Analytic consistency and neural correlates of peak alpha frequency in the study of pain

https://doi.org/10.1016/j.jneumeth.2021.109460Get rights and content

Highlights

  • Electroencephalography (EEG) processing pipelines vary widely in pain research.

  • Pain populations are heterogeneous and so are relationships with neural markers.

  • Potential EEG pain marker is robust against many common pre-processing differences.

  • EEG pain marker sensitive to calculation method and EEG region-of-interest.

  • EEG pain marker correlated with underlying activity in pain-related brain regions.

Abstract

Background

Several studies have found evidence of reduced resting-state peak alpha frequency (PAF) in populations with pain. However, the stability of PAF from different analytic pipelines used to study pain has not been determined and underlying neural correlates of PAF have not been validated in humans.

New method

For the first time we compare analytic pipelines and the relationship of PAF to activity in the whole brain and thalamus, a hypothesized generator of PAF. We collected resting-state functional magnetic resonance imaging (rs-fMRI) data and subsequently 64 channel resting-state electroencephalographic (EEG) from 47 healthy men, controls from an ongoing study of chronic prostatitis (a pain condition affecting men). We identified important variations in EEG processing for PAF from a review of 17 papers investigating the relationship between pain and PAF. We tested three progressively complex pre-processing pipelines and varied four postprocessing variables (epoch length, alpha band, calculation method, and region-of-interest [ROI]) that were inconsistent across the literature.

Results

We found a single principal component, well-represented by the average PAF across all electrodes (grand-average PAF), explained > 95% of the variance across participants. We also found the grand-average PAF was highly correlated among the pre-processing pipelines and primarily impacted by calculation method and ROI. Across methods, interindividual differences in PAF were correlated with rs-fMRI-estimated activity in the thalamus, insula, cingulate, and sensory cortices.

Conclusions

These results suggest PAF is a relatively stable marker with respect to common pre and post-processing methods used in pain research and reflects interindividual differences in thalamic and salience network function.

Introduction

Chronic pain is very impactful. A challenge of studying chronic pain is that it is an inherently subjective experience. Therefore, identifying objective markers is a pressing need. Markers of particular interest would be those that could measure the predisposition to developing chronic pain. Resting-state peak alpha frequency (PAF) measured from electroencephalography (EEG) has been proposed as such a marker (Furman et al., 2019). Previous work has also shown that resting-state PAF is reduced in populations with neuropathic pain in persistent abdominal pain as a result of chronic pancreatitis (Vries et al., 2013), neuropathic pain as a result of spinal cord injury (Boord et al., 2007, Sato et al., 2017), and increased subjective perception of tonic heat pain (Nir et al., 2010, Raghuraman et al., 2019). Interindividual differences in PAF may relate to awareness and the sampling of sensory inputs (Angelakis et al., 2004, Mierau et al., 2017), and thus may reflect an individual’s response to acute pain in a way that may influence the transition to chronic pain.

However, the existing literature is not conclusive in establishing a relationship between PAF and pain: studies on chronic back pain (Schmidt et al., 2012), central neuropathic pain in multiple sclerosis patients (Krupina et al., 2019), and persistent pain after breast cancer treatment (van den Broeke et al., 2013) did not find a relationship between slowed PAF and pain. Together the positive and negative findings of the existing literature are often interpreted as suggesting either a differential relationship between PAF and pain of different origins, or as lack of a strong relationship between PAF and pain altogether. Within the pain field there is also substantial variation in how PAF is computed and interpreted, which complicates the comparison of results across studies.

In order to interpret and compare results across the field in a meaningful way, we must first be certain that differences in findings are not simply a byproduct of differences in processing pipelines. While previously published papers have focused on developing a method that best estimates the “true” PAF value (Corcoran et al., 2018), the goal of this paper is instead to analyze previously employed methods and determine whether differences in processing decisions may impact the final PAF calculation. We reviewed 17 papers with complete available methods investigating the relationship between PAF and pain and found that there were major differences in data filtering, whether data were re-referenced, what kind of artifact removal was performed, epoching, alpha band bounds, and formula for calculating PAF. Furthermore, comparisons between PAF and other measures of neural activity are infrequent, limiting the interpretation of PAF in terms of underlying cortical and subcortical activity patterns. The thalamus is a hypothesized generator of the alpha rhythm and has been implicated in alpha rhythm alterations, but the association between thalamic activity and PAF in humans has not been thoroughly explored.

Given these uncertainties, in this study we aimed to assess the robustness of PAF to three different pre-processing pipelines as well as four processing variables (seven epoch lengths, two formulas for calculating PAF, three different alpha bands, and five regions-of-interest [ROIs]: grand-average, frontal, parietal, occipital, sensorimotor) representative of what is currently being used in the pain literature. Additionally, we aimed to associate PAF with fractional amplitude of low-frequency fluctuations (fALFF), a resting-state functional magnetic resonance imaging (rs-fMRI)-derived measure of local activity. To accomplish these aims, we analyzed data from 47 healthy men, a subset of an ongoing study of urologic chronic pelvic pain. The dataset included 64-channel resting-state electroencephalographic (EEG) and resting-state functional magnetic resonance imaging (rs-fMRI) from all participants. We first ran three of the most representative EEG pre-processing pipelines and then varied post-processing parameters (epoch length, PAF formula, alpha band bounds, and ROI) as determined by our review of the PAF and pain literature. We then determined whether there was a relationship between the ROI averages and whole brain as well as thalamic activity in healthy men.

Section snippets

Participants

Forty-seven healthy men who were the control group with complete EEG and rs-fMRI data for an ongoing study of chronic prostatitis were entered into this study. Participants were included if they were older than 18 years of age, able to participate in the informed consent process, safe to be scanned by magnetic resonance imaging, had no diagnosis of chronic prostatitis, had no active urinary, anal, or genital infection, and no severe, urgent, or debilitating medical condition. All aspects of the

Results

All processing variables of interest are reported in Table 1, and information about subject populations (sample size, clinical versus healthy, pathology if applicable) as well as the relationship between PAF and pain are reported in Supplementary Table 1. Detailed criteria and steps for the literature review process are reported in Supplementary Fig. 1.

The mean number of ICs removed using MARA for all participants was 27.85 ± 9.54 (mean±STD). The total amount of data removed due to the 80 mV

Discussion

Despite the novelty and promise of PAF, a few key issues complicate the interpretation of past data and comparison across publications in the pain field. The first issue is a lack of consistency in the pre-processing pipelines and post-processing variables for EEG data. We reviewed 17 papers investigating the relationship between PAF and pain and found that there were major differences in how labs were pre-processing their data, (re-referencing, artifact removal, etc), epoch length, bounds of

CRediT authorship contribution statement

Natalie McLain: Conceptualization, Methodology, Formal analysis, Visualization, Writing – original draft. Moheb Yani.: Software, Formal analysis, Investigation, Writing – Review & Editing. Jason Kutch: Conceptualization, Methodology, Software, Writing – original draft, Funding acquisition.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Acknowledgements

We would like to thank Arike Coker for manuscript support and helpful discussions. This work was supported, in part, by grants DK110669 and DK121724 from the National Institute of Diabetes and Digestive and Kidney Diseases (NIH/NIDDK).

References (66)

  • A.K. Hegarty et al.

    Salience network functional connectivity is spatially heterogeneous across sensorimotor cortex in healthy humans

    Neuroimage

    (2020)
  • K. Jann et al.

    Association of individual resting state EEG alpha frequency and cerebral blood flow

    Neuroimage

    (2010)
  • L.A. Kilpatrick et al.

    Alterations in resting state oscillations and connectivity in sensory and motor networks in women with interstitial cystitis/painful bladder syndrome

    J. Urol.

    (2014)
  • L.B. Kisler et al.

    Abnormal alpha band power in the dynamic pain connectome is a marker of chronic pain with a neuropathic component

    Neuroimage Clin.

    (2020)
  • W. Klimesch

    EEG alpha and theta oscillations reflect cognitive and memory performance: a review and analysis

    Brain Res. Brain Res. Rev.

    (1999)
  • A. Mierau et al.

    State-dependent alpha peak frequency shifts: experimental evidence, potential mechanisms and functional implications

    Neuroscience

    (2017)
  • A. Molins et al.

    Quantification of the benefit from integrating MEG and EEG data in minimum l2-norm estimation

    Neuroimage

    (2008)
  • N. Ngernyam et al.

    The effects of transcranial direct current stimulation in patients with neuropathic pain from spinal cord injury

    Clin. Neurophysiol.

    (2015)
  • L. Parkes et al.

    An evaluation of the efficacy, reliability, and sensitivity of motion correction strategies for resting-state functional MRI

    Neuroimage

    (2018)
  • M.C. Salinsky et al.

    Test-retest reliability in EEG frequency analysis

    Electroencephalogr. Clin. Neurophysiol.

    (1991)
  • S. Schwab et al.

    Discovering frequency sensitive thalamic nuclei from EEG microstate informed resting state fMRI

    Neuroimage

    (2015)
  • K.R.A. Van Dijk et al.

    The influence of head motion on intrinsic functional connectivity MRI

    Neuroimage

    (2012)
  • A. Vuckovic et al.

    Dynamic oscillatory signatures of central neuropathic pain in spinal cord injury

    J. Pain.

    (2014)
  • A. Vuckovic et al.

    Electroencephalographic predictors of neuropathic pain in subacute spinal cord injury

    J. Pain.

    (2018)
  • C.B. Ab Aziz et al.

    The role of the thalamus in modulating pain

    Malays. J. Med. Sci.

    (2006)
  • P. Boord et al.

    Electroencephalographic slowing and reduced reactivity in neuropathic pain following spinal cord injury

    Spinal Cord.

    (2007)
  • A.W. Corcoran et al.

    Toward a reliable, automated method of individual alpha frequency (IAF) quantification

    Psychophysiology

    (2018)
  • J. Corlier et al.

    Repetitive transcranial magnetic stimulation treatment of major depressive disorder and comorbid chronic pain: response rates and neurophysiologic biomarkers

    Psychol. Med.

    (2021)
  • B.N. Cuffin et al.

    Comparison of the magnetoencephalogram and electroencephalogram

    Electroencephalogr. Clin. Neurophysiol.

    (1979)
  • A. Eklund et al.

    Cluster failure: Why fMRI inferences for spatial extent have inflated false-positive rates

    Proc. Natl. Acad. Sci. US A

    (2016)
  • R. Freynhagen et al.

    painDETECT: a new screening questionnaire to identify neuropathic components in patients with back pain

    Curr. Med. Res. Opin.

    (2006)
  • A.J. Furman et al.

    Sensorimotor peak alpha frequency is a reliable biomarker of prolonged pain sensitivity

    Cereb. Cortex

    (2020)
  • A.J. Furman et al.

    Cerebral peak alpha frequency reflects average pain severity in a human model of sustained, musculoskeletal pain

    J. Neurophysiol.

    (2019)
  • Cited by (4)

    • The reliability of two prospective cortical biomarkers for pain: EEG peak alpha frequency and TMS corticomotor excitability

      2023, Journal of Neuroscience Methods
      Citation Excerpt :

      As there are currently no clinically meaningful brain biomarkers available for predicting the transition to chronic pain (Zhang et al., 2021), such results are promising. Indeed, there has been a recent increase in research using PAF (De Martino et al., 2021; May et al., 2021; McLain et al., 2022; Valentini et al., 2022) and CME (Cavaleri et al., 2019; Chowdhury et al., 2022; Costa et al., 2019) in the context of predicting and/or modulating pain sensitivity, suggesting evaluations of the reliability of these measures is critical. The mechanisms through which PAF and CME are associated with pain remain unclear.

    • Paired associative stimulation applied to the cortex can increase resting-state functional connectivity: A proof of principle study

      2022, Neuroscience Letters
      Citation Excerpt :

      The data was not re-referenced and coherence was calculated on the entire epoch. Previous research has shown that this method of calculating coherence does not impact the consistency of frequency based neural correlates [36]. The resulting coherence value from each electrode pair varies in value between 0 and 1 with a value closer to 1 representing stronger resting-state intracortical connectivity.

    View full text