Classifying Schizophrenia Subtypes via Resting- State EEG Complexity Networks

Schizophrenia (SZ), a network disorder, features abnormal functional connectivity, but fMRI limitations hinder clinical use. While EEG offers convenience, traditional complexity measures like sample entropy (SampEn) inadequately capture spatiotemporal network dynamics and yield inconsistent SZ findings. We investigated functional network alterations in SZ subtypes—deficit (DS) and non-deficit (NDS)—using a novel EEG complexity network approach and aimed for classification against healthy controls (HC). We analyzed resting-state EEG (64 channels, 500Hz) from 19 DS patients, 19 NDS patients, and 30 HC. Data underwent preprocessing, bandpass filtering (δ:0.5-4Hz, θ:4-8Hz, α:8-13Hz, β:13-40Hz), and artifact removal (ICA). Dynamic SampEn time series were calculated per electrode. Complexity networks were constructed using Spearman correlations; topological features (global efficiency, local efficiency, strength) were analyzed. Support Vector Machine (SVM) was used for classification. Traditional SampEn differentiated groups only in δ band. The network approach revealed significant topological alterations: DS exhibited the highest local efficiency (δ, θ, α) and lowest global efficiency (δ, α), while NDS showed the lowest global efficiency (θ) and highest local efficiency (β). SVM achieved 96.3% classification accuracy, optimal in δ/θ bands. This novel EEG complexity network method effectively distinguishes HC, DS, and NDS, demonstrating strong potential for clinical application, particularly in outpatient settings. Validation with larger cohorts and task-state EEG is warranted.