Using the Excitation/Inhibition Ratio to Optimize the Classification of Autism and Schizophrenia

Abstract

The excitation/inhibition (E/I) ratio has been shown to be elevated in both autism spectrum disorder (ASD) and schizophrenia (SZ), relative to neurotypical controls. However, the degree of E/I imbalance overlap and differentiation between SZ and ASD is not known. Our main objectives were therefore (1) to quantify group differences in the E/I ratio between controls, ASD and SZ, and (2) to assess the potential of the E/I ratio for differential diagnosis. We collected resting state fMRI (rsfMRI) and phenotypic data from 55 controls, 30 ASD, and 39 SZ, ages 18 to 35 (IQ>80). For each participant, we computed the Hurst exponent (H), an indicator of the E/I ratio, for the timecourses of 53 independent components covering the entire brain. Next, using Optimal Classification Trees (OCT), we ran a classification analysis on the two clinical groups using five incremental feature sets (i.e., models): (1) Positive and Negative Syndrome Scale (PANSS) and the Autism Diagnostic Observation Schedule (ADOS) only; (2) PANSS, ADOS, Bermond–Vorst Alexithymia Questionnaire (BVAQ), Empathy Quotient (EQ), and IQ; (3) H only; (4) H, PANSS and ADOS; (5) H, PANSS, ADOS, BVAQ, EQ and IQ. We observed decreased H (i.e., increase in E/I ratio) in ASD and SZ compared to controls, and in SZ compared to ASD in the Cerebellar, Sensorimotor, Visual and Cognitive Control networks. The OCT classification showed a consistent increase in discrimination accuracy across models between ASD and SZ, suggesting that the E/I ratio in combination with phenotypic measures can contribute to differential diagnosis in adults.

Electrophysiologically-defined excitation-inhibition autism neurosubtypes

“…Thus, the subtypes are different not only in the directionality of their difference relative to TD, but also in the topography of effects, potentially indicative of differential impact of separate brain regions. It is unclear how such results may hold up against other past work looking at H or related measures in autism in EEG 34,35 and rsfMRI data 11,[36][37][38] . Such past work does not examine data-driven E:I subtypes and instead rely on the clinical diagnosis to conduct case-control modeling.…”