Abstract
BACKGROUND: Schizophrenia is a severe psychiatric disorder marked by specific cognitive and clinical disturbances, for which neuroimaging biomarkers remain elusive. Novel theoretical and computational frameworks, such as integrated information decomposition, offer promising approaches to provide interpretable biomarkers for neuroimaging alterations in schizophrenia, potentially capturing disruptions relevant to consciousness and self-experience. METHODS: In this preliminary methodological exploration study, resting-state functional MRI (rsFMRI) data from 72 patients with schizophrenia and 74 healthy controls were retrieved and analyzed. Integrated information decomposition was leveraged to assess pairwise brain connectivity according to redundant, transferred, and synergistic components of information processing, as well as an overall metric of emergent consciousness/information integration: Φ. Clinical correlates with the Positive and Negative Syndrome Scale and the Wechsler Adult Intelligence Scale were assessed by partial Spearman correlations. Diagnostic accuracy was assessed through L1-regularized logistic regressions, after 5-fold cross-validation. RESULTS: Redundancy was positively correlated with intelligence quotient (IQ) across both groups (rho = 0.187, p-value = 0.033). Within patients, information metrics were positively correlated with stereotyped thinking (min rho = 0.343, max p-value = 0.006) and preoccupation (min rho = 0.250, max p-value = 0.046). Positive symptoms were positively correlated with redundancy (min rho = 0.250, max p-value = 0.047). Promising diagnostic accuracy was reached with Φ (balanced accuracy = 64.38%, area under the curve = 70.89%), redundancy (balanced accuracy = 84.93%, area under the curve = 92.30%), and synergy (balanced accuracy = 65.75%, area under the curve = 70.93%). CONCLUSIONS: These preliminary findings suggest that information metrics may offer clinically relevant, interpretable biomarkers for schizophrenia.