Advances in EEG-Based Neural Mechanism Analysis and Decoding Methods for Pain

Abstract

Objective pain assessment remains challenging, particularly in patients with limited capacity for self-report. Electroencephalography (EEG), with its high temporal resolution, non-invasiveness, and clinical accessibility, has become an important tool for investigating pain-related neural dynamics and developing objective pain decoding models. This review summarizes recent advances in EEG-based pain decoding, focusing on neural oscillatory mechanisms, signal processing strategies, and machine learning (ML) and deep learning (DL) approaches. Current evidence indicates that acute pain is mainly associated with stimulus-evoked oscillatory responses, especially gamma-band activity related to pain intensity, whereas chronic pain is characterized by altered resting-state rhythms, abnormal low-frequency activity, and reorganization of large-scale brain networks. Traditional ML methods offer interpretability and are suitable for small-sample studies but rely heavily on handcrafted features. In contrast, DL models can automatically learn complex spatiotemporal representations and generally achieve higher decoding performance, although their generalizability and interpretability remain limited. Key challenges include signal artifacts, inter-individual variability, inconsistent experimental paradigms, and a lack of standardized public datasets. Future studies should emphasize multimodal fusion, transfer learning, explainable artificial intelligence, standardized validation, and clinical translation toward real-time and personalized pain monitoring.