Abnormal brain network reconfiguration in neuropsychiatric disorders across cognitive decline, Depression, and Schizophrenia

Abstract

Objective Neuropsychiatric disorders are characterized by high complexity and comorbidity, imposing a substantial burden on both patients and society. However, their elusive pathogenic mechanisms impede accurate clinical diagnosis and effective interventions. To overcome this challenge, the present study proposes a novel framework to quantify and characterize these disorders. Methods Routine electroencephalogram (EEG) recordings are acquired from 236 subjects, including patients with Alzheimer’s disease (AD), mild cognitive impairment (MCI), major depressive disorder (MDD), schizophrenia, and healthy controls (HCs). Time-varying functional brain networks are constructed by phase locking value (PLV) analysis on band-pass filtered EEG signals. Subsequently, the nodal behavior characteristics within these dynamic brain networks are quantified by integrating robust dynamic community detection algorithms and network reconfiguration metrics. Results Significant intergroup differences in network reconfiguration metrics are identified based on the dynamic community structures (FDR-corrected p < 0.001 ). Lower cohesion strength is observed across all neuropsychiatric disorders compared to healthy controls, consistent across all frequency bands and recording sites. When six machine learning classifiers are trained on these metrics, the maximum classification accuracies exceeded 80%. Since lower cohesion strength is a prominent potential biomarker for neuropsychiatric disorders, it was then selected as the independent input feature for random forest classifier, and the classification accuracy achieved 0.85 for schizophrenia group, 0.88 for both the MCI and MDD group, and 0.82 for the AD group. Conclusions Our findings indicate that the framework based on dynamic network reconfiguration metrics effectively captures both the shared and disorder-specific alterations in brain network dynamics among neuropsychiatric disorders. Significance Dynamic community structure advances our understanding of the pathological mechanisms underlying neuropsychiatric disorders. This study provides novel insights that may inform the development of more targeted and effective therapeutic strategies.