Namgil Lee1, Jong-Min Kim2. 1. Department of Information Statistics, Kangwon National University, Chuncheon, Gangwon 24341, South Korea. Electronic address: namgil.lee@kangwon.ac.kr. 2. Statistics Discipline, Division of Sciences and Mathematics, University of Minnesota-Morris, Morris, MN 56267-2134, USA. Electronic address: jongmink@morris.umn.edu.
Abstract
BACKGROUND: Recent studies showed that functional connectivity (FC) in the human brain is not static but can dynamically change across time within time scales of seconds to minutes. NEW METHOD: This study introduces a new statistical method called the copula time-varying correlation for dynamic functional connectivity (dFC) analysis from functional magnetic resonance imaging (fMRI) data. RESULTS: Compared to other state-of-the-art statistical measures of dynamic correlation such as the dynamic conditional correlation (DCC), the proposed method can be effectively applied to data having asymmetric or non-normal distributions. COMPARISON WITH EXISTING METHODS: Numerical simulations were conducted under various kinds of time-varying correlations and distributions, and it was demonstrated that the proposed method was superior to the DCC-based method for asymmetric and non-normal distributions. CONCLUSIONS: FMRI data of 138 human participants watching a Pixar animated movie were analyzed by the proposed method based on five a priori selected brain regions in the cortex. Based on statistical group analysis results, it was discovered that (1) the correlation between the left temporoparietal junction (LTPJ) and the primary visual cortex (V1) and the correlation between the dorsal posterior cingulate cortex (dPCC) and V1 were significantly higher for older age groups (5yo-Adult) more often than for younger age groups (3yo-4yo), and (2) the right temporoparietal junction (RTPJ), LTPJ, and dPCC were significantly correlated in all age groups at most of the scanning time periods.
BACKGROUND: Recent studies showed that functional connectivity (FC) in the human brain is not static but can dynamically change across time within time scales of seconds to minutes. NEW METHOD: This study introduces a new statistical method called the copula time-varying correlation for dynamic functional connectivity (dFC) analysis from functional magnetic resonance imaging (fMRI) data. RESULTS: Compared to other state-of-the-art statistical measures of dynamic correlation such as the dynamic conditional correlation (DCC), the proposed method can be effectively applied to data having asymmetric or non-normal distributions. COMPARISON WITH EXISTING METHODS: Numerical simulations were conducted under various kinds of time-varying correlations and distributions, and it was demonstrated that the proposed method was superior to the DCC-based method for asymmetric and non-normal distributions. CONCLUSIONS: FMRI data of 138 humanparticipants watching a Pixar animated movie were analyzed by the proposed method based on five a priori selected brain regions in the cortex. Based on statistical group analysis results, it was discovered that (1) the correlation between the left temporoparietal junction (LTPJ) and the primary visual cortex (V1) and the correlation between the dorsal posterior cingulate cortex (dPCC) and V1 were significantly higher for older age groups (5yo-Adult) more often than for younger age groups (3yo-4yo), and (2) the right temporoparietal junction (RTPJ), LTPJ, and dPCC were significantly correlated in all age groups at most of the scanning time periods.