Evaluation and calibration of functional network modeling methods based on known anatomical connections

Abstract

1-Abstract EnglishRecent studies have identified large scale brain networks based on the spatio-temporal structure of spontaneous fluctuations in resting-state fMRI data. It is expected that functional connectivity based on resting-state data is reflective of-but not identical to the underlying anatomical connectivity.However, which functional connectivity analysis methods reliably predict the network structure remains unclear. Here we tested and compared network connectivity analysis methods by applying them to fMRI restingstate time-series obtained from the human visual cortex. The methods evaluated here are those previously tested against simulated data in Smith et al. (Neuroimage, 2011).To this end, we defined regions within retinotopic visual areas V1, V2, and V3 according to their eccentricity in the visual field, delineating central, intermediate, and peripheral eccentricity regions of interest (ROIs). These ROIs served as nodes in the models we study. We based our evaluation -anatomical connectivity in the monkey visual cortex. For each evaluated method, we computed the fractional rate of detecting connections known to -th percentile of the 4 distribution of interaction magnitudes of those connections not expected to exist.Under optimal conditions, including session duration of 68 minutes, a relatively small network consisting of 9 nodes and artifact-free regression of the global effect, each of the top methods predicted the expected connections with 75%-83% c-sensitivity. Partial Correlation performed best (PCorr; 83%), followed by Regularized Inverse Covariance (ICOV; 79%), Bayesian Network methods (BayesNet; 77%), Correlation (75%), andGeneral Synchronization measures (75%). With decreased session duration, these top methods saw decreases in c-sensitivities, achieving 66%-78% and 60%-70% for 34 and 17 minute sessions, respectively. With a short resting-state fMRI scan of 8.5 minutes (TR = 2s), none of the methods predicted the real network well, with ICOV (53%) and PCorr (51%) performing best. With increased complexity of the network from 9 to 36 nodes, multivariate methods including PCorr and BayesNet saw a decrease in performance. However, this decrease became small when using data from a long (68 minutes) session. Artifact-free regression of the global effect significantly increased the c-sensitivity of all top-performing methods. In an overall evaluation across all tests we performed, PCorr, ICOV and BayesNet set themselves somewhat above all other methods. 5We propose that data-based calibration based on known anatomical connections be integrated into future network studies, in order to maximize sensitivity and reduce false positives. Studies have also demonstrated changes in fMRI-measured brain activity when the subjects are in a state of rest. Importantly, it has been shown that these spontaneous (i.e. without any external stimulation)fluctuations in fMRI signals demonstrate functional connectivity over large parts of the human brain (Xiong et al., 1999). These large parts of the brain correspo...

Ongoing Slow Fluctuations in V1 Impact on Visual Perception

“…This result is in line with previous findings about the intrinsic organization of the occipital cortex (Cohen et al, 2008; Kiviniemi et al, 2009; Allen et al, 2011; Craddock et al, 2012; Dawson et al, 2013). …”

Initial Validation for the Estimation of Resting-State fMRI Effective Connectivity by a Generalization of the Correlation Approach

“…P-correlation and four alternative methods from Smith et al (2011), specifically, “Granger B1,” “Gen Synch S1,” “LiNGAM,” and “Patel's conditional dependence measure,” were compared by the accuracy criteria (${A}$), since under both synthetic and experimental scenarios, these methods have been tested and have relatively good performances among all the others (Smith et al, 2011; Dawson et al, 2013). The computation of these methods were done by software provided by Prof. S.M.…”

“…Smith. Granger B1, a pairwise Granger causality estimation method which provides the best performance among Granger causality approaches (Smith et al, 2011; Dawson et al, 2013), uses the Bayesian Information Criterion to estimate the lag up to 1 TR. Gen Synch S1 is a nonlinear synchronization method with respect to the time lag 1 TR.…”

“…Prior work has shown its high sensitivity for detecting the existence of network architectures under both simulated and experimental scenarios (Smith et al, 2011; Dawson et al, 2013). This paper describes methodology for analyzing rs-fMRI data using a generalization of well-established correlation ideas.…”

Subject-specific functional parcellation via Prior Based Eigenanatomy

“…These issues are particularly salient in blood oxygen level dependent (BOLD) fMRI where recent papers have highlighted the sensitivity of this modality to specific analysis choices (Carp, 2012; Haller and Bartsch, 2009). Network analysis of functional connectivity within the brain from BOLD data has received a significant amount of attention and is notorious for being sensitive to analysis decisions (Dawson et al, 2012; Eke et al, 2012). …”