Disease-specific network pattern of perinatal depression revealed by Common Orthogonal Basis Extraction
Yueheng Peng, Jihan Wang, Xianyong Fan, Yue Yu, Guolin He, Dawazhuoma, Lu Jiang, Tingting Zhang, Silangquncuo, Chanlin Yi, Dezhong Yao, Bin Lv, Peng Xu, Kaibo Shi
Abstract
The clinical diagnosis of perinatal depression (PD) presents considerable challenge, as it is much harder to identify than non-perinatal depression. Psychologically, common emotional fluctuations during pregnancy are easily confounded with depressive symptoms, leading to missed and incorrect diagnoses. Neurologically, pregnancy-induced alterations in brain activity could obscure neuroimaging features specific to PD. Therefore, this study introduced an innovative approach that combined brain network analysis with Common Orthogonal Basis Extraction (COBE) to identify a PD-Specific Network Pattern from resting-state brain networks, as well as validating its efficacy in diagnosis and assessment. Resting-state electroencephalography (EEG) data were collected from 21 patients with PD and 20 healthy pregnant (HP) individuals, from which functional brain networks were constructed. An optimized COBE method was then employed to extract Exclusive Network Pattern for each group, as well as Common Network Pattern shared by all participants (PD + HP). This process enabled the identification of the PD-Specific Network Pattern that most consistent with neural mechanisms of PD. Based on the PD-Specific Network Pattern, PD-Specific Features were derived and applied to train support vector machine and multiple linear regression models, which respectively performed individual-level classification and assessment. This study effectively addressed the limitation of traditional neuroimaging techniques in the diagnosis of PD, providing a new avenue for objective screening and dynamic monitoring of perinatal depression. • A modified Common Orthogonal Basis Extraction (COBE) method was proposed to filter out pregnancy-induced neural interference. • EEG-based Disease-Specific Network Pattern was identified to capture the core neural pathology of perinatal depression. • Disease-Specific Features enabled quantitative assessment of perinatal depression, surpassing subjective scales. • This study provided a novel, objective, and efficient tool for the clinical screening and monitoring of perinatal depression.