Basic neural processing mechanisms of live human face viewing

Yale University, New Haven CT

Investigators

Abstract

Project Summary INTRODUCTION. Face perception is central to social functioning and commonly disrupted in psychosis. Prior research found that face processing in the dorsal and lateral visual streams is modulated by visual behavior, pointing to the existence of a distributed face visual processing neural network which is an extension of visual acquisition and is a candidate for dysfunction in mental illness. Building upon the Bayesian Brain hypothesis which posits that the brain is an inference machine which models the environment to make predictions and drive behavior, we theorized that a lateral-dorsal network emerges during dynamic face viewing as a function of fixation behavior, enabling emotional expression recognition. Further, this network is a candidate driver of the emotion recognition deficits common in psychosis. AIM 1 is thus to test hypotheses that a fully connected lateral-dorsal network emerges during dynamic face viewing as a function of fixation behavior, using dynamic causal modelling (DCM). AIM 2 is to test hypotheses that this network is related to expression perception and social functioning, with disruptions emerging in those with psychosis and attenuated in those who are clinical high risk for psychosis(CHR). HYPOTHESES. A fully connected lateral-dorsal network is hypothesized to be the model of best fit for face viewing electrocortical data. Mean fixation duration is posited to modulate the intra-lateral and lateral to dorsal connections as mean fixation is theorized to determine precision of bottom-up stimuli. Significant differences in behavior and parameter estimates are hypothesized to occur between those with and without psychosis, with the CHR group as an intermediate phenotype. A mixed effects model including parameter estimates for lateral edges and mean fixation duration is hypothesized to predict expression accuracy and social functioning across pooled subjects. APPROACH. Participants (n=30 each of those with and without psychosis, and those who are CHR) will view dynamic RADIATE faces11 and control videos normalized for brightness and movement, reporting perceived emotions per event. 128-channel scalp EEG and 1000hz binocular eye tracking will be acquired, supplemented by the Global Assessment of Functioning (GAF)12. Data will be simulated using biophysically realistic head and cortex DCMs13,14 with various connectivity patterns between lateral and dorsal streams, and compared to the recorded data. The best fit model will be identified with Bayesian model selection and single subject parameters will be estimated with Variational Bayes15. Average parameters and behavioral measures per group will be compared to identify significant differences. Mixed effects models will be generated to relate parameter estimates and mean fixation duration to emotion perception accuracy and GAF score. SIGNIFICANCE. These hypotheses may shed light on a core deficit in psychosis—social functioning—through identification of a mechanism of action—atypical acquisition of dynamic facial information leading to disrupted connectivity in a lateral-dorsal neural network and disrupting emotional expression perception. If true, it will enable new avenues for targeted treatment and symptom amelioration.