Abstract:Coherence in the phase of oscillatory neuronal activity indicates functional interaction among brain regions at rest. Infra-slow fluctuations in BOLD signal (band 5 - 0.01 to 0.027 Hz - and band 4 - 0.027 to 0.073 Hz) has been observed using resting state fMRI when employing frequency-domain analysis, and has been previously shown to be altered in schizophrenia. In the current study, we examined the strength and the dynamic variance of phase coherence in these 2 bands (using a sliding window approach) among 6 large-scale brain networks (default-mode, fronto-parietal, salience, sensorimotor, visual, cerebellar) in 129 drug-naïve patients with schizophrenia and 197 healthy controls. Our motivation was to exploit the large-scale resting-state slow-wave oscillatory dynamics to parse the heterogeneity of schizophrenia. Four 6*6 matrices depicting patient vs. control differences in dynamic variance and mean of phase coherence (vPC and mPC respectively) in the slow 4 and slow 5 bands were constructed from resting state fMRI time series obtained from 6 networks based on 160 nodes of Dosenbach’s atlas. Deviations in mean/variance of phase coherence among the 6 networks were identified after FDR correction for each matrix (p<0.05). A latent profile analysis (LPA) was undertaken on the basis of the identified deviant features in the patient group. LPA is a Finite Mixture Modelling approach to identify naturally occurring sub-groups of patients on the basis of multivariable data. The identified subgroups were compared in terms of the severity of clinical symptoms across van der Gaag’s 5 factors of PANSS scale. Patients with schizophrenia showed increased vPC between salience-sensorimotor and visual-cerebellar networks in band 5; decreased vPC between DMN-sensorimotor and DMN-cerebellar networks in band 4. Patients also had a decrease in mPC between DMN-visual and DMN-cerebellar networks in band 5. We were able to identify 3 subgroups of patients using LPA. SZ1 (n=28) and SZ2 (n=45) had higher overall burden of symptoms compared to SZ3 (n=56). SZ2 had the highest burden of negative syndrome score and showed most deviance from healthy controls (5 out of 7 features significantly different from the healthy cohort). SZ1 had the highest burden of positive syndrome scores and had 4 out of 7 features deviant from HC. In contrast, SZ3 had least deviation from HC (3 out of 7 features) and also had less symptom burden across all symptom dimensions. Various abnormalities have been reported in the interactions among the large-scale networks in schizophrenia, with lack of consistency ascribed to syndromic heterogeneity. We illustrate how deviations in time-varying nature of slow-wave oscillations in resting state fMRI can be exploited to meaningfully reduce heterogeneity of this illness. The 3 subgroups thus identified not only show differential symptom burden but also exhibit hierarchical deviation from a normative group of healthy controls (SZ2>SZ1>SZ3>HC). To our knowledge this is the first attempt to stratify ‘neurotypes’ among drug-naïve patients with schizophrenia on the basis of large-scale network dynamics. Given the widespread availability of resting-fMRI data, we anticipate independent replication of our results in the near future.

Neuropsychiatric Disorder Subtyping Via Clustered Deep Learning Classifier Explanations

Redefining and subtyping of major depression based on brain functional connectivity signatures with high generalization: an ensemble hybrid framework

Identifying Subgroups of Major Depressive Disorder Using Brain Structural Covariance Networks and Mapping of Associated Clinical and Cognitive Variables

Explainable fuzzy clustering framework reveals divergent default mode network connectivity dynamics in schizophrenia

O3.1. NEUROTYPING UNTREATED FIRST EPISODE SCHIZOPHRENIA ON THE BASIS OF SLOW-WAVE RESTING-STATE DYNAMICS

Two neurostructural subtypes: results of machine learning on brain images from 4,291 individuals with schizophrenia

A Novel Explainable Fuzzy Clustering Approach for fMRI Dynamic Functional Network Connectivity Analysis

Multi-modal deep learning from imaging genomic data for schizophrenia classification

Identification of transdiagnostic psychiatric disorder subtypes using unsupervised learning

Computational framework for detection of subtypes of neuropsychiatric disorders based on DTI-derived anatomical connectivity

Sparse deep neural networks on imaging genetics for schizophrenia case–control classification

Two Distinct Neuroanatomical Subtypes of Schizophrenia Revealed Using Machine Learning.

Subtyping Schizophrenia Patients Based on Patterns of Structural Brain Alterations

Identification of psychiatric disorder subtypes from functional connectivity patterns in resting-state electroencephalography

Dissociation between neuroanatomical and symptomatic subtypes in schizophrenia

Neurostructural subgroup in 4291 individuals with schizophrenia identified using the subtype and stage inference algorithm

Subtyping Autism Spectrum Disorder Via Joint Modeling of Clinical and Connectomic Profiles.

Mapping Relationships among Schizophrenia, Bipolar and Schizoaffective Disorders: A Deep Classification and Clustering Framework Using Fmri Time Series

Towards a New Classification of Stable Phase Schizophrenia into Major and Simple Neuro-Cognitive Psychosis: Results of Unsupervised Machine Learning Analysis

Evaluation of boundaries between mood and psychosis disorder using dynamic functional network connectivity (dFNC) via deep learning classification

An Explainable and Robust Deep Learning Approach for Automated Electroencephalography-based Schizophrenia Diagnosis