Granger Causality Estimation of Brain Connectivity in Autism Spectrum Disorders
Disruption in interregional functional and anatomical connectivity has been at the center of neurobiological accounts of autism spectrum disorder (ASD) (Just et al., 2012; Maximo, Cadena, & Kana, 2014). While insights from these models are valuable, functional connectivity does not provide the time-lagged causality and directionality of connectivity. Effective connectivity, however, provides information about the influence one system exerts over another in a given experimental context (Büchel and Friston, 2000). Only a few previous fMRI studies have utilized effective connectivity to understand the neural mechanisms in ASD (Deshpande et al., 2013; Shen et al., 2012; Shih et al., 2010; Wicker et al., 2008).
The main objective of this fMRI study was to examine effective connectivity differences underlying self-other processing in individuals with ASD using Granger causality method.
Eighteen high-functioning adolescents and adults with ASD and 18 age-and-IQ-matched typically developing control (TD) adults participated in this study. Participants made “yes” or “no” judgments about whether an adjective, presented visually, described them (self) or their favorite teacher (other). The data were collected using a 3T MRI scanner. Mean time series was extracted from 5 different regions of interest (ROIs) for all participants: left inferior parietal lobule (LIPL), left medial prefrontal cortex (LMPFC), supplementary motor area (SMA), and the pars opercularis (LIFO) and pars triangularis (LIFT) aspects of the left inferior frontal gyrus (LIFG). The extracted time series were normalized and the hemodynamic response de-convolved using a cubature Kalman filter (Havliceck et al., 2011) to get the underlying neuronal responses, which were put into the multivariate autoregressive model (Deshpande et al., 2010). Connectivity matrices were obtained and FDR corrected t-tests were performed to determine group differences.
For self condition, ASD participants showed weaker effective connectivity than TD from LIFT to LIPL and to LMPFC, from LIFO to LIPL and LMPFC, and from LIPL to LIFT (all p < 0.05 corrected). For other condition, ASD participants showed weaker effective connectivity from LIFO to LIPL and to MPFC. Effective connectivity during self condition was stronger for ASD than TD from LIPL to LIFO and to MPFC, and from MPFC to all other regions; LIFT; LIPL (all p < 0.05 corrected) . For other condition, effective connectivity in ASD was stronger for LIPL to LIFO, and bidirectionally between LIPL and LMPFC.
Overall, this study found strong causal information transfer among three brain areas, LIFG, MPFC, and IPL, with most of the weaker connections of ASD originating from LIFG to IPL or MPFC. The MPFC has been associated with theory-of-mind and thoughts about others (Ebner et al., 2011), and the LIFG and LIPL have been found to have a role in self-other processing (Decety and Somerville, 2003; Kelley et al., 2002). Overall, the findings of this study underscore altered patterns of information flow in participants with ASD during social cognition (Deshpande et al., 2013) and supplement functional connectivity findings in ASD.