T52. ABERRANT SENSORY PRECISION IN FIRST-EPISODE PSYCHOSIS: A 7-TESLA RESTING-STATE FMRI AND STROOP-TASK STUDY

Abstract

Background: Predictive coding and active inference framework states that subjects with psychotic disorders overly ascribe confidence (precision) to both sensory information and prior beliefs, leading to delusions and hallucinations. When considered in terms of brain connectivity, we predict that this computational aberrance should map onto strong forward and backward connections between sensory and higher cortical areas, in particular the anterior insula -deemed to be a critical node in the salience network. We tested this prediction in untreated, acute first-episode-psychosis (FEP) using both cognitive and neurophysiological data. Method(s): Twenty FEP and 20 healthy control (HC) subjects (matched on age and gender) performed the Stroop task and were subsequently scanned during task-free state with ultra-high-field (7 Tesla) functional magnetic resonance imaging. From a theory-driven modeling perspective, we fit hierarchical (Bayesian) drift-diffusion models (HDDM) to the distributions of reaction time and accuracy data and spectral dynamic causal models (DCM) to the fMRI data. Three competing HDDM tested the hypotheses that group differences in behavioral data would be accounted for aberrant prior precision (H1, mapped onto the starting point parameter), aberrant sensory precision (H2, mapped onto the drift-rate parameter), or both aberrant prior precision and aberrant prior beliefs (H3, mapped onto the starting point and drift-rate parameters). With DCM, we evaluated the differences in parameter estimates (beta) of effective connections (i.e., causal influence) between the right inferior occipital gyrus (rIOG) and the right anterior insula (rAI). Specifically, we performed Bayesian parameter averaging at a group level and a matched analysis (Bayesian estimations of differences in means) at a subject level to evaluate whether FEP subjects would show stronger forward rIOG -> rAI (indexing aberrant sensory precision) and stronger backward rAI -> rIOG connections (indexing aberrant prior beliefs) than HC subjects. Result(s): In the Stroop task, FEP performed less accurately (posterior proportion, PP, of the most credible values of difference in means, PP = .97 and more slowly, PP = 1.0) than HC. The HDDM representing prior and sensory precision (H3) accounted for this difference, FEP showed larger prior precision (PP = 0.93) and lower sensory precision (PP = 1.0) than HC. The spectral DCM showed that the influence from the sensory area to the insula (rIOG -> rAI) was negative (implying an attenuation) in both groups during resting state (betaHC = -0.29, betaFEP = -0.13; PP =1). However, FEP showed less sensory attenuation than HC (PP = 0.95). Furthermore, in both groups the insular cortex positively influenced the sensory area, indicating a facilitatory effect (rAI -> rIOG, betaHC = 0.3, betaFEP = 0.12; PP =1). However, this effect was weaker in FEP than in HC (PP = 0.98). Discussion(s): We have demonstrated that low sensory attenuation in subjects with psychosis is associated with strong forward connections from sensory to higher cortical areas. Contrary to our expectation, we observed a reduction, rather than a compensatory increase in backward connectivity. However, aberrant prior precision during the Stroop task suggests the hypothesis that strong backward connections could be context-dependent, elicited in highly uncertain situations (e.g., during the resolution of incongruent stimuli in task fMRI) -which is congruent with our previous findings. Mapping a neurocomputational construct (aberrant sensory precision) onto visual-insular effective connectivity provides empirical support to disrupted hierarchical information processing in early stages of psychosis.