Accurate classification of pain experiences using wearable electroencephalography in adolescents with and without chronic musculoskeletal pain

Abstract

ObjectiveWe assessed the potential of using EEG to detect cold thermal pain in adolescents with and without chronic musculoskeletal pain.MethodsThirty-nine healthy controls (15.2 ± 2.1 years, 18 females) and 121 chronic pain participants (15.0 ± 2.0 years, 100 females, 85 experiencing pain ≥12-months) had 19-channel EEG recorded at rest and throughout a cold-pressor task (CPT). Permutation entropy, directed phase lag index, peak frequency, and binary graph theory features were calculated across 10-second EEG epochs (Healthy: 292 baseline / 273 CPT epochs; Pain: 1039 baseline / 755 CPT epochs). Support vector machine (SVM) and logistic regression models were trained to classify between baseline and CPT conditions separately for control and pain participants.ResultsSVM models significantly distinguished between baseline and CPT conditions in chronic pain (75.2% accuracy, 95% CI: 71.4%–77.1%; p < 0.0001) and control (74.8% accuracy, 95% CI: 66.3%–77.6%; p < 0.0001) participants. Logistic regression models performed similar to the SVM (Pain: 75.8% accuracy, 95% CI: 69.5%–76.6%, p < 0.0001; Controls: 72.0% accuracy, 95% CI: 64.5%–78.5%, p < 0.0001). Permutation entropy features in the theta frequency band were the largest contributor to model accuracy for both groups.ConclusionsOur results demonstrate that subjective pain experiences can accurately be detected from electrophysiological data, and represent the first step towards the development of a point-of-care system to detect pain in the absence of self-report.

Behavioral modeling and neuroimaging of impaired risky decision making in patients with chronic musculoskeletal pain

“…The distinctive risky decision-making behaviors and deactivated PFC activity during risky decision making of chronic pain patients provide an effective index of automatically discerning the chronic pain patients from HCs and measuring the intervention efficacy. Several machine learning methods have achieved satisfactory performance in the diagnosis of chronic pain patients by combining cognitive tasks and neuroimage tools 86 , 87 . Further studies could utilize artificial intelligence, such as machine learning and deep learning, to combine cognitive and neuroimaging data for precise diagnosis and treatment of chronic pain patients.…”

“…Several machine learning methods have achieved satisfactory performance in the diagnosis of chronic pain patients by combining cognitive tasks and neuroimage tools. 86 , 87 Further studies could utilize artificial intelligence, such as machine learning and deep learning, to combine cognitive and neuroimaging data for precise diagnosis and treatment of chronic pain patients. However, there were several limitations in our study which warrant discussion.…”

PainVision-based evaluation of brain potentials: a novel approach for quantitative pain assessment

“…In this study, QST was conducted with a commercial device (PV, PS-2100, Nipro Co., Osaka, Japan) by delivering percutaneous electrical stimuli at given intensity. In addition to electoral stimulation, several attempts have made to combine ERPs with other phenotype of sensory input (i.e., mechanical, cold, and Frontiers in Bioengineering and Biotechnology frontiersin.org thermal stimuli) in previous reports (Teel et al, 2022;Anders et al, 2023). We determined to apply PainVision system to achieve a replicable and stable behavioral result by controlling the extract output and duration of peripheral stimuli (Table 1), which is especially critical for data reproducibility in the EEG research.…”

In search of a composite biomarker for chronic pain by way of EEG and machine learning: where do we currently stand?

“…Moreover, the decoder performed only at 57% accuracy—close to chance-level, leaving much room for improvement. In more recent years, researchers’ primary goal in improving decoding performance has been motivated primarily by the goal of optimizing model generalization, where the application of SVM classifiers has led to an improvement in accuracy with up to 93.7% ( Misra et al, 2017 ; Kragel et al, 2018 ; Levitt et al, 2020 ; Buchanan et al, 2021 ; Lendaro et al, 2021 ; Zolezzi et al, 2021 ; Teel et al, 2022 ; Topaz et al, 2022 ).…”