Multi-Frequency EEG Connectomics Uncovers Insula-Network Subtypes in Somatic Symptom Disorder
Shuzhi Zhao1, Chongyuan Lian2, Xue Shi2
1CAS Key Laboratory of Human-Machine Intelligence-Synergy Systems, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China; Guangdong-Hong Kong-Macao Joint Laboratory of Human-Machine Intelligence-Synergy Systems, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, Shenzhen, China.
Biological Psychiatry
|July 29, 2026
Summary
Researchers identified three distinct subtypes of Somatic Symptom Disorder (SSD) using electroencephalography (EEG) connectomics and contrastive learning, paving the way for personalized treatments.
Area of Science:
- Neuroscience
- Computational Psychiatry
- Medical Informatics
Background:
- Somatic Symptom Disorder (SSD) presents significant clinical heterogeneity, hindering treatment effectiveness, with over 40% of patients not responding to standard interventions.
- Existing diagnostic criteria may not fully capture the neurobiological underpinnings of SSD, necessitating novel approaches for patient stratification.
Purpose of the Study:
- To develop and validate a framework integrating multi-frequency electroencephalography (EEG) connectomics with contrastive learning to identify distinct neurobiological subtypes of SSD.
- To correlate identified subtypes with clinical symptom dimensions and validate their reproducibility in an independent cohort.
Main Methods:
- A contrastive variational autoencoder with Gaussian mixture modeling (CVAE-GM) was employed using resting-state EEG connectomics data from 1,419 SSD patients.
- Subtypes were identified and correlated with symptom dimensions using cross-validated canonical correlation analysis.
- Reproducibility was assessed in an independent external cohort of 530 patients.
Main Results:
- Three robust subtypes were identified, characterized by dominant connectivity in somatomotor (SMN), central executive (CEN), and limbic (LN) networks.
- Distinct associations were found between subtypes and clinical dimensions: SMN-subtype with somatic burden, CEN-subtype with negative event reactivity, and LN-subtype with emotional symptoms.
- The insula served as a convergent hub, with subtype differentiation linked to its coupling with specific brain regions; external validation achieved 85% accuracy and 0.87 AUC.
Conclusions:
- An EEG-based connectomic framework can effectively investigate neurobiological heterogeneity in SSD.
- Insula-centered network features are promising for future mechanistic stratification of SSD patients.
- This approach offers potential for developing more targeted and effective interventions for SSD.
