Improved estimation of subject-level functional connectivity using full and partial correlation with empirical Bayes shrinkage

Amanda F. Mejia, Mary Beth Nebel, Anita D. Barber, Ann Choe, James J Pekar, Brian S Caffo, Martin Lindquist

Research output: Contribution to journalArticle

Abstract

Reliability of subject-level resting-state functional connectivity (FC) is determined in part by the statistical techniques employed in its estimation. Methods that pool information across subjects to inform estimation of subject-level effects (e.g., Bayesian approaches) have been shown to enhance reliability of subject-level FC. However, fully Bayesian approaches are computationally demanding, while empirical Bayesian approaches typically rely on using repeated measures to estimate the variance components in the model. Here, we avoid the need for repeated measures by proposing a novel measurement error model for FC describing the different sources of variance and error, which we use to perform empirical Bayes shrinkage of subject-level FC towards the group average. In addition, since the traditional intra-class correlation coefficient (ICC) is inappropriate for biased estimates, we propose a new reliability measure denoted the mean squared error intra-class correlation coefficient (ICCMSE) to properly assess the reliability of the resulting (biased) estimates. We apply the proposed techniques to test-retest resting-state fMRI data on 461 subjects from the Human Connectome Project to estimate connectivity between 100 regions identified through independent components analysis (ICA). We consider both correlation and partial correlation as the measure of FC and assess the benefit of shrinkage for each measure, as well as the effects of scan duration. We find that shrinkage estimates of subject-level FC exhibit substantially greater reliability than traditional estimates across various scan durations, even for the most reliable connections and regardless of connectivity measure. Additionally, we find partial correlation reliability to be highly sensitive to the choice of penalty term, and to be generally worse than that of full correlations except for certain connections and a narrow range of penalty values. This suggests that the penalty needs to be chosen carefully when using partial correlations.

Original languageEnglish (US)
Pages (from-to)478-491
Number of pages14
JournalNeuroImage
Volume172
DOIs
StatePublished - May 15 2018

Fingerprint

Bayes Theorem
Connectome
Research Design
Magnetic Resonance Imaging

Keywords

  • Bayesian statistics
  • Connectome
  • Functional connectivity
  • Measurement error
  • Partial correlation
  • Reliability
  • Resting-state fMRI
  • Shrinkage

ASJC Scopus subject areas

  • Neurology
  • Cognitive Neuroscience

Cite this

Improved estimation of subject-level functional connectivity using full and partial correlation with empirical Bayes shrinkage. / Mejia, Amanda F.; Nebel, Mary Beth; Barber, Anita D.; Choe, Ann; Pekar, James J; Caffo, Brian S; Lindquist, Martin.

In: NeuroImage, Vol. 172, 15.05.2018, p. 478-491.

Research output: Contribution to journalArticle

@article{ef3164b995004b6189dfb6670b6ee71c,
title = "Improved estimation of subject-level functional connectivity using full and partial correlation with empirical Bayes shrinkage",
abstract = "Reliability of subject-level resting-state functional connectivity (FC) is determined in part by the statistical techniques employed in its estimation. Methods that pool information across subjects to inform estimation of subject-level effects (e.g., Bayesian approaches) have been shown to enhance reliability of subject-level FC. However, fully Bayesian approaches are computationally demanding, while empirical Bayesian approaches typically rely on using repeated measures to estimate the variance components in the model. Here, we avoid the need for repeated measures by proposing a novel measurement error model for FC describing the different sources of variance and error, which we use to perform empirical Bayes shrinkage of subject-level FC towards the group average. In addition, since the traditional intra-class correlation coefficient (ICC) is inappropriate for biased estimates, we propose a new reliability measure denoted the mean squared error intra-class correlation coefficient (ICCMSE) to properly assess the reliability of the resulting (biased) estimates. We apply the proposed techniques to test-retest resting-state fMRI data on 461 subjects from the Human Connectome Project to estimate connectivity between 100 regions identified through independent components analysis (ICA). We consider both correlation and partial correlation as the measure of FC and assess the benefit of shrinkage for each measure, as well as the effects of scan duration. We find that shrinkage estimates of subject-level FC exhibit substantially greater reliability than traditional estimates across various scan durations, even for the most reliable connections and regardless of connectivity measure. Additionally, we find partial correlation reliability to be highly sensitive to the choice of penalty term, and to be generally worse than that of full correlations except for certain connections and a narrow range of penalty values. This suggests that the penalty needs to be chosen carefully when using partial correlations.",
keywords = "Bayesian statistics, Connectome, Functional connectivity, Measurement error, Partial correlation, Reliability, Resting-state fMRI, Shrinkage",
author = "Mejia, {Amanda F.} and Nebel, {Mary Beth} and Barber, {Anita D.} and Ann Choe and Pekar, {James J} and Caffo, {Brian S} and Martin Lindquist",
year = "2018",
month = "5",
day = "15",
doi = "10.1016/j.neuroimage.2018.01.029",
language = "English (US)",
volume = "172",
pages = "478--491",
journal = "NeuroImage",
issn = "1053-8119",
publisher = "Academic Press Inc.",

}

TY - JOUR

T1 - Improved estimation of subject-level functional connectivity using full and partial correlation with empirical Bayes shrinkage

AU - Mejia, Amanda F.

AU - Nebel, Mary Beth

AU - Barber, Anita D.

AU - Choe, Ann

AU - Pekar, James J

AU - Caffo, Brian S

AU - Lindquist, Martin

PY - 2018/5/15

Y1 - 2018/5/15

N2 - Reliability of subject-level resting-state functional connectivity (FC) is determined in part by the statistical techniques employed in its estimation. Methods that pool information across subjects to inform estimation of subject-level effects (e.g., Bayesian approaches) have been shown to enhance reliability of subject-level FC. However, fully Bayesian approaches are computationally demanding, while empirical Bayesian approaches typically rely on using repeated measures to estimate the variance components in the model. Here, we avoid the need for repeated measures by proposing a novel measurement error model for FC describing the different sources of variance and error, which we use to perform empirical Bayes shrinkage of subject-level FC towards the group average. In addition, since the traditional intra-class correlation coefficient (ICC) is inappropriate for biased estimates, we propose a new reliability measure denoted the mean squared error intra-class correlation coefficient (ICCMSE) to properly assess the reliability of the resulting (biased) estimates. We apply the proposed techniques to test-retest resting-state fMRI data on 461 subjects from the Human Connectome Project to estimate connectivity between 100 regions identified through independent components analysis (ICA). We consider both correlation and partial correlation as the measure of FC and assess the benefit of shrinkage for each measure, as well as the effects of scan duration. We find that shrinkage estimates of subject-level FC exhibit substantially greater reliability than traditional estimates across various scan durations, even for the most reliable connections and regardless of connectivity measure. Additionally, we find partial correlation reliability to be highly sensitive to the choice of penalty term, and to be generally worse than that of full correlations except for certain connections and a narrow range of penalty values. This suggests that the penalty needs to be chosen carefully when using partial correlations.

AB - Reliability of subject-level resting-state functional connectivity (FC) is determined in part by the statistical techniques employed in its estimation. Methods that pool information across subjects to inform estimation of subject-level effects (e.g., Bayesian approaches) have been shown to enhance reliability of subject-level FC. However, fully Bayesian approaches are computationally demanding, while empirical Bayesian approaches typically rely on using repeated measures to estimate the variance components in the model. Here, we avoid the need for repeated measures by proposing a novel measurement error model for FC describing the different sources of variance and error, which we use to perform empirical Bayes shrinkage of subject-level FC towards the group average. In addition, since the traditional intra-class correlation coefficient (ICC) is inappropriate for biased estimates, we propose a new reliability measure denoted the mean squared error intra-class correlation coefficient (ICCMSE) to properly assess the reliability of the resulting (biased) estimates. We apply the proposed techniques to test-retest resting-state fMRI data on 461 subjects from the Human Connectome Project to estimate connectivity between 100 regions identified through independent components analysis (ICA). We consider both correlation and partial correlation as the measure of FC and assess the benefit of shrinkage for each measure, as well as the effects of scan duration. We find that shrinkage estimates of subject-level FC exhibit substantially greater reliability than traditional estimates across various scan durations, even for the most reliable connections and regardless of connectivity measure. Additionally, we find partial correlation reliability to be highly sensitive to the choice of penalty term, and to be generally worse than that of full correlations except for certain connections and a narrow range of penalty values. This suggests that the penalty needs to be chosen carefully when using partial correlations.

KW - Bayesian statistics

KW - Connectome

KW - Functional connectivity

KW - Measurement error

KW - Partial correlation

KW - Reliability

KW - Resting-state fMRI

KW - Shrinkage

UR - http://www.scopus.com/inward/record.url?scp=85041699592&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85041699592&partnerID=8YFLogxK

U2 - 10.1016/j.neuroimage.2018.01.029

DO - 10.1016/j.neuroimage.2018.01.029

M3 - Article

C2 - 29391241

AN - SCOPUS:85041699592

VL - 172

SP - 478

EP - 491

JO - NeuroImage

JF - NeuroImage

SN - 1053-8119

ER -