My collaborator Genevieve Yang is presenting a poster at the 2015 Society for Neuroscience meeting, on Sunday, October 18, in the morning session. Our poster is titled: Cortical hierarchy underlies preferential connectivity disturbances in schizophrenia. Our abstract is below:
Schizophrenia (SCZ) is a neuropsychiatric illness associated with abnormal neural connectivity. In particular, patients show prefrontal cortex (PFC) hypo-connectivity, assessed by computing blood-oxygen level-dependent (BOLD) signal correlations. However, recent studies reveal elevated BOLD signal variance in SCZ, which may impact correlations, computed as covariance normalized by variance. We hypothesized that functional connectivity (using covariance) may be elevated in SCZ, but that this may occur in the context of elevated signal variance. Further, we hypothesized that preferential PFC effects may intrinsically arise from the information-processing hierarchy and corresponding physiological consequences, which we tested via biophysically grounded computational modeling. We conducted resting-state fMRI in 161 SCZ and 164 matched healthy subjects, assessing group differences in connectivity and BOLD variance. Both voxel-wise and network-level analyses were performed. To mechanistically inform fMRI findings, we used a large scale neural network model to simulate a well-known synaptic hypothesis of SCZ pathology—namely excitation/inhibition (E/I) imbalance, and analyzed resulting in silico ‘BOLD’ signals. Empirically, we observed hyper-connectivity in PFC and other associative regions in SCZ, with concurrent increases in BOLD variance. These effects were absent in a comparison group of bipolar patients (N=73). In initial simulations of E/I imbalance, we observed global elevations in covariance and variance of model-generated BOLD signals. To investigate our empirical associative effects, we extended our model to reflect known differences in associative vs. sensory neuronal dynamics. This extended model reproduced preferential associative effects, and predicted that covariance and variance elevations would be positively correlated, which we confirmed empirically. Collectively, we show that elevations in BOLD covariance and variance in chronic SCZ co-occur and are strongly related phenomena. Thus, some connectivity elevations may not be fully captured by correlation measures that normalize connectivity by variance. Hypo-connectivity seen in previous studies may be reconciled with our findings by considering the effect of elevated variance in reducing correlation-based measures. We also computationally demonstrate how a common cellular-level mechanism can produce elevations in covariance and variance of BOLD signals. Further, we show how this global perturbation can produce preferential effects in associative regions due to neural differences arising from the intrinsic functional cortical hierarchy.