Abstract
High-throughput experimental methods in neuroscience have led to an explosion of techniques for measuring complex interactions and multi-dimensional patterns. However, whether sophisticated measures of emergent phenomena can be traced back to simpler, low-dimensional statistics is largely unknown. To explore this question, we examined resting-state functional magnetic resonance imaging (rs-fMRI) data using complex topology measures from network neuroscience. Here we show that spatial and temporal autocorrelation are reliable statistics that explain numerous measures of network topology. Surrogate time series with subject-matched spatial and temporal autocorrelation capture nearly all reliable individual and regional variation in these topology measures. Network topology changes during aging are driven by spatial autocorrelation, and multiple serotonergic drugs causally induce the same topographic change in temporal autocorrelation. This reductionistic interpretation of widely used complexity measures may help link them to neurobiology.
This is a preview of subscription content, access via your institution
Access options
Access Nature and 54 other Nature Portfolio journals
Get Nature+, our best-value online-access subscription
$29.99 /Â 30Â days
cancel any time
Subscribe to this journal
Receive 12 print issues and online access
$209.00 per year
only $17.42 per issue
Buy this article
- Purchase on SpringerLink
- Instant access to full article PDF
Prices may be subject to local taxes which are calculated during checkout
Similar content being viewed by others
Data availability
The HCP data are available at https://www.humanconnectome.org/study/hcp-young-adult. The Yale-TRT data are available at http://fcon_1000.projects.nitrc.org/indi/retro/yale_trt.html. The Cam-CAN data are available at https://www.cam-can.org/index.php?content=dataset. The LSD data and psilocybin data are available upon reasonable request.
Code availability
We prepared a software package that allows the principal analyses in this paper to be performed quickly and easily. The âspatiotemporalâ Python package, which can be installed through pip or downloaded at https://github.com/murraylab/spatiotemporal, offers a more user-friendly way of applying the analyses described here.
The raw source code used to perform the analyses in this paper can be downloaded at https://github.com/murraylab/spatial_and_temporal_paper. Code was implemented using the standard Python stack99,100 and other libraries101,102. Source code was checked for correctness using software verification techniques103.
References
Buckner, R. L., Krienen, F. M. & Yeo, B. T. T. Opportunities and limitations of intrinsic functional connectivity MRI. Nat. Neurosci. 16, 832â837 (2013).
Smith, S. M. et al. Functional connectomics from resting-state fMRI. Trends Cogn. Sci. 17, 666â682 (2013).
Fornito, A., Zalesky, A. & Bullmore, E. Fundamentals of Brain Network Analysis (Academic Press, 2016).
Rubinov, M. & Sporns, O. Complex network measures of brain connectivity: uses and interpretations. Neuroimage 52, 1059â1069 (2010).
Betzel, R. F. et al. Changes in structural and functional connectivity among resting-state networks across the human lifespan. Neuroimage 102, 345â357 (2014).
Damoiseaux, J. S. Effects of aging on functional and structural brain connectivity. Neuroimage 160, 32â40 (2017).
Preller, K. H. et al. Changes in global and thalamic brain connectivity in LSD-induced altered states of consciousness are attributable to the 5-HT2A receptor. eLife 7, e35082 (2018).
Luppi, A. I. et al. LSD alters dynamic integration and segregation in the human brain. Neuroimage 227, 117653 (2021).
Murphy, K., Birn, R. M. & Bandettini, P. A. Resting-state fMRI confounds and cleanup. Neuroimage 80, 349â359 (2013).
Friston, K. J. et al. Analysis of fMRI time-series revisited. Neuroimage 2, 45â53 (1995).
Burt, J. B., Helmer, M., Shinn, M., Anticevic, A. & Murray, J. D. Generative modeling of brain maps with spatial autocorrelation. Neuroimage 220, 117038 (2020).
Huang, Z., Liu, X., Mashour, G. A. & Hudetz, A. G. Timescales of intrinsic BOLD signal dynamics and functional connectivity in pharmacologic and neuropathologic states of unconsciousness. J. Neurosci. 38, 2304â2317 (2018).
Sethi, S. S., Zerbi, V., Wenderoth, N., Fornito, A. & Fulcher, B. D. Structural connectome topology relates to regional BOLD signal dynamics in the mouse brain. Chaos 27, 047405 (2017).
Fallon, J. et al. Timescales of spontaneous fMRI fluctuations relate to structural connectivity in the brain. Netw. Neurosci. 4, 788â806 (2020).
Fascianelli, V., Tsujimoto, S., Marcos, E. & Genovesio, A. Autocorrelation structure in the macaque dorsolateral, but not orbital or polar, prefrontal cortex predicts response-coding strength in a visually cued strategy task. Cereb. Cortex 29, 230â241 (2019).
Arbabshirani, M. R. et al. Autoconnectivity: A new perspective on human brain function. J. Neurosci. Methods 323, 68â76 (2019).
Honey, C. J. et al. Slow cortical dynamics and the accumulation of information over long timescales. Neuron 76, 423â434 (2012).
Song, H. F., Kennedy, H. & Wang, X.-J. Spatial embedding of structural similarity in the cerebral cortex. Proc. Natl Acad. Sci. USA 111, 16580â16585 (2014).
Glasser, M. F. et al. A multi-modal parcellation of human cerebral cortex. Nature 536, 171â178 (2016).
Noble, S. et al. Influences on the testâretest reliability of functional connectivity MRI and its relationship with behavioral utility. Cereb. Cortex 27, 5415â5429 (2017).
Shafto, M. A. et al. The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) study protocol: a cross-sectional, lifespan, multidisciplinary examination of healthy cognitive ageing. BMC Neurol. 14, 204 (2014).
Li, C., Wang, H., de Haan, W., Stam, C. J. & Mieghem, P. V. The correlation of metrics in complex networks with applications in functional brain networks. J. Stat. Mech. 2011, P11018 (2011).
Vértes, P. E. et al. Simple models of human brain functional networks. Proc. Natl Acad. Sci. USA 109, 5868â5873 (2012).
Roberts, J. A. et al. The contribution of geometry to the human connectome. Neuroimage 124, 379â393 (2016).
Cicchetti, D. V. & Sparrow, S. A. Developing criteria for establishing interrater reliability of specific items: applications to assessment of adaptive behavior. Am. J. Ment. Defic. 86, 127â137 (1981).
Koo, T. K. & Li, M. Y. A guideline of selecting and reporting intraclass correlation coefficients for reliability research. J. Chiropr. Med. 15, 155â163 (2016).
Murray, J. D. et al. A hierarchy of intrinsic timescales across primate cortex. Nat. Neurosci. 17, 1661â1663 (2014).
Raut, R. V., Snyder, A. Z. & Raichle, M. E. Hierarchical dynamics as a macroscopic organizing principle of the human brain. Proc. Natl Acad. Sci. USA 117, 20890â20897 (2020).
Arbabshirani, M. R. et al. Impact of autocorrelation on functional connectivity. Neuroimage 102, 294â308 (2014).
Zalesky, A., Fornito, A. & Bullmore, E. On the use of correlation as a measure of network connectivity. Neuroimage 60, 2096â2106 (2012).
Afyouni, S., Smith, S. M. & Nichols, T. E. Effective degrees of freedom of the Pearsonâs correlation coefficient under autocorrelation. Neuroimage 199, 609â625 (2019).
Shafiei, G. et al. Topographic gradients of intrinsic dynamics across neocortex. eLife 9, e62116 (2020).
Keitel, A. & Gross, J. Individual human brain areas can be identified from their characteristic spectral activation fingerprints. PLoS Biol. 14, e1002498 (2016).
Finn, E. S. et al. Functional connectome fingerprinting: identifying individuals using patterns of brain connectivity. Nat. Neurosci. 18, 1664â1671 (2015).
Waller, L. et al. Evaluating the replicability, specificity, and generalizability of connectome fingerprints. Neuroimage 158, 371â377 (2017).
Baria, A. et al. Linking human brain local activity fluctuations to structural and functional network architectures. Neuroimage 73, 144â155 (2013).
Noble, S., Scheinost, D. & Constable, R. T. A decade of testâretest reliability of functional connectivity: a systematic review and meta-analysis. Neuroimage 203, 116157 (2019).
Cole, M. W., Yang, G. J., Murray, J. D., RepovÅ¡, G. & Anticevic, A. Functional connectivity change as shared signal dynamics. J. Neurosci. Methods 259, 22â39 (2016).
Achard, S. A resilient, low-frequency, small-world human brain functional network with highly connected association cortical hubs. J. Neurosci. 26, 63â72 (2006).
Betzel, R. F. et al. Generative models of the human connectome. Neuroimage 124, 1054â1064 (2016).
Sullivan, E. V. & Pfefferbaum, A. Diffusion tensor imaging and aging. Neurosci. Biobehav. Rev. 30, 749â61 (2006).
Watanabe, T., Rees, G. & Masuda, N. Atypical intrinsic neural timescale in autism. eLife 8, e42256 (2019).
Preller, K. H. et al. Psilocybin induces time-dependent changes in global functional connectivity. Biol. Psychiatry 88, 197â207 (2020).
Persson, B., Heykants, J. & Hedner, T. Clinical pharmacokinetics of ketanserin. Clin. Pharmacokinet. 20, 263â279 (1991).
Grady, C. L. & Garrett, D. D. Understanding variability in the BOLD signal and why it matters for aging. Brain Imaging Behav. 8, 274â283 (2013).
Zerbi, V. et al. Rapid reconfiguration of the functional connectome after chemogenetic locus coeruleus activation. Neuron 103, 702â718 (2019).
Shafiei, G. et al. Dopamine signaling modulates the stability and integration of intrinsic brain networks. Cereb. Cortex 29, 397â409 (2018).
Turchi, J. et al. The basal forebrain regulates global resting-state fMRI fluctuations. Neuron 97, 940â952 (2018).
Davey, C. E., Grayden, D. B., Egan, G. F. & Johnston, L. A. Filtering induces correlation in fMRI resting state data. Neuroimage 64, 728â740 (2013).
Cantwell, G. T. et al. Thresholding normally distributed data creates complex networks. Phys. Rev. E 101, 062302 (2020).
Ito, T., Hearne, L. J. & Cole, M. W. A cortical hierarchy of localized and distributed processes revealed via dissociation of task activations, connectivity changes, and intrinsic timescales. Neuroimage 221, 117141 (2020).
Chen, J. E. et al. Resting-state âphysiological networksâ. Neuroimage 213, 116707 (2020).
Drew, P. J., Mateo, C., Turner, K. L., Yu, X. & Kleinfeld, D. Ultra-slow oscillations in fMRI and resting-state connectivity: neuronal and vascular contributions and technical confounds. Neuron 107, 782â804 (2020).
Baracchini, G. et al. Inter-regional BOLD signal variability is an organizational feature of functional brain networks. Neuroimage 237, 118149 (2021).
Kriegeskorte, N., Cusack, R. & Bandettini, P. How does an fMRI voxel sample the neuronal activity pattern: compact-kernel or complex spatiotemporal filter? Neuroimage 49, 1965â1976 (2010).
West, K. L. et al. BOLD hemodynamic response function changes significantly with healthy aging. Neuroimage 188, 198â207 (2019).
Cohen, Z., Bonvento, G., Lacombe, P. & Hamel, E. Serotonin in the regulation of brain microcirculation. Prog. Neurobiol. 50, 335â362 (1996).
Geerligs, L., Tsvetanov, K. A. & Henson, R. N. Challenges in measuring individual differences in functional connectivity using fMRI: the case of healthy aging. Hum. Brain Mapp. 38, 4125â4156 (2017).
Honey, C. J. et al. Predicting human resting-state functional connectivity from structural connectivity. Proc. Natl Acad. Sci. USA 106, 2035â2040 (2009).
Braga, R. M. & Buckner, R. L. Parallel interdigitated distributed networks within the individual estimated by intrinsic functional connectivity. Neuron 95, 457â471 (2017).
Kong, R. et al. Individual-specific areal-level parcellations improve functional connectivity prediction of behavior. Cereb. Cortex 31, 4477â4500 (2021).
Balthazar, M. L. F., de Campos, B. M., Franco, A. R., Damasceno, B. P. & Cendes, F. Whole cortical and default mode network mean functional connectivity as potential biomarkers for mild Alzheimerâs disease. Psychiatry Res. 221, 37â42 (2014).
Wiedermann, M., Donges, J. F., Kurths, J. & Donner, R. V. Spatial network surrogates for disentangling complex system structure from spatial embedding of nodes. Phys. Rev. E 93, 042308 (2016).
DemirtaÅ, M. et al. Hierarchical heterogeneity across human cortex shapes large-scale neural dynamics. Neuron 101, 1181â1194 (2019).
Morgan, S. E., Achard, S., Termenon, M., Bullmore, E. T. & Vértes, P. E. Low-dimensional morphospace of topological motifs in human fMRI brain networks. Netw. Neurosci. 2, 285â302 (2018).
Robinson, E. C. et al. MSM: a new flexible framework for multimodal surface matching. Neuroimage 100, 414â426 (2014).
Friedman, L., Glover, G. H., Krenz, D. & Magnotta, V. Reducing inter-scanner variability of activation in a multicenter fMRI study: role of smoothness equalization. Neuroimage 32, 1656â1668 (2006).
Scheinost, D., Papademetris, X. & Constable, R. T. The impact of image smoothness on intrinsic functional connectivity and head motion confounds. Neuroimage 95, 13â21 (2014).
Shen, X., Tokoglu, F., Papademetris, X. & Constable, R. Groupwise whole-brain parcellation from resting-state fMRI data for network node identification. Neuroimage 82, 403â415 (2013).
Taylor, J. R. et al. The Cambridge Centre for Ageing and Neuroscience (Cam-CAN) data repository: structural and functional MRI, MEG, and cognitive data from a cross-sectional adult lifespan sample. Neuroimage 144, 262â269 (2017).
Tzourio-Mazoyer, N. et al. Automated anatomical labeling of activations in SPM using a macroscopic anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15, 273â289 (2002).
Bullmore, E. et al. Colored noise and computational inference in neurophysiological (fMRI) time series analysis: resampling methods in time and wavelet domains. Hum. Brain Mapp. 12, 61â78 (2001).
Wagenmakers, E.-J., Farrell, S. & Ratcliff, R. Estimation and interpretation of 1/fα noise in human cognition. Psychon. Bull. Rev. 11, 579â615 (2004).
Maxim, V. et al. Fractional Gaussian noise, functional MRI and Alzheimerâs disease. Neuroimage 25, 141â158 (2005).
Achard, S., Bassett, D. S., Meyer-Lindenberg, A. & Bullmore, E. Fractal connectivity of long-memory networks. Phys. Rev. E 77, 036104 (2008).
Sela, R. J. & Hurvich, C. M. Computationally efficient methods for two multivariate fractionally integrated models. J. Time Ser. Anal. 30, 631â651 (2009).
Achard, S. & Gannaz, I. Multivariate wavelet whittle estimation in long-range dependence. J. Time Ser. Anal. 37, 476â512 (2015).
Lu, Z. Analysis of stationary and non-stationary long memory processes: estimation, applications and forecast. PhD thesis, Ãcole normale supérieure de Cachan - ENS Cachan (2019); https://theses.hal.science/tel-00422376/document
Hassani, H., Leonenko, N. & Patterson, K. The sample autocorrelation function and the detection of long-memory processes. Phys. A 391, 6367â6379 (2012).
Scheffer, M. et al. Early-warning signals for critical transitions. Nature 461, 53â59 (2009).
Bullmore, E. & Sporns, O. Complex brain networks: graph theoretical analysis of structural and functional systems. Nat. Rev. Neurosci. 10, 186â198 (2009).
Mantegna, R. Hierarchical structure in financial markets. Eur. Phys. J. B 11, 193â197 (1999).
Kruskal, J. B. On the shortest spanning subtree of a graph and the traveling salesman problem. Proc. Am. Math. Soc. 7, 48â48 (1956).
Termenon, M., Jaillard, A., Delon-Martin, C. & Achard, S. Reliability of graph analysis of resting state fMRI using test-retest dataset from the Human Connectome Project. Neuroimage 142, 172â187 (2016).
McGraw, K. O. & Wong, S. P. Forming inferences about some intraclass correlation coefficients. Psychol. Methods 1, 30â46 (1996).
Horien, C. et al. Considering factors affecting the connectome-based identification process: comment on Waller et al. Neuroimage 169, 172â175 (2018).
Zang, Y., Jiang, T., Lu, Y., He, Y. & Tian, L. Regional homogeneity approach to fMRI data analysis. Neuroimage 22, 394â400 (2004).
Lin, L. I.-K. A concordance correlation coefficient to evaluate reproducibility. Biometrics 45, 255 (1989).
Newman, M. E. J. Assortative mixing in networks. Phys. Rev. Lett. 89, 208701 (2002).
Latora, V. & Marchiori, M. Efficient behavior of small-world networks. Phys. Rev. Lett. 87, 198701 (2001).
Newman, M. E. J. & Girvan, M. Finding and evaluating community structure in networks. Phys. Rev. E Stat. Nonlin. Soft Matter Phys. 9, 026113 (2004).
Brandes, U. A faster algorithm for betweenness centrality. J. Math. Sociol. 25, 163â177 (2001).
Timmer, J. & König, M. On generating power law noise. Astron. Astrophys. 300, 707â710 (1995).
Davies, P. I. & Higham, N. J. Numerically stable generation of correlation matrices and their factors. BIT Numer. Math. 40, 640â651 (2000).
Theiler, J., Eubank, S., Longtin, A., Galdrikian, B. & Farmer, J. D. Testing for nonlinearity in time series: the method of surrogate data. Physica D 58, 77â94 (1992).
Maslov, S. & Sneppen, K. Specificity and stability in topology of protein networks. Science 296, 910â913 (2002).
Storn, R. & Price, K. Differential evolutionâa simple and efficient heuristic for global optimization over continuous spaces. J. Glob. Optim. 11, 341â359 (1997).
Bassett, D. S. & Sporns, O. Network neuroscience. Nat. Neurosci. 20, 353â364 (2017).
Harris, C. R. et al. Array programming with NumPy. Nature 585, 357â362 (2020).
Virtanen, P. et al. SciPy 1.0: fundamental algorithms for scientific computing in python. Nat. Methods 17, 261â272 (2020).
LaPlante, R. A., Douw, L., Tang, W. & Stufflebeam, S. M. The connectome visualization utility: software for visualization of human brain networks. PLoS ONE 9, e113838 (2014).
Vallat, R. Pingouin: statistics in Python. J. Open Source Softw. 3, 1026 (2018).
Shinn, M. Refinement type contracts for verification of scientific investigative software. in Verified Software. Theories, Tools, and Experiments (eds Chakraborty, S. & Navas, J. A.) 143â160 (Springer, 2020).
Acknowledgements
We thank J. Burt for help with plotting; A. Howell for assistance with HCP data; and T. Ito and J. C. C. Vila for helpful discussions. Funding was provided by European Molecular Biology Organization grant ALTF 712-2021, the Winston Churchill Foundation of the United States and the Gruber Foundation (M.S.); National Institute of Mental Health (NIMH) grant K00MH122372 (S.N.); the Swiss National Science Foundation (P2ZHP1\161626) (K.H.P.); Agence Nationale de la Recherche (ANR-20-NEUC-003-01) and MIAI@Grenoble Alpes (ANR-19-P3IA-0003) (S.A.); the Usona Institute (2015-2056), the Heffter Research Institute (1-190420), the Swiss Neuromatrix Foundation (2015-103 and 2016-0111) and the Swiss National Science Foundation under the framework of Neuron Cofund (01EW1908) (F.X.V.); the National Institute for Health and Care Research (NIHR) (Senior Investigator Award) and the NIHR Cambridge Biomedical Research Centre (E.T.B.); SFARI (Pilot Award) (A.A. and J.D.M.); and the NIMH (R01MH112746) (J.D.M.). Data collection and sharing for this project were provided, in part, by the Cambridge Centre for Ageing and Neuroscience (Cam-CAN). Cam-CAN funding was provided by the UK Biotechnology and Biological Sciences Research Council (grant BB/H008217/1), together with support from the UK Medical Research Council and the University of Cambridge.
Author information
Authors and Affiliations
Contributions
M.S. and E.T.B. conceived the research. M.S. designed the experiments. M.S. and A.H. performed the experiments. M.S. and J.D.M. analyzed and interpreted results. K.H.P., J.L.J., F.M., S.N., D.S., R.T.C., J.H.K., F.X.V. and A.A. contributed data, methodology and resources. M.S., L.T. and S.A. performed the mathematical analysis. D.L., E.T.B. and J.D.M. provided supervision and funding. M.S., A.H. and L.T. wrote the first draft of the manuscript. All authors edited, revised and approved the manuscript.
Corresponding author
Ethics declarations
Competing interests
K.H.P. is currently an employee of Hoffmann-La Roche. J.H.K. has consulting agreements (less than $5,000 per year) with the following: Aptinyx; Atai Life Sciences; AstraZeneca Pharmaceuticals; Biogen; Biomedisyn Corporation; Bionomics; Boehringer Ingelheim; Cadent Therapeutics; Clexio Bioscience; COMPASS Pathways; Concert Pharmaceuticals; Epiodyne; EpiVario; Greenwich Biosciences; Heptares Therapeutics; Janssen Research & Development; Jazz Pharmaceuticals; Otsuka America Pharmaceutical; Perception Neuroscience Holdings; Spring Care; Sunovion Pharmaceuticals; Takeda Industries; and Taisho Pharmaceutical Company. J.H.K. serves on the scientific advisory boards of Biohaven Pharmaceuticals; BioXcel Therapeutics (Clinical Advisory Board); Cadent Therapeutics (Clinical Advisory Board); Cerevel Therapeutics; EpiVario; Eisai; Jazz Pharmaceuticals; Lohocla Research Corporation; Novartis Pharmaceuticals Corporation; PsychoGenics; Neumora Therapeutics; Tempero Bio; and Terran Biosciences. J.H.K. is on the board of directors of Freedom Biosciences. J.H.K. has stock and/or stock options in Biohaven Pharmaceuticals; Sage Pharmaceuticals; Spring Care; Biohaven Pharmaceuticals Medical Sciences; EpiVario; Neumora Therapeutics; Terran Biosciences; and Tempero Bio. J.H.K. is editor of Biological Psychiatry with income greater than $10,000. D.L. is a co-founder of Neurogazer. A.A. and J.D.M. are co-founders of Manifest Technologies and serve on the technical advisory board of Neumora Therapeutics. E.T.B. serves on the scientific advisory board of Sosei Heptares and as a consultant for GlaxoSmithKline. The remaining authors declare no competing interests.
Peer review
Peer review information
Nature Neuroscience thanks Janine Bijsterbosch, Javier Gonzalez-Castillo, Lucina Uddin and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Additional information
Publisherâs note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data
Extended Data Fig. 1 SA and TA are important features of rs-fMRI.
Data from the HCP-GSR (a-h, N=850 subjects), Yale-TRT (i-p, N=12 subjects, 4 sessions each), and Cam-CAN (q-t, N=646 subjects) datasets. For all subpanels, unless otherwise indicated, * indicates pâ<â0.05 and ** indicates Pâ<â0.01 on a two-sided test, and boxplots indicate the median, first/third quartiles, and range, with outliers hidden for visualization. (a,i,q) Correlation across all subjects in the HCP dataset, with Bonferroni FWER corrected two-sided P-values. (b,i) Test-retest reliability of graph metrics, quantified by intraclass correlation coefficient (ICC). Error bars indicate 95% CI. Following each bar is a string of three characters indicating significance: the first indicates significantly less than SA-λ, the second SA-\(\infty\), and the third global TA-Î1, where # indicates Pâ<â.01, + indicates Pâ<â.05, and â indicates Pâ>â.05, by a one-sided bootstrap resampling procedure. Inset scatterplots show correlation across subjects for two example sessions. (c,k,r) Correlation across subjects between graph metrics and SA-λ, SA-\(\infty\), or global TA-Î1. âSA-λ + SA-\(\infty\)â and âallâ indicate a cross-validated linear model with two or three terms, respectively, using Bonferroni FWER corrected two-sided P-values. (d,l,s) The brain map depicting regional TA-Î1, averaged across all subjects. (e,m) Distribution of reliability for each brain region. (f,n) Mean fraction of subjects correctly identified by a fingerprinting analysis. Points indicate identification performance on each of six possible test-retest pairs from the four sessions. (g,o,t) Correlation across regions of regional TA-Î1 with nodal graph metrics for each subject. (k) The regional TA-Î1 for each region plotted against its reliability.
Extended Data Fig. 2 Correlation of motion and parcel size with spatial and temporal autocorrelation.
(a-c) For each dataset, the mean framewise displacement (âmotionâ) is compared to (a) SA-λ, (b) SA-\(\infty\), and (c) TA-Î1, with inset Spearman correlation (rs). (d) The average regional TA-Î1 across all subjects is plotted against the regionâs parcel size, with inset Spearman correlation. Parcel size is measured in surface area for HCP and HCP-GSR, and number of voxels for Yale-TRT and Cam-CAN. (e) Subjectâs mean framewise displacement (âmotionâ) is plotted against the subjectâs age. * indicates Spearman correlation two-sided Pâ<â.05, and ** indicates Pâ<â.01. HCP: N=883, HCP-GSR: N=850, Yale-TRT: N=12 subjects, 4 sessions each, Cam-CAN: N=646, LSD: N=24, Psilocybin: N=23 subjects.
Extended Data Fig. 3 TA-Î1 captures individual variation in long memory dynamics.
(a-c,h-j) TA-Î1, the first lag term in the ACF, is more reliable than higher lag terms. (a,h) For each lag k, we computed the reliability of the corresponding term in the ACF, ACFx(k), as measured by ICC. The median reliability across brain regions decreased as k increased, and regional TA-Î1 (that is, ACFx(1)) maximized median reliability. Additionally, d had similar reliability to regional TA-Î1. (b,i) To confirm this occurs for each brain region individually, we found the difference in reliability between the ACFx(1) and ACFx(k) (ÎICC), which also decreased across lags. d was similar to regional TA-Î1. (c,j) Global TA-Î1 reliability also decreased with increasing lag, and d averaged across regions had similar reliability to global TA-Î1, as measured by ICC. Error bars indicate 95% confidence interval. Thus, the reliability at short compared to long time lags, and the similarity to d, motivate a prioritization for TA-Î1 over higher lags. (d,k,o) TA-Î1 is predictive of higher terms of the ACF. TA-Î1 was used to predict higher ACF terms in a regression model (see Methods). Mean cross-validated R2 is shown, where error bars (sometimes hidden under the line) indicate maximum and minimum cross-validated R2. (e-g,l-n,p-r) Individual variation in d, the long-memory or fractional integration term from an ARFIMA(0,d,0) model, can be captured by individual variation in TA-Î1. (e,l,p) regional TA-Î1 averaged across subjects is highly correlated with regional d averaged across subjects. (e: P=0, l: P=10â260, p: P=.10) (f,m,q) Global TA-Î1 is highly correlated with d averaged across regions within a subject. (f: P=0, m: P=10â101 q: 10â61) (g,n,r) Without averaging, across regions and subjects, d is correlated with TA-Î1. (g: P=0, n: P=0, r: P=0) For all figures, unless otherwise indicated, rs indicates Spearman correlation, where * indicates two-sided Pâ<â.05, and ** indicates Pâ<â.01, N=883 subjects.
Extended Data Fig. 4 Relationship between regional TA-Î1 and regional homogeneity.
Comparisons are shown for the HCP (a-c, N=883) and Yale-TRT (d-f, N=12, 4 sessions each) datasets. (a,d) The global TA-Î1 averaged across subjects, compared to the regional homogeneity (ReHo) averaged across subjects. Each point represents a parcel. ** indicates two-sided Spearman correlation Pâ<â.01. (a: P=10â18, d: P=10â39) (b,e) Distribution of correlations of global TA-Î1 and regional homogeneity. Compare to Fig. 1j. (c,f) Average fingerprinting performance of regional homogeneity, compared to global TA-Î1 and chance. Points are overlaid for each pair of datasets compared. Compare to Fig. 1i.
Extended Data Fig. 5 Correlation of model and data graph metrics for all models.
For each model, each subjectâs empirical graph metrics are plotted against model graph metrics for metrics from Fig. 2e. Spearman correlation (rs) and Linâs concordance (Lin) are inset. * indicates Spearman correlation two-sided Pâ<â.05, and ** indicates Pâ<â.01, N=883 subjects. Zalesky matching operates at the level of the FC matrix, and thus, TA-Î1 could not be computed. Edge reshuffle operates at the level of the graph, preventing any of these measures from being computed.
Extended Data Fig. 6 Schematics of all models.
All models considered are described in their corresponding schematics. The spatiotemporal model is shown in Fig. 2a.
Extended Data Fig. 7 Model fitting under geodesic distance.
We analyzed the data and model using geodesic instead of Euclidean distance in the right hemisphere. (a-b) We computed SA-λ and SA-\(\infty\) under geodesic distance for the right hemisphere. The geodesic and Euclidean distances lead to highly correlated measurements for SA-λ (a) and SA-\(\infty\) (b). * indicates Spearman correlation two-sided Pâ<â.05, and ** indicates Pâ<â.01. (a: P=0, b: P=0) (c) Comparison of reliability of graph measures to SA-λ and SA-\(\infty\) when computed using geodesic distance, as measured using ICC. Following each bar is a string of three characters: the first indicates significantly less than SA-λ, the second SA-\(\infty\), and the third global TA-Î1, where # indicates Pâ<â.01, + indicates Pâ<â.05, and - indicates Pâ>â.05 by a bootstrap resampling procedure on a one-sided test. Compare to Fig. 1d. (d) Correlation of graph metrics with SA-λ, SA-\(\infty\), a linear model of both (âSA-λ + SA-\(\infty\)â), and a linear model incorporating both of these plus global TA-Î1 (âallâ), when computed using geodesic distance instead of Euclidean distance. Compare to Fig. 1e. * indicates Spearman correlation two-sided Pâ<â.05, and ** indicates Pâ<â.01. (e-g) The spatiotemporal model and all comparison models were fit using geodesic instead of Euclidean distance. (e) The similarity of graph metrics in the model and data, as measured by Linâs concordance, under geodesic distance. Compare to Fig. 2e. (f) The similarity of degree distribution in the model and data under geodesic distance. Compare to Fig. 2f. (g) The similarity of nodal graph metrics in the model and data under geodesic distance. Compare to Fig. 2g. For all panels, N=883 subjects.
Extended Data Fig. 8 Comparison of model fits for all datasets.
(a,d,g) Linâs concordance between model and data for each model. Bars represent the mean across four (HCP-GSR), six (Yale-TRT), or one (Cam-CAN) scanning sessions, and points indicate the Linâs concordance between the model and data for each session. For comparison, black indicates Linâs concordance between separate sessions from the same subject in the HCP-GSR and Yale-TRT datasets, where dots indicate pairs of sessions. (b,e,h) Log-log degree distribution for each model compared to the data (black). (c,f,i) Distribution of Linâs concordance of nodal metrics between model and data for each region. Boxplots show median, first/third quartiles, and range, with outliers hidden for visualization, N=883 subjects. Compare to Fig. 2e-g. Statistics of these fits can be found in Table S1.
Extended Data Fig. 9 Relationship between the spatiotemporal model and economical clustering (EC) model.
(a) For three given values of the EC cluster parameter (γ), the distance parameter (η) was varied and the TA-\({\Delta }_{1}^{{{{\rm{gen}}}}}\) parameter of the best fit spatiotemporal model is shown. (b) For three given values of the EC distance parameter, the EC cluster parameter was varied and the SA-λgen parameter of the best fit spatiotemporal model is shown. Error bars show standard error across 10 simulations of the EC model. (e-g) The Homogeneous variant of the model was simulated for different values of parameters SA-λgen and TA-\({\Delta }_{1}^{{{{\rm{gen}}}}}\). For each timeseries metric (e), weighted graph metric (f), and unweighted graph metric (g), the metric value is plotted as a heatmap. (h) The economical clustering (EC) model was simulated for different values of its distance and cluster parameters. The value of each of the graph metrics is plotted as a heatmap. For all panels, N=883 subjects.
Extended Data Fig. 10 SA and TA under serotonergic modulation.
(a-c,e) Analysis of scans after administration of LSD (top) or psilocybin (bottom). (a) Correlation of graph metrics across all subjects, with Bonferroni FWER corrected two-sided P-values, * indicates Pâ<â.05, ** indicates Pâ<â.01. Compare to Figure 1b. (b) Correlation across subjects between graph metrics and SA-λ, SA-\(\infty\), or global TA-Î1. âSA-λ+SA-\(\infty\)â and âallâ columns indicate a linear model with leave-one-out cross-validation and Bonferroni FWER corrected two-sided P-values, * indicates Pâ<â.05, ** indicates Pâ<â.01. Compare to Figure 1e. (c) The brain map depicting regional TA-Î1, averaged across all subjects. (d) Correlation of regional TA-Î1 in HCP with those under LSD (top) or psilocybin (bottom). Spearman correlationâ>â.85, two-sided significance Pâ<â.0001 of the correlation of HCP and LSD or psilocybin regional TA-Î1 determined with SA-preserving scrambles11. Gray points indicate the placebo condition. (e) Correlation across regions of regional TA-Î1 with nodal graph metrics for each subject. Boxplots show the median, first/third quartiles, and outlier-excluded minimum and maximum of the distribution Compare to Figure 1j. (f-j) Difference between drug and control across subjects for LSD (blue), psilocybin (pink), and LSD with ketanserin (green), for both early and late scans. Metrics plotted are TA as quantified by TA-Î1 (f), SA as quantified by SA-λ and SA-\(\infty\) (g), the residual of a regression model using motion to predict TA-Î1, subject motion as quantified by mean framewise displacement (i), and graph metrics (j). * indicates Pâ<â.05, ** indicates Pâ<â.01, two-sided Wilcoxon sign-rank test. Boxplots show the median, first/third quartiles, and outlier-excluded minimum and maximum of the distribution. For all panels, LSD: N=24, Psilocybin: N=23 subjects.
Supplementary information
Supplementary Information
Supplementary Notes 1 and 2 and Supplementary Table 1.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Shinn, M., Hu, A., Turner, L. et al. Functional brain networks reflect spatial and temporal autocorrelation. Nat Neurosci 26, 867â878 (2023). https://doi.org/10.1038/s41593-023-01299-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1038/s41593-023-01299-3
This article is cited by
-
The application value of Rs-fMRI-based machine learning models for differentiating mild cognitive impairment from Alzheimer's disease: a systematic review and meta-analysis
Neurological Sciences (2024)
-
Complex topology meets simple statistics
Nature Neuroscience (2023)