 39 L ParaHippocampal gyrus 28 35. 2 8 2 ParaCL (3) Default mode networks Bi paracentral lobule 6 46.8 8 9 66 Precuneus (2) SPL (9) Bi precuneus 7 68.7 6 72 39 R superior parietal lobule 7 52. 8 54 66 Bi middle cingulate cortex 23 39.5 24 3 L superior parietal lobule 5 52.4 2 45 63 Precuneus (44) Visual networks Bi precuneus 7 65.2 54 5 MTG (77) PCC (39) L middle temporal gyrus 39 42.3 48 72 9 L posterior cingulate 3 67.7 2 57 5 R middle temporal gyrus 39 44.5 48 63 9 R posterior cingulate 3 67.8 5 54 5 LingualG (79) PCC (28) R lingual gyrus 9 5. 2 72 6 Bi posterior cingulate cortex 23 65.7 3 36 27 L lingual gyrus 8 44.7 8 75 6 R AG (64) Cuneus (6) R angular gyrus 4 62.6 45 66 42 Bi cuneus 8 6.8 84 24 R Precuneus 7 35. 6 63 39 CalcarineG (46) R superior frontal gyrus 8 3.4 24 33 54 L calcarine gyrus 3 66. 9 69 9 ACC (26) R calcarine gyrus 3 63.3 5 66 9 Bi anterior cingulate cortex 32 57. 3 45 3 Cuneus (27) L AG (75) Bi cuneus 7 62.6 3 87 3 L angular gyrus 39 56.7 48 66 33 FFG (43) L Precuneus 7 49. 6 33 L fusiform gyrus 37 5.5 27 48 2 R angular gyrus 39 3.7 5 63 3 R fusiform gyrus 37 52.5 3 42 5 L superior frontal gyrus 8 22. 8 39 5 R MOG (82) L medial frontal gyrus 2.2 3 5 6 R middle occipital gyrus 9 47. 36 84 6 MiFG+SFG (48) L MOG (89) L middle frontal gyrus 8 33. 24 24 48 L middle occipital gyrus 8 42.4 27 9 L superior frontal gyrus 8 34. 33 48 MOG (8) R middle frontal gyrus 8 29.8 2 27 45 R middle occipital gyrus 8 7.8 3 93 L MTG+IFG (87) L middle occipital gyrus 8 65.6 27 96 3 L middle temporal gyrus 2 42.4 57 39 3 SOG (6) L inferior frontal gyrus 44 39.2 54 8 2 L superior occipital gyrus 9 45. 36 8 3 Cerebellar networks R superior occipital gyrus 9 47. 39 75 36 R CB (2) Cognitive control networks R cerebellum (crus 2) 5.6 33 75 39 R IPL (67) L cerebellum (crus 2) 26.9 3 78 36 R inferior parietal lobule 4 52.3 45 39 5 L CB (7) R middle frontal gyrus 46 33.3 48 42 5 L cerebellum (crus 2) 38.2 36 6 42 R inferior frontal gyrus 44 29.6 54 9 24 CB (32) R inferior temporal gyrus 37 27.7 57 54 9 Bi cerebellum (VI) 43.9 5 66 24

A DVARS (std) EFFECT OF POST-PROCESSING STEPS ON MOTION-RELATED VARIANCE 3 2 Subject 24 Subject 52 Original TCs, all components Original TCs, just ICNs Regression of realignment parameters Regression + outlier removal B AVERAGE FC, ORIGINAL SC AUD SM VIS CC FD (mm) 2 2 4 6 8 2 4 Frame # (TR = 2 s) DM CB SC AUD SM VIS CC C AVERAGE FC, FULLY POST-PROCESSED DM CB DVARS (std) 3 2 Subject 267 Subject 349 +.62 correlation (r).62 FD (mm) 2 2 4 6 8 2 4 Frame # (TR = 2 s) D DIFFERENCE (POST-PROCESSED ORIGINAL) FD (mm) DVARS (std) 3 Subject 36 2 2 Subject 367 correlation (r) +.. 2 4 6 8 2 4 Frame # (TR = 2 s) Figure S. Effects of motion removal at the subject (A) and group levels (B-D). In (A), we display the framewise displacement (FD, bottom panel) computed from realignment parameters, and the root-mean-square of the temporal differential of components TCs (DVARS, top panel, see Power JD et al., 22) for the three example subjects (left column) and three additional subjects chosen at random (right column). From the large spikes in DVARS, it is clear that the original TCs of all components (solid blue line) as well as just ICN components (dotted blue line) are contaminated with motion-related variance. Regression of realignment parameters and their derivatives (green line) reduces motion influence in some cases, but additional outlier removal (black line) offers much greater improvement. The effect of post-processing can be seen in the difference between stationary FC matrices computed from the original TCs (B) and fully post-processed TCs (regression of realignment parameters and outlier removal) (C), as shown in panel (D). In agreement with (Power JD et al., 22), we observe motioninduced increases in short-range connections (e.g., between SM ICNs) and, to a lesser extent, decreases in longrange connections (e.g., between SC and VIS ICNs).

Subcortical Networks Component 57, Caudate 7.4 X = 2 mm Y = mm Z = 5 mm 7.4 Auditory networks Component 8, Putamen Component 99, STG 7.4 38. X = 2 mm Y = 2 mm Z = 6 mm 7.4 X = 5 mm Y = 33 mm Z = 8 mm 38. X = 2 mm Y = 9 mm Z = 3 mm X = 5 mm Y = 3 mm Z = 2 mm Component 3, Putamen Component 98, STG 76.4 47.8 X = 27 mm Y = 6 mm Z = 3 mm 76.4 X = 63 mm Y = 24 mm Z = mm 47.8 X = 3 mm Y = 3 mm Z = mm X = 57 mm Y = 2 mm Z = 3 mm Component 5, Thalamus 52. X = 9 mm Y = 8 mm Z = 6 mm 52. Figure S2. SMs of ICNs. ICNs are divided into the same groups shown in Figure 2 and are thresholded at t > 2, where one-sample t-statistics have been computed across the single-subject SMs. Sagittal, coronal, and axial slices are shown at the maximal t-statistic for clusters larger than 3 cm 3.

Somatomotor networks Component 2, PreCG Component 37, ParaCL 74.8 65.8 X = 57 mm Y = 6 mm Z = 27 mm 74.8 X = 5 mm Y = 24 mm Z = 66 mm 65.8 Component 45, PoCG 52.4 X = 54 mm Y = 9 mm Z = 3 mm Component, R PoCG X = 57 mm Y = 24 mm Z = 36 mm 52.4 57.6 X = 42 mm Y = 24 mm Z = 6 mm Component 4, L PoCG 57.6 X = 57 mm Y = 2 mm Z = 39 mm Component 3, ParaCL 57.5 46.7 X = 42 mm Y = 24 mm Z = 57 mm 57.5 X = 8 mm Y = 9 mm Z = 66 mm 46.7 Component 58, SMA Component 9, SPL 46.3 52.4 X = 5 mm Y = 3 mm Z = 48 mm 46.3 X = 2 mm Y = 45 mm Z = 63 mm 52.4

Visual networks Component 77, MTG 44.5 Component 27, Cuneus 62.6 X = 48 mm Y = 72 mm Z = 9 mm 44.5 X = 5 mm Y = 87 mm Z = 3 mm 62.6 Component 43, FFG X = 48 mm Y = 63 mm Z = 9 mm 52.5 Component 79, LingualG 5. X = 27 mm Y = 48 mm Z = 2 mm 52.5 X = 2 mm Y = 72 mm Z = 6 mm 5. X = 3 mm Y = 42 mm Z = 5 mm Component 82, R MOG X = 8 mm Y = 75 mm Z = 6 mm 47. Component 6, Cuneus 6.8 X = 36 mm Y = 84 mm Z = 6 mm Component 89, L MOG 47. X = mm Y = 84 mm Z = 24 mm 6.8 42.4 Component 46, CalcarineG 66. X = 27 mm Y = 9 mm Z = mm 42.4 X = 9 mm Y = 69 mm Z = 9 mm 66.

Cognitive control networks Component 67, R IPl 52.3 Visual networks continued Component 8, MOG X = 45 mm Y = 39 mm Z = 5 mm 52.3 7.8 X = 3 mm Y = 93 mm Z = mm 7.8 X = 48 mm Y = 42 mm Z = 5 mm Component 93, ITG 42.9 X = 27 mm Y = 96 mm Z = 3 mm Component 6, SOG X = 5 mm Y = 63 mm Z = 9 mm 42.9 47. X = 36 mm Y = 8 mm Z = 29 mm 47. X = 5 mm Y = 6 mm Z = 9 mm Component 56, PreCG 36.7 X = 39 mm Y = 75 mm Z = 36 mm X = 48 mm Y = 6 mm Z = 48 mm 36.7 X = 5 mm Y = 6 mm Z = 48 mm

Cognitive control networks continued () Component 47, MiFG 42.3 Component 9, MCC 37.5 X = 45 mm Y = 5 mm Z = 33 mm 42.3 X = 5 mm Y = 8 mm Z = 39 mm 37.5 Component 7, pinsula 5.5 X = 42 mm Y = 9 mm Z = 33 mm Component 88, IFG 39. X = 45 mm Y = 3 mm Z = 3 mm 5.5 X = 48 mm Y = 39 mm Z = mm 39. X = 45 mm Y = 6 mm Z = 3 mm Component 76, L IPL 5.6 X = 48 mm Y = 33 mm Z = 9 mm Component 2, ainsula 56.7 X = 42 mm Y = 5 mm Z = 5 mm 5.6 X = 33 mm Y = 24 mm Z = 6 mm 56.7 X = 48 mm Y = 33 mm Z = 2 mm X = 3 mm Y = 24 mm Z = 3 mm

Cognitive control networks continued (2) Component 59, IPL 49. X = 57 mm Y = 45 mm Z = 39 mm 49. Component 66, R STG+IFG 43.9 X = 57 mm Y = 45 mm Z = 42 mm Component 36, MiFG X = 57 mm Y = 48 mm Z = 5 mm Component 8, PHG 43.9 45.6 37.2 X = 3 mm Y = 54 mm Z = 6 mm 45.6 X = 24 mm Y = 8 mm Z = 2 mm 37.2 X = 3 mm Y = 54 mm Z = 2 mm Component 29, SMA X = 2 mm Y = 8 mm Z = 2 mm 45. X = 5 mm Y = 5 mm Z = 6 mm 45.

Default-mode networks Component 64, R AG 62.6 Component 2, Precuneus 68.7 X = 45 mm Y = 66 mm Z = 42 mm 62.6 X = 6 mm Y = 72 mm Z = 39 mm 68.7 X = 24 mm Y = 33 mm Z = 54 mm X = mm Y = 24 mm Z = 3 mm Component 44, Precuneus 65.2 X = 6 mm Y = 63 mm Z = 39 mm Component 26, ACC 57. X = mm Y = 54 mm Z = 5 mm 65.2 Component 39, PCC 67.8 X = 5 mm Y = 45 mm Z = 3 mm 57. Component 75, L AG 56.7 X = 5 mm Y = 54 mm Z = 5 mm 67.8 Component 28, PCC 65.7 X = 48 mm Y = 66 mm Z = 33 mm 56.7 X = 5 mm Y = 36 mm Z = 27 mm 65.7 X = mm Y = 6 mm Z = 33 mm

Default-mode networks continued Component 48, MiFG+SFG Cerebellar networks Component 2, R CB 5.6 34. X = 33 mm Y = 75 mm Z = 39 mm 5.6 X = mm Y = 33 mm Z = 48 mm Component 87, L MTG+IFG 34. Component 7, L CB 38.2 42.4 X = 36 mm Y = 6 mm Z = 42 mm 38.2 X = 57 mm Y = 39 mm Z = 3 mm 42.4 Component 32, CB 43.9 X = 54 mm Y = 8 mm Z = 2 mm X = 5 mm Y = 66 mm Z = 24 mm 43.9

.5 AUC.2.9 COM (Hz).22 SC AUD SM A AREA UNDER THE CURVE (AUC) OF FC SPECTRA, AVERAGED OVER SUBJECTS B CENTER OF MASS (COM) OF FC SPECTRA, AVERAGED OVER SUBJECTS VIS CC DM CB ventricular cortical WM ICNs ARTs CB WM motion vascular susceptibility SC AUD SM VIS CC DM CB ventricular cortical WM CB WM motion vascular susceptibility ICNs ARTs AUC of FC spectra between all ICNs...9.8.7 t = 39.4, P =.3e 4.7.8.9.. COM of FC spectra between ICNs (Hz)...9 t = 27.7, P = 2.7e 95.9.. AUC of FC spectra between all ARTs COM of FC spectra between all ARTs (Hz) Figure S3. Differences in FC variability between ICNs and ARTs. FC spectra were evaluated in terms of the area under the curve (AUC) and the center-of-mass (COM), both non-parametric descriptors of the spectral shape that are robust to noise. FC oscillations between ICNs are larger variable (A, top) and are focused at lower frequencies (B, top). Comparing the FC spectral properties within subjects between all ICNs (average over quadrant II) and all ARTs (average over quadrant IV) yields highly consistent and significant differences (A-B, bottom).

k = 2 k = 3 k = 4 k = 5 k = 6 k = 7 k = 8 k = 9 k = 933 (64%) 7 (57%) 53 (5%) 248 (4%) 73 (39%) 3 (34%) 879 (29%) 82 (27%) 755 (25%) S 93 (36%) 975 (32%) 593 (2%) 497 (6%) 383 (3%) 243 (8%) 227 (8%) 25 (7%) 29 (7%) S2 34 (%) 295 (%) 283 (9%) 279 (9%) 265 (9%) 263 (9%) 233 (8%) 23 (8%) S3 67 (2%) 49 (6%) 387 (3%) 357 (2%) 225 (7%) 28 (9%) 83 (6%) S4 57 (7%) 487 (6%) 49 (4%) 393 (3%) 35 (2%) 39 (%) S5 37 (%) 289 (%) 243 (8%) 37 (%) 265 (9%) S6 423 (4%) 333 (%) 263 (9%) 249 (8%) S7 463 (5%) 299 (%) 279 (9%) S8 255 (8%) 225 (7%) S9 3 (%) S Figure S4. Cluster centroids for k = 2 to. For each k, the k-means algorithm was applied with 5 repetitions to the subject exemplar windows (326 instances). The gray rectangle highlights the clustering result presented in the main text. The total number and percentage of occurrences is listed above each centroid.

A CLUSTER CENTROIDS FROM BOOTSTRAP RESAMPLES S S2 S3 S4 S5 S6 S7 63 (35%) 93 (6%) 28 (9%) 377 (2%) 44 (5%) 28 (9%) 393 (3%) 69 (35%) 27 (7%) 277 (9%) 353 (2%) 45 (5%) 35 (%) 4 (3%) 35 (34%) 237 (8%) 286 (%) 367 (2%) 369 (2%) 275 (9%) 44 (5%) 973 (32%) 24 (8%) 27 (9%) 365 (2%) 425 (4%) 3 (%) 455 (5%) 2 (34%) 263 (9%) 267 (9%) 39 (%) 429 (4%) 263 (9%) 425 (4%) B CLUSTER CENTROIDS FROM SPLIT-HALF ANALYSIS 535 (35%) 3 (7%) 2 (8%) 85 (2%) 235 (5%) 29 (8%) 29 (4%) HALF SAMPLE 5 SAMPLE 4 SAMPLE 3 SAMPLE 2 SAMPLE 489 (33%) 5 (8%) 45 (%) 87 (2%) 95 (3%) 59 (%) 29 (4%) HALF 2 Figure S5. Cluster centroids for bootstrap resamples (A) and split-half samples (B). In (A), subjects were resampled and the k-means algorithm was applied with 5 repetitions to the subject exemplar windows ( 3 to 4 instances depending on the sample). In (B), the subjects were split into two groups and the k-means algorithm was applied with 5 repetitions to the subject exemplars in that group ( 5 instances). The total number and percentage of occurrences is listed above each centroid.

A EXAMPLE FC TIMESERIES CalcarineG (46) to PreCG (56) CalcarineG (46) to MCC (9) CalcarineG (46) to pinsula (7) ORIGINAL DATA Correlation (z).5.5 subject 24 5 5 2 25 time (s) B AVERAGE OF SUBJECT EXEMPLARS C CLUSTER CENTROIDS (k = 7) S S2 S3 S4 S5 S6 S7 3 (34%) 243 (8%) 265 (9%) 357 (2%) 49 (4%) 289 (%) 423 (4%) SR: CONSISTENT FOURIER PHASE SHIFT Correlation (z).5.5 5 5 2 25 time (s) 496 (6%) 4 (3%) 249 (8%) 458 (5%) 5 (7%) 473 (6%) 438 (4%) SR2: INCONSISTENT FOURIER PHASE SHIFT 674 (55%) (%) (%) 2 (%) 39 (34%) 232 (8%) 6 (2%) Correlation (z).5.5 5 5 2 25 time (s) Figure S6. Clustering results with surrogate datasets. (A) Demonstration of surrogate data creation from the original FC timeseries (top) using a consistent Fourier phase shift (SR, middle) and inconsistent phase shift (SR2, bottom). In SR, the FC timeseries maintain their phase relationships. In SR2, phase relationships (and thus covariance structures) are disrupted. Neither consistent nor inconsistent Fourier phase shifting alters the mean, variance, or temporal autocorrelation of individual FC timeseries, as demonstrated by the very similar average subject exemplars shown in (B) for original data (top) SR (middle) and SR2 (bottom). (C) Cluster centroids for original data (top, identical to Figure S4, k = 7), SR (middle), and SR2 (bottom). Clustering of exemplars from SR2 fails to replicate the centroids observed in original data and SR : two clusters (S2 and S3) are empty and all others strongly resemble the mean shown in (B). This suggests it is the patterns of covariance (preserved in SR), rather than distinctions in mean or variance, that drive the clustering.

A SIMULATED DYNAMIC NEURAL FUNCTIONAL CONNECTIVITY state state 2 state 3 state 4 state 2 + FC Amplitude 2 2 5 5 2 25 3 Time (s) B CLUSTER CENTROIDS.6 centroid centroid 2 centroid 3 centroid 4.4 k = 4.2 Cluster validity index 2 4 6 8 2 Number of clusters (k) C STATE TRANSITIONS Distance from centroid 5 4 3 2 5 5 2 25 3 Time (s) centroid centroid 2 centroid 3 centroid 4 estimated state true state Figure S7. Validation of clustering approach with simulated data. (A) Simulated BOLD timeseries for ten nodes are generated under a model of dynamic neural connectivity where 4 states are possible (State 2 repeats). Simulation parameters (TR = 2 s, 48 volumes) are matched to experimental data; connectivity states are modeled after clusters observed in real data (e.g., Figure 5). (B) Windowed covariance matrices are estimated from the simulated timeseries and are subjected to k-means clustering with the L distance metric. The elbow criterion correctly identifies k = 4 clusters, and cluster centroids show high similarity to the true states. (C) The distance of each window to each cluster centroid. The assignment of individual windows to states is very accurate. Distances and state assignments are plotted at the time point corresponding to the center of the window.

2024 © sciencedocbox.com Privacy Policy | Terms of Service | Feedback