655: An open-access multimodal physiological dataset for pain biomarker research

Topic: Experimental (Non-clinical) Human Studies
Subtopic: Non -specific to any diagnosis, Electrophysiology/Clinical neurophysiology (e.g. microneurography, EEG, MEG, OPMs, TMS), Assessment and Measurement of Pain/Nociception, Artificial Intelligence/Machine Learning

Background and Aims: Pain is a complex, subjective experience influenced by contextual, affective, and cognitive factors1,2. Recently, machine learning has been increasingly applied to EEG data to identify pain biomarkers3,4. However, several methodological limitations have hindered their development, such as small sample sizes and the lack of external validation across independent datasets. While large datasets are increasingly made publicly available, most focus on a single pain modality and do not offer adequate non-painful aversive control conditions. We therefore present a dataset combining multimodal physiological signals with diverse painful and aversive control conditions.

Methods: EEG (n = 100) and physiological (galvanic skin response, electrocardiogram; n = 51) data were collected from healthy adults during tonic pain (heat, chemical) and aversive non-painful (sound, orofacial disgust) tasks. Participants provided continuous pain or unpleasantness ratings. Open- and closed-eyes resting-state recordings, and sociodemographic and psychological (depression, anxiety, pain catastrophization) data are available. EEG data was standardized to BIDS format and preprocessed via a pipeline based on DISCOVER-EEG5. Technical validation included assessing EEG signal integrity and autonomic responses plausibility. Cluster-based permutation tests and spectral analyses assessed the sensitivity of EEG signals to subjective ratings and stimulus intensity across all frequency bands.

Results: Technical validation confirmed high-quality EEG data (98.0% ± 3.0% good channels; 98.1% ± 2.1% epoch retention). Significant spectral modulations distinguishing different intensities of perception of painful or unpleasant states across tasks within theta, alpha, beta, and gamma bands were identified. Significant intensity-dependent effects were also found in the auditory and thermal tasks in all frequency bands. Autonomic metrics were biologically plausible, showing expected stimulus-evoked peaks followed by a gradual return to baseline. Reporting unpleasantness was associated with significantly higher skin conductance level (SCL) for the auditory task. SCL was significantly modulated by stimulus intensity in the thermal pain task.

Conclusions: This dataset provides a robust resource for developing and validating pain biomarkers, as well as for original studies aiming to better understand the electrophysiology of pain. By including non-painful aversive control conditions, it allows researchers to distinguish pain-specific patterns from general salience-driven responses. The multimodal nature of the data also facilitates the discovery of composite biomarkers which could better capture the multidimensional complexity of pain6. By offering a large-scale and comprehensive database, this initiative supports global collaborative efforts toward more generalizable and reproducible pain research.

References: 1. Melzack, R., & Casey, K.L., In: The Skin Senses. (1968).
2. Fields, H. L., Pain. (1987)
3. Davis, K. D., et al., Nat. Rev. Neurol. (2020)
4. Mari, T., et al., J Pain. (2022)
5. Gil Ávila, C., et al., Sci Data. (2023)
6. Zhang, L.-B., et al., Cell Rep Med. (2024)

Presenting Author: Alyson Champagne
Submitting Author: Alyson Champagne

Senior Author: Michel-Pierre Coll

IASP 2026 Global Year theme of Neuropathic Pain: No

• Lorie-Ève Barrette: NA - Université Laval
• Mathieu Roy: Ph D - McGill University
• Michel-Pierre Coll: Ph D - Université Laval