pith. sign in

arxiv: 2605.16326 · v1 · pith:NIVBMV4Anew · submitted 2026-05-05 · 🧬 q-bio.QM · cs.AI· cs.LG· eess.SP

A Machine Learning Framework for EEG-Based Prediction of Treatment Efficacy in Chronic Neck Pain

Xiru Wang , Aiden Li , Hongzhao Tan , Stevie Foglia , Aimee Nelson , Zhen Gao This is my paper

Pith reviewed 2026-05-21 00:33 UTC · model grok-4.3

classification 🧬 q-bio.QM cs.AIcs.LGeess.SP
keywords machine learningEEGchronic neck paintreatment predictionpreprocessing pipelinemotor imageryresting-statepersonalized therapy
0
0 comments X

The pith

A machine learning framework uses tailored EEG preprocessing to predict treatment efficacy for chronic neck pain.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper presents a machine learning framework that processes electroencephalography recordings to forecast which therapies will succeed for individual patients with chronic neck pain. The approach details distinct preprocessing steps for resting-state EEG, including baseline removal, channel rejection, re-referencing, filtering, independent component analysis and power spectral density computation, and for motor execution or imagery tasks, with added quantification of event-related desynchronization and synchronization plus electromyography correlation. An accompanying review of prior machine learning studies on clinical EEG informs the overall strategy. A sympathetic reader would care because current chronic neck pain care relies on trial-and-error selection of treatments, which this framework aims to replace with data-driven personalization that could lower disability rates and ease demands on healthcare resources.

Core claim

The paper claims that a combined preprocessing and review effort will develop a robust predictive model that uses EEG to forecast treatment efficacy in chronic neck pain and thereby support personalized healthcare strategies.

What carries the argument

The tailored EEG preprocessing pipeline, which removes baseline signals, identifies and excludes bad channels, re-references, applies bandpass and notch filters plus independent component analysis for resting-state data, quantifies event-related desynchronization and synchronization after aligning motor task signals to triggers, and correlates smoothed electromyography with EEG channels.

If this is right

  • Enables selection of effective treatments for chronic neck pain on an individualized basis rather than through repeated trials.
  • Reduces overall burden on healthcare systems by improving the match between patients and therapies.
  • Supports development of personalized healthcare strategies specifically for chronic pain management.
  • Provides a template for incorporating both resting-state and task-based EEG signals into clinical prediction tasks.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same preprocessing steps could be tested for predicting treatment response in other forms of chronic pain or related neurological conditions.
  • Collecting follow-up outcome data from treated patients would allow direct training and validation of the predictive models on real efficacy labels.
  • Correlations between EEG and electromyography might highlight motor-related markers that influence which therapies succeed in pain relief.

Load-bearing premise

The preprocessing steps will extract EEG features that contain information predictive of treatment efficacy.

What would settle it

A machine learning model trained on the preprocessed features predicts actual treatment outcomes in new patients at no better than chance accuracy.

Figures

Figures reproduced from arXiv: 2605.16326 by Aiden Li, Aimee Nelson, Hongzhao Tan, Stevie Foglia, Xiru Wang, Zhen Gao.

Figure 1
Figure 1. Figure 1: Standard 10–20 system montage for the 64-channel recording configuration used in this study. The two earlobe electrodes A1 and A2 (visible at the left and right of the head outline) serve as references for the remaining channels [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: Time-courses of 15 Independent Components (ICA000–ICA014) decomposed from a single subject's EEG recording. Component ICA000 (top trace) exhibits high-amplitude transient deflections characteristic of eye￾blink and movement artifacts, and would be excluded from the reconstructed signal. 2.5 Power Spectral Density Analysis After artifact repair, resting-state EEG was characterised in the frequency domain ov… view at source ↗
Figure 3
Figure 3. Figure 3: Scalp topographies of average band power computed on resting-state EEG for the four classical frequency bands (Delta 0–4 Hz, Theta 4–8 Hz, Alpha 8–12 Hz, Beta 12–30 Hz). Colour scales are in μV²/Hz. A complete preprocessing pipeline for the resting-state EEG was implemented in Python within the Jupyter Notebook environment. EEG recordings were imported from HDF5 files using the h5py library. Most preproces… view at source ↗
Figure 4
Figure 4. Figure 4: Scalp topographies of trigger-locked spectrotemporal power for the four frequency bands (delta, theta, alpha, beta). Topographies are computed by averaging baseline-corrected spectrotemporal representations across epochs and across all 1 Hz bins within each band. Colour scale in arbitrary units [PITH_FULL_IMAGE:figures/full_fig_p007_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Per-channel time-courses of trigger-locked event-related desynchronization/synchronization (ERDS) for delta, theta, alpha and beta bands. Time 0 marks trigger onset; the −1 to 0 s window is used as the baseline. Each coloured line corresponds to one EEG channel. EMG measures the electrical activity generated when motor neurons stimulate skeletal muscle fibres. Higher absolute correlations between the EEG o… view at source ↗
Figure 6
Figure 6. Figure 6: Preprocessed EEG and EMG signals within a single trigger-locked epoch. The dashed grey vertical line marks trigger onset. The green-shaded region indicates the analysis window over which EEG–EMG correlations are computed. Top: EMG channel; Bottom: a subset of central EEG channels (C1–C6, Cz) [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Scalp topography of absolute EEG–EMG correlation values, averaged across all epochs, for each of the recorded EEG channels. Higher values (darker red) indicate stronger cortico-muscular coupling during the cued movement attempt [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗
read the original abstract

Chronic neck pain is a leading cause of disability worldwide, and current treatment selection remains largely trial and error. We present a machine learning framework that uses electroencephalography to predict treatment efficacy in patients with chronic neck pain, with the goal of supporting individualized therapy and reducing the burden on healthcare systems. The framework centers on a rigorous data preprocessing stage tailored to the characteristics of each EEG recording type. For resting-state EEG, the preprocessing pipeline comprises baseline signal removal, bad channel identification and exclusion, re-referencing, bandpass and notch filtering, Independent Component Analysis, and power spectral density analysis. For motor execution and motor imagery recordings, the same initial steps are applied, after which signals are aligned to trigger events so that event-related desynchronization (ERD) and event-related synchronization (ERS) can be quantified. Synchronously recorded electromyography data are bandpass filtered and smoothed with a moving average, then correlated with the corresponding EEG channels to characterize the EEG EMG relationship during attempted movement. In parallel, we performed an extensive literature review of machine learning models applied to clinical EEG (763 records initially screened, 16 patient and 47 healthy-control studies retained), to inform the post-processing strategy. Through this combined preprocessing and review effort, we aim to develop a robust predictive model that can support personalized healthcare strategies in chronic pain management.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a machine learning framework for EEG-based prediction of treatment efficacy in chronic neck pain. It details preprocessing pipelines for resting-state EEG (baseline removal, bad channel identification, re-referencing, bandpass/notch filtering, ICA, and PSD analysis) and for motor execution/imagery tasks (event alignment, ERD/ERS quantification, and EEG-EMG correlation after bandpass filtering and smoothing of EMG). An extensive literature review (763 records screened, 16 patient and 47 healthy-control studies retained) is performed to inform post-processing and model strategy, with the stated goal of developing a robust predictive model to support personalized therapy.

Significance. If implemented and validated, the framework could contribute to reducing trial-and-error in chronic pain treatment by identifying EEG-derived biomarkers. The preprocessing steps are standard and well-motivated for artifact handling and feature extraction, while the literature review provides a broad foundation for algorithm selection. The work is prospective rather than empirical, so its immediate impact is limited to methodological guidance.

major comments (2)
  1. [Abstract] Abstract and overall framework description: the claim that the described preprocessing plus literature review will enable a 'robust predictive model' rests on the untested assumption that features such as PSD, ERD/ERS, and EEG-EMG correlations contain information predictive of treatment outcomes; no patient data application, classifier training, cross-validation, or performance metrics (accuracy, AUC, etc.) are reported anywhere in the manuscript.
  2. [Framework Description] Preprocessing sections for resting-state and motor tasks: while the steps (ICA, PSD, ERD/ERS quantification) are listed, there is no specification of how components are selected/rejected in ICA, how ERD/ERS time windows or frequency bands are chosen for chronic neck pain patients, or any validation that these features survive artifact correction in this population.
minor comments (2)
  1. [Literature Review] The literature review retention numbers (16 patient + 47 healthy-control studies) are stated but the manuscript does not indicate which specific models or features from those studies are adopted in the proposed pipeline.
  2. Notation for ERD/ERS quantification and EEG-EMG correlation is introduced without explicit formulas or pseudocode, which would improve reproducibility.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive review and for recognizing the potential of this prospective framework. We address the major comments point by point below, clarifying the scope of the work as a methodological proposal informed by literature rather than an empirical validation study.

read point-by-point responses

  1. Referee: [Abstract] Abstract and overall framework description: the claim that the described preprocessing plus literature review will enable a 'robust predictive model' rests on the untested assumption that features such as PSD, ERD/ERS, and EEG-EMG correlations contain information predictive of treatment outcomes; no patient data application, classifier training, cross-validation, or performance metrics (accuracy, AUC, etc.) are reported anywhere in the manuscript.

    Authors: We agree that the manuscript describes a proposed framework and does not report empirical results from patient data, model training, or performance metrics. The abstract and text state the goal of developing such a model through the outlined preprocessing and literature review (763 records screened, 16 patient studies retained). To address the concern, we will revise the abstract and introduction to explicitly frame the contribution as a methodological pipeline intended to support future empirical validation, removing any implication of immediate predictive performance. The feature choices are motivated by the retained literature on EEG in pain and motor tasks, but we acknowledge these remain untested in this specific application. revision: yes

  2. Referee: [Framework Description] Preprocessing sections for resting-state and motor tasks: while the steps (ICA, PSD, ERD/ERS quantification) are listed, there is no specification of how components are selected/rejected in ICA, how ERD/ERS time windows or frequency bands are chosen for chronic neck pain patients, or any validation that these features survive artifact correction in this population.

    Authors: We accept that greater specificity is needed. In the revised manuscript we will expand the preprocessing sections to detail ICA component rejection criteria (e.g., based on topographic maps, temporal dynamics, and spectral signatures consistent with ocular or muscular artifacts, following conventions in the reviewed studies). For ERD/ERS we will specify the frequency bands (mu 8–12 Hz and beta 13–30 Hz) and time windows (typically 0–2 s post-event) drawn from the 47 healthy-control and 16 patient studies retained in the literature review. While direct validation of feature robustness after correction in chronic neck pain patients is not possible in the current prospective manuscript and will require future data collection, we will add a note referencing how these parameters have been shown to survive similar preprocessing in related clinical EEG literature. revision: yes

Circularity Check

0 steps flagged

No circularity: forward-looking framework proposal with no derivations or fitted results

full rationale

The paper describes a proposed EEG preprocessing pipeline (baseline removal, ICA, PSD, ERD/ERS, EEG-EMG correlation) and a literature review to inform future model development, but contains no equations, no model training, no fitted parameters, no predictions on data, and no self-citations that bear load on any central claim. The manuscript is explicitly a framework proposal without claiming achieved results or reductions of outputs to inputs by construction. This is the normal case of a self-contained descriptive proposal.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The description relies on standard domain assumptions from EEG signal processing and clinical machine learning rather than introducing new free parameters, axioms, or entities.

axioms (1)
  • domain assumption Independent Component Analysis can reliably separate neural signals from artifacts in EEG recordings.
    Invoked as part of the resting-state preprocessing pipeline.

pith-pipeline@v0.9.0 · 5788 in / 1150 out tokens · 38327 ms · 2026-05-21T00:33:29.805365+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

  • IndisputableMonolith/Cost/FunctionalEquation.lean washburn_uniqueness_aczel unclear
    ?
    unclear

    Relation between the paper passage and the cited Recognition theorem.

    For resting-state EEG, the preprocessing pipeline comprises baseline signal removal, bad channel identification and exclusion, re-referencing, bandpass and notch filtering, Independent Component Analysis, and power spectral density analysis.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

23 extracted references · 23 canonical work pages

  1. [1]

    Global, regional, and national burden of neck pain, 1990 –2020, and projections to 2050: a systematic analysis of the Global Burden of Disease Study 2021,

    GBD 2021 Neck Pain Collaborators, "Global, regional, and national burden of neck pain, 1990 –2020, and projections to 2050: a systematic analysis of the Global Burden of Disease Study 2021," The Lancet Rheumatology, vol. 6, no. 3, pp. e142–e155, 2024, doi: 10.1016/S2665-9913(23)00321-1

    work page doi:10.1016/s2665-9913(23)00321-1 2021

  2. [2]

    Global, regional, and national burden of neck pain in the general population, 1990–2017: systematic analysis of the Global Burden of Disease Study 2017,

    S. Safiri, A. A. Kolahi, D. Hoy, et al., "Global, regional, and national burden of neck pain in the general population, 1990–2017: systematic analysis of the Global Burden of Disease Study 2017," BMJ, vol. 368, p. m791, 2020, doi: 10.1136/bmj.m791

    work page doi:10.1136/bmj.m791 1990

  3. [3]

    A systematic review of the effect of waiting for treatment for chronic pain,

    M. E. Lynch, F. Campbell, A. J. Clark, et al., "A systematic review of the effect of waiting for treatment for chronic pain," Pain, vol. 136, no. 1–2, pp. 97–116, May 2008, doi: 10.1016/j.pain.2007.06.018

    work page doi:10.1016/j.pain.2007.06.018 2008

  4. [4]

    Brain imaging biomarkers for chronic pain,

    I. Mackey, P. Hassan, L. Bertulis -Fernandes, et al., "Brain imaging biomarkers for chronic pain," Frontiers in Neurology, vol. 12, p. 734821, 2022, doi: 10.3389/fneur.2021.734821

    work page doi:10.3389/fneur.2021.734821 2022

  5. [5]

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

    A. Mussigmann, B. Bardel, and J.-P. Lefaucheur, "In search of a composite biomarker for chronic pain by way of EEG and machine learning: where do we currently stand?" Frontiers in Neuroscience, vol. 17, p. 1186418, 2023, doi: 10.3389/fnins.2023.1186418

    work page doi:10.3389/fnins.2023.1186418 2023

  6. [6]

    Exploring electroencephalographic chronic pain biomarkers: a mega - analysis,

    E. S. May, S. Taesler, S. Tiemann, et al. , "Exploring electroencephalographic chronic pain biomarkers: a mega - analysis," eBioMedicine, vol. 120, p. 105902, 2025, doi: 10.1016/j.ebiom.2025.105902

    work page doi:10.1016/j.ebiom.2025.105902 2025

  7. [7]

    Deep learning -based electroencephalography analysis: a systematic review,

    Y. Roy, H. Banville, I. Albuquerque, A. Gramfort, T. H. Falk, and J. Faubert, "Deep learning -based electroencephalography analysis: a systematic review," Journal of Neural Engineering, vol. 16, no. 5, p. 051001, 2019, doi: 10.1088/1741-2552/ab260c. 15

    work page doi:10.1088/1741-2552/ab260c 2019

  8. [8]

    An information -maximization approach to blind separation and blind deconvolution,

    A. J. Bell and T. J. Sejnowski, "An information -maximization approach to blind separation and blind deconvolution," Neural Computation, vol. 7, no. 6, pp. 1129– 1159, Nov. 1995, doi: 10.1162/neco.1995.7.6.1129

    work page doi:10.1162/neco.1995.7.6.1129 1995

  9. [9]

    Removing electroencephalographic artifacts by blind source separation,

    T.-P. Jung, S. Makeig, C. Humphries, T. -W. Lee, M. J. McKeown, V. Iragui, and T. J. Sejnowski, "Removing electroencephalographic artifacts by blind source separation," Psychophysiology , vol. 37, no. 2, pp. 163 –178, Mar. 2000, doi: 10.1111/1469-8986.3720163

    work page doi:10.1111/1469-8986.3720163 2000

  10. [10]

    The ten -twenty electrode system of the International Federation,

    H. H. Jasper, "The ten -twenty electrode system of the International Federation," Electroencephalography and Clinical Neurophysiology, vol. 10, pp. 371–375, 1958

    work page 1958

  11. [11]

    Ardizzone, M.G

    G. Pfurtscheller, "Event-related synchronization (ERS): an electrophysiological correlate of cortical areas at rest," Electroencephalography and Clinical Neurophysiology, vol. 83, no. 1, pp. 62–69, Jul. 1992, doi: 10.1016/0013- 4694(92)90133-3

    work page doi:10.1016/0013- 1992

  12. [12]

    Event -related EEG/MEG synchronization and desynchronization: basic principles,

    G. Pfurtscheller and F. H. Lopes da Silva, "Event -related EEG/MEG synchronization and desynchronization: basic principles," Clinical Neurophysiology, vol. 110, no. 11, pp. 1842–1857, Nov. 1999, doi: 10.1016/S1388- 2457(99)00141-8

    work page doi:10.1016/s1388- 1999

  13. [13]

    Removal of artifacts from EEG signals: a review,

    X. Jiang, G.-B. Bian, and Z. Tian, "Removal of artifacts from EEG signals: a review," Sensors, vol. 19, no. 5, p. 987, Feb. 2019, doi: 10.3390/s19050987

    work page doi:10.3390/s19050987 2019

  14. [14]

    The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms,

    P. D. Welch, "The use of fast Fourier transform for the estimation of power spectra: a method based on time averaging over short, modified periodograms," IEEE Transactions on Audio and Electroacoustics, vol. AU-15, no. 2, pp. 70–73, Jun. 1967, doi: 10.1109/TAU.1967.1161901

    work page doi:10.1109/tau.1967.1161901 1967

  15. [15]

    MEG and EEG data analysis with MNE -Python,

    A. Gramfort, M. Luessi, E. Larson, et al. , "MEG and EEG data analysis with MNE -Python," Frontiers in Neuroscience, vol. 7, p. 267, Dec. 2013, doi: 10.3389/fnins.2013.00267

    work page doi:10.3389/fnins.2013.00267 2013

  16. [16]

    S. J. Luck, An Introduction to the Event-Related Potential Technique, 2nd ed. Cambridge, MA, USA: MIT Press, 2014

    work page 2014

  17. [17]

    An EEG-EMG correlation-based brain-computer interface for hand orthosis supported neuro-rehabilitation,

    A. Chowdhury, H. Raza, Y. K. Meena, A. Dutta, and G. Prasad, "An EEG-EMG correlation-based brain-computer interface for hand orthosis supported neuro-rehabilitation," Journal of Neuroscience Methods, vol. 312, pp. 1– 11, Jan. 2019, doi: 10.1016/j.jneumeth.2018.11.010

    work page doi:10.1016/j.jneumeth.2018.11.010 2019

  18. [18]

    Support-vector networks,

    C. Cortes and V. Vapnik, "Support-vector networks," Machine Learning, vol. 20, no. 3, pp. 273–297, Sep. 1995, doi: 10.1007/BF00994018

    work page doi:10.1007/bf00994018 1995

  19. [19]

    Machine Learning , author =

    L. Breiman, "Random forests," Machine Learning, vol. 45, no. 1, pp. 5 –32, Oct. 2001, doi: 10.1023/A:1010933404324

    work page doi:10.1023/a:1010933404324 2001

  20. [20]

    Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining , pages =

    T. Chen and C. Guestrin, "XGBoost: a scalable tree boosting system," in Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining (KDD '16), San Francisco, CA, USA, Aug. 2016, pp. 785–794, doi: 10.1145/2939672.2939785

    work page doi:10.1145/2939672.2939785 2016

  21. [21]

    1997, Neural Computation, 9, 1735, 10.1162/neco.1997.9.8.1735

    S. Hochreiter and J. Schmidhuber, "Long short- term memory," Neural Computation, vol. 9, no. 8, pp. 1735 – 1780, Nov. 1997, doi: 10.1162/neco.1997.9.8.1735

    work page doi:10.1162/neco.1997.9.8.1735 1997

  22. [22]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, et al. , "Attention is all you need," in Advances in Neural Information Processing Systems (NeurIPS), vol. 30, 2017, pp. 5998–6008

    work page 2017

  23. [23]

    Shared neural signatures in functional neurological disorder and chronic pain: a multimodal narrative review,

    S. Kannan, K. Patel, D. Di Basilio, et al., "Shared neural signatures in functional neurological disorder and chronic pain: a multimodal narrative review," BMJ Neurology Open, vol. 7, no. 2, p. e001032, 2025, doi: 10.1136/bmjno- 2025-001032

    work page doi:10.1136/bmjno- 2025