195: Toward Objective Chronic Pain Assessment: Decoding Pain Intensity from EEG Microstate Dynamics

Topic: Clinical Science
Subtopic: Non-specific to any diagnosis, Electrophysiology/Clinical neurophysiology (e.g. microneurography, EEG, MEG, OPMs, TMS)

Background and Aims: Chronic pain is a disorder characterized by aberrant temporal dynamics of large-scale brain networks. However, traditional analysis methods often overlook the rapid temporal evolution of these networks. EEG microstate analysis offers a solution by capturing sub-second brain dynamics that reflect large-scale brain network organization. This study aimed to characterize microstate dynamics across pain intensity levels and evaluate their predictive utility for intensity classification.

Methods: Resting-state EEG data were recorded from 100 chronic pain patients, who were stratified according to Visual Analog Scale (VAS) scores (mild: n=18; moderate: n=56; severe: n=26). The microstate classes were identified using modified k-means clustering. Temporal parameters (mean duration, occurrence, time coverage, and transition probability) were extracted. Group differences and clinical correlations were assessed using linear mixed-effects (LME) models and multiple regression, controlling for age and sex. Additionally, a ResNet-based classifier was developed to distinguish intensity levels using these microstate-derived features.

Results: Group comparisons revealed significant pain intensity–dependent alterations in microstate dynamics. Regression analyses further demonstrated associations between microstate temporal parameters and pain intensity. Higher VAS was associated with increased microstate C duration (β=4.01, p=0.017), occurrence (β=0.40, p=0.046) and coverage (β=2.95, p=0.028). Conversely, VAS was negatively associated with microstate D duration (β=−4.69, p=0.019) and coverage (β=−3.64, p=0.046), with no effect on occurrence (p=0.124). Using microstate parameters as features, the ResNet model achieved high performance in pain intensity classification (accuracy 81.82%, sensitivity 83.70%, specificity 81.46%, F1 81.90%).

Conclusions: Our findings reveal that chronic pain intensity is encoded in the temporal reorganization of large-scale brain networks. Increased microstate C and reduced microstate D temporal engagement suggest a shift toward salience-network dominance with diminished dorsal attention network involvement. The microstate parameters hold promise as objective biomarkers for clinical pain assessment.

References: [1] Brain dysfunction in chronic pain patients assessed by resting-state electroencephalography: Erratum. Pain 2020;161:1684–1684. doi:10.1097/j.pain.0000000000001948.
[2] Jajcay N, Hlinka J. Towards a dynamical understanding of microstate analysis of M/EEG data. NeuroImage 2023;281:120371. doi:10.1016/j.neuroimage.2023.120371.
[3] Li J, Li N, Shao X, Chen J, Hao Y, Li X, Hu B. Altered Brain Dynamics and Their Ability for Major Depression Detection Using EEG Microstates Analysis. IEEE Trans Affective Comput 2023;14:2116–2126. doi:10.1109/TAFFC.2021.3139104.
[4] Tait L, Zhang J. MEG cortical microstates: Spatiotemporal characteristics, dynamic functional connectivity and stimulus-evoked responses. NeuroImage 2022;251:119006. doi:10.1016/j.neuroimage.2022.119006.
[5] Zhou Y, Gong L, Yang Y, Tan L, Ruan L, Chen X, Luo H, Ruan J. Spatio-temporal dynamics of resting-state brain networks are associated with migraine disability. J Headache Pain 2023;24:13. doi:10.1186/s10194-023-01551-y.

Presenting Author: Chao Wang
Submitting Author: chao wang

Senior Author: Guihua Tian

IASP 2026 Global Year theme of Neuropathic Pain: No

• Guihua Tian - Beijing Friendship Hospital, Capital Medical University