Background: Deficits in neurocognition and social cognition are drivers of reduced functioning in schizophrenia spectrum disorders, with potentially shared neurobiological underpinnings. Many studies have sought to identify brain-based biomarkers of these clinical variables using a priori dichotomies (e.g., good vs. poor cognition, deficit vs. nondeficit syndrome).
Methods: We evaluated a fully data-driven approach to do the same by building and validating a brain connectivity-based biomarker of social cognitive and neurocognitive performance in a sample using resting-state and task-based functional magnetic resonance imaging (n = 74 healthy control participants, n = 114 persons with schizophrenia spectrum disorder, 188 total). We used canonical correlation analysis followed by clustering to identify a functional connectivity signature of normal and poor social cognitive and neurocognitive performance.
Results: Persons with poor social cognitive and neurocognitive performance were differentiated from those with normal performance by greater resting-state connectivity in the mirror neuron and mentalizing systems. We validated our findings by showing that poor performers also scored lower on functional outcome measures not included in the original analysis and by demonstrating neuroanatomical differences between the normal and poorly performing groups. We used a support vector machine classifier to demonstrate that functional connectivity alone is enough to distinguish normal and poorly performing participants, and we replicated our findings in an independent sample (n = 75).
Conclusions: A brief functional magnetic resonance imaging scan may ultimately be useful in future studies aimed at characterizing long-term illness trajectories and treatments that target specific brain circuitry in those with impaired cognition and function.
Keywords: Biomarker; Functional outcomes; Imaging; Machine learning; Resting-state fMRI; Schizophrenia.
Copyright © 2018 Society of Biological Psychiatry. All rights reserved.