Pith. sign in

REVIEW 3 major objections 5 minor 30 references

Feasibility of simultaneous EEG-fMRI at 0.55 T: Recording, Denoising, and Functional Mapping

T0 review · 3 major / 5 minor · reviewed 2026-08-04 · deepseek-v4-flash

Pith's one-line read The paper claims simultaneous EEG-fMRI is feasible at 0.55 T, that the ballistocardiogram artifact is smaller there than at 3 T, and that a 12 Hz visual EEG power envelope tracks BOLD in visual cortex.

desk verdict First 0.55T EEG-fMRI feasibility study with solid artifact characterization; the EEG-BOLD coupling claim is an over-read because the EEG regressor is collinear with the stimulus design. read the letter →

arxiv 2602.13489 v2 pith:B2PBWWXD submitted 2026-02-13 eess.SP

classification eess.SP
keywords simultaneousEEG-fMRI0.55Tballistocardiogramartifactgradientsteady-statevisualevokedpotentialBOLDneurovascularcouplingmid-field
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

This proof-of-concept study asks whether simultaneous EEG and fMRI can be recorded on a modern 0.55 T scanner despite the artifacts MRI introduces into the EEG. Using a flickering checkerboard task in two participants, the authors show that standard gradient and heartbeat artifact subtraction preserves the alpha rhythm and the 12 Hz steady-state visual evoked potential. They further show that the power envelope of that 12 Hz EEG signal, convolved with the hemodynamic response function, produces a V1 activation map similar to the standard block-design map. The authors argue that the lower field strength reduces the ballistocardiogram artifact, making 0.55 T a promising platform for multimodal neuroimaging, especially for people with implanted devices. If correct, this would extend EEG-fMRI to settings and populations where high-field scanners are impractical.

What carries the argument

The load-bearing mechanism is the artifact-reduction chain: gradient artifacts are removed by average artifact subtraction synchronized to the MRI clock, the ballistocardiogram is subtracted using the ECG channel, and independent component analysis cleans residuals. The multimodal link is made by the EEG-informed predictor—the amplitude envelope of the 12 Hz steady-state visual evoked potential at Oz, band-pass filtered, Hilbert-transformed, downsampled to the 3-s TR, and convolved with a canonical hemodynamic response function. This predictor is the bridge between electrophysiology and BOLD, and the paper's central comparison is its correlation map versus the standard boxcar map.

What would settle it

Record the same visual task at 0.55 T with a design in which the 12 Hz flicker is modulated independently of the block schedule (e.g., amplitude-modulated or random-duration stimulation); if the EEG predictor retains unique explained variance in a joint GLM with the boxcar, the neurovascular-coupling claim is supported, and if it does not, the reported spatial correspondence is explained by timing alone. Separately, measure peak BCG amplitude in the same participant at 0.55 T and 3 T with identical electrodes and cap; a null or reversed difference would falsify the BCG-reduction claim.

Watch

Extended reading notes

Core claim

On its own terms, the paper establishes three connected findings: (1) task-based BOLD activation in visual cortex is detectable at 0.55 T when ten minutes of task data are concatenated; (2) the ballistocardiogram artifact amplitude is qualitatively smaller than typical 3 T reports, and standard average-artifact-subtraction plus ICA clean the EEG while preserving physiological rhythms; and (3) a data-driven regressor built from the 12 Hz SSVEP power at electrode Oz predicts the BOLD time course in visual cortex in a pattern matching the stimulus boxcar. The authors state plainly that this is a pilot proof-of-concept with N=2, and the headline claim is that combined EEG-fMRI at 0.55 T is feasi

Load-bearing premise

The load-bearing premise is that the 12 Hz SSVEP envelope is an independent neural regressor rather than a restatement of the stimulus schedule, and that the BCG reduction claim can be trusted from qualitative comparison to published 3 T reports rather than a same-subject baseline.

Editorial extensions

If this is right

  • Standard EEG artifact correction pipelines, developed for high-field scanners, transfer directly to 0.55 T and leave the alpha rhythm and SSVEP intact.
  • Task-based fMRI at 0.55 T requires longer acquisitions (here, two 5-minute runs concatenated) to reach activation strengths comparable to higher-field studies.
  • The 12 Hz SSVEP power envelope can be used as a data-driven regressor for BOLD at 0.55 T, opening the door to EEG-informed functional mapping at mid-field.
  • If the BCG reduction is confirmed, 0.55 T may support EEG-fMRI in patients with ferromagnetic implants or other populations excluded from 3 T and 7 T scanners.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Inference: because the checkerboard flickers in 24-s blocks, the 12 Hz envelope is strongly driven by the stimulus schedule; after HRF convolution the EEG predictor is nearly collinear with the boxcar, so the spatial overlap in Fig. 6 is expected from timing alone and does not independently prove neurovascular coupling. A task with a neural regressor that is decoupled from the stimulus (e.g., rand
  • Inference: the BCG 'reduction' is reported by qualitative comparison to typical 3 T literature values, not by a within-study head-to-head. Repeating the same EEG protocol in the same subject at 0.55 T and 3 T would settle the claim quantitatively.
  • Inference: the feasibility result, if replicated, reframes the usual trade-off—0.55 T sacrifices BOLD SNR but gains a cleaner electromagnetic environment; extended scan time can partly compensate for the SNR loss while the EEG benefits persist, which is a different cost-geometry than at high field.
  • Inference: a natural extension would be to test whether the reduced BCG allows shorter or no ECG-based subtraction, and whether cleaned EEG preserves low-frequency (delta/theta) and event-related potentials that are often degraded at 3 T.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. This proof-of-concept manuscript reports simultaneous EEG-fMRI recordings on a 0.55 T scanner in two healthy participants performing a 12 Hz flickering checkerboard block-design visual task. The authors characterize gradient and ballistocardiogram (BCG) artifacts, describe a denoising pipeline based on average artifact subtraction, pulse artifact subtraction, and ICA, and report preserved alpha rhythm and a focal occipital SSVEP power modulation after cleaning. They further demonstrate statistically significant BOLD activation in visual cortex from 10 minutes of concatenated task data, and they construct an EEG-informed regressor from the 12 Hz SSVEP envelope at Oz, which they correlate voxel-wise with BOLD. The spatial map from this EEG predictor is shown to resemble the standard block-design activation map, and the authors interpret this correspondence as evidence of neurovascular coupling and as validation of the multimodal integration pipeline at 0.55 T.

Significance. If the central claims hold, the paper would provide a valuable demonstration that simultaneous EEG-fMRI is technically feasible on modern high-performance 0.55 T systems, with a potentially lighter BCG artifact burden than at 3 T, and would support the use of mid-field scanners for multimodal neuroimaging in populations poorly served by high-field systems. The strengths of the study include the use of real simultaneous recordings rather than simulations, the characterization of EEG spectral content (alpha and SSVEP harmonics) after denoising, and the explicit use of extended acquisition to compensate for reduced BOLD sensitivity. However, the multimodal validation is weakened by a constructional confounding: the EEG predictor is derived from the same 12 Hz stimulus envelope that defines the block-design timing, so the observed spatial correspondence does not independently establish EEG-BOLD coupling. The BCG reduction claim similarly lacks a quantitative baseline. These issues are localizable and fixable, but they are load-bearing for the paper's most novel conclusions.

major comments (3)
  1. [Methods II.D and Fig. 6] The EEG-informed predictor is constructed from the 12 Hz SSVEP power envelope at Oz during a 24 s on/off checkerboard block design (Methods II.D.1). Because the flicker frequency equals the stimulus frequency and the envelope is averaged over 24 s blocks, the envelope tracks the boxcar timing almost exactly; after convolution with the SPM-HRF, the 'data-driven' regressor is nearly collinear with the block-design regressor, with correlation likely exceeding 0.9. Consequently, the spatial correspondence between maps A and B in Fig. 6 is expected from stimulus timing alone and cannot validate EEG-BOLD coupling or neurovascular coupling. The statement in Methods II.D.2 that these are 'two independent spatial maps' is inaccurate. The authors should report the actual correlation between the two predictors, include both regressors in a single GLM to assess incremental variance, and/or use a par
  2. [Results III.B, Discussion IV] The claim that the BCG artifact is reduced at 0.55 T compared to conventional 3 T systems is supported only by qualitative reference to 'typical reports from high-field (3T) systems.' No in-study 3 T baseline or quantitative metric (e.g., peak-to-peak artifact amplitude, RMS, artifact-to-signal ratio, or spatial topography) is provided. Since this reduced BCG burden is a central feasibility advantage motivating the work, the claim needs quantitative support: either a same-protocol 3 T recording analyzed with the same pipeline, or a clearly specified published dataset quantified with the same metric. As written, the claim is not falsifiable from the presented data.
  3. [Results III.C and Fig. 6] The multimodal spatial-correspondence result is reported as 'similar spatial pattern' based on single-participant maps, without any quantitative overlap measure or reproducibility statistic. With N=2, the manuscript cannot support generalizable statements about the reliability of EEG-informed mapping at 0.55 T. Please report per-participant activation maps, a quantitative overlap measure (e.g., Dice coefficient between EEG-derived and block-design clusters), and the correlation between regressors. The Discussion already acknowledges the pilot size, but the main conclusions should be tempered accordingly.
minor comments (5)
  1. [Fig. 6 caption] The caption states 'from a 5-minute run,' while Methods II.A/II.C describe a 10-minute concatenated dataset. Please clarify whether the map is from a single 5-min run or the concatenated data.
  2. [Methods II.C] The ICA step is described only as 'applied to isolate residual artifacts' with 'visual inspection.' Please specify the number of components retained/rejected, the criteria for component classification, and the number of bad channels excluded per participant.
  3. [Results III.B and Fig. 4] The PSD plots are shown without error bars or statistical comparison across participants. Since the claim is about signal preservation, including per-participant spectra or confidence intervals would strengthen the presentation.
  4. [Abstract and Introduction] The abstract and introduction use 'demonstrate' for BCG reduction and multimodal coupling. Given the pilot nature and the confounds noted above, 'suggest' or 'provide preliminary evidence' would be more accurate.
  5. [Throughout] There are minor typos and formatting artifacts (e.g., 'totheB 0 field inhomogeneities' in the Introduction; 'electrophysiologicalpower' missing space in the Discussion). A careful copyedit is advised.

Circularity Check

1 steps flagged · score 6.0 of 10

EEG-informed predictor is collinear with the 24 s block design, so the EEG-BOLD spatial correspondence is forced by construction.

  1. renaming known result [Methods II.D.1–II.D.2; Results III.C (Fig. 6); Discussion IV]
    "the preprocessed Oz time series was bandpass filtered between 11 Hz and 13 Hz (20th-order IIR filter) to isolate the target checkerboard frequency; ... this power envelope was downsampled to the fMRI repetition time (TR = 3.0 s) and convolved with a canonical Hemodynamic Response Function (SPM-HRF) ... This generated two independent spatial maps representing voxels synchronized with the experimental timing and voxels synchronized with the subject’s actual neural power at 12 Hz."

    The checkerboard is flickered at 12 Hz in 24 s blocks alternating with rest (II.A). The 11–13 Hz SSVEP envelope of Oz is therefore a filtered version of the task boxcar; after convolution with the same SPM-HRF it is nearly collinear with the boxcar predictor. Calling the resulting maps 'two independent spatial maps' is incorrect: any stimulus-driven BOLD activation will correlate with both regressors regardless of whether the EEG contains genuine neural information. The V1 correspondence in Fig. 6 is thus a replay of the block design and cannot validate EEG-BOLD/neurovascular coupling.

full rationale

The recording/denoising and artifact-characterization parts of the paper are self-contained and credible: PSD shows alpha preservation, SSVEP topography is focal over occipital electrodes, and ICA/AAS pipelines are standard. The BCG-reduction claim is weaker (no in-study 3T baseline), but that is an unsupported comparison, not circularity. The load-bearing circular step is the multimodal validation: the EEG predictor is constructed from the 12 Hz SSVEP elicited by the 24 s checkerboard blocks, so it encodes the same boxcar timing as the 'hypothesis-driven' predictor. The observed spatial similarity (Fig. 6) is therefore expected by construction and does not demonstrate that the EEG power envelope carries independent information about hemodynamic coupling. This reduces the central EEG-BOLD claim to a restatement of the block-design result, although the feasibility conclusions and denoising evidence remain independent.

Assumptions & free parameters 0 free parameters · 4 assumptions · 0 invented entities

No model parameters are fitted to data. The canonical SPM-HRF, 11-13 Hz band, 20th-order IIR filter, TR=3s, and 10-minute concatenation are preset pipeline choices from prior tools/literature. This does not compensate for missing quantitative comparisons or N=2 variability.

assumptions (4)
  • domain assumption BCG artifact amplitude scales with B0, so artifact magnitude at 0.55T is lower than at 3T.
    Invoked in the introduction and results to interpret the observed artifact; not directly measured against 3T in this study.
  • domain assumption Standard AAS + pulse-subtraction + ICA preserves true neural EEG after artifact removal at 0.55T.
    Used to claim alpha rhythm and SSVEP fidelity; no quantitative validation of residual artifact or neural source.
  • domain assumption Oz 12 Hz SSVEP envelope is a valid neural driver for visual-cortex BOLD.
    Used to build the EEG-informed predictor; confounded by stimulus-locked task design (Methods II.D).
  • domain assumption 10-minute concatenated acquisition yields BOLD sensitivity comparable to higher field.
    Relies on the authors' own prior ISMRM abstract [19], not replicated in this paper.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Feasibility of simultaneous EEG-fMRI at 0.55 T: Recording, Denoising, and Functional Mapping." pith.science (2026). https://pith.science/paper/B2PBWWXD

@misc{pith2026260213489,
  author       = {Pith},
  title        = {Pith review of: Feasibility of simultaneous EEG-fMRI at 0.55 T: Recording, Denoising, and Functional Mapping},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B2PBWWXD}},
  note         = {Machine review of arXiv:2602.13489}
}
read the original abstract

Simultaneous recording of electroencephalography (EEG) and functional MRI (fMRI) can provide a more complete view of brain function by merging high temporal and spatial resolutions. This proof-of-concept study presents initial evidence for the feasibility of simultaneous EEG-fMRI at 0.55T in a visual task. We characterize the gradient and ballistocardiogram (BCG) artifacts inherent to this environment and demonstrate that the lower field strength suggests a reduction in the magnitude of the BCG artifact compared to high-field (1.5T, 3T, 7T) systems. This reduction shows promise for facilitating effective denoising while preserving signal integrity. Furthermore, we tested a multimodal integration pipeline that uses the EEG power envelope to compute a predictor of the hemodynamic BOLD response, demonstrating the potential for EEG-based estimation of neurovascular coupling in this environment. We demonstrate that combined EEG-fMRI at 0.55T is feasible and represents a promising environment for multimodal neuroimaging.

Figures

Figures reproduced from arXiv: 2602.13489 by the authors.

Figure 2
Figure 2. Representative example of EEG signal showing the raw signal across [PITH_FULL_IMAGE:figures/full_fig_p002_2.png] view at source ↗
Figure 3
Figure 3. Visual activation and statistical reliability at 0.55T. Representative whole-brain axial montage of t-statistic maps from a single participant, generated [PITH_FULL_IMAGE:figures/full_fig_p003_3.png] view at source ↗
Figure 6
Figure 6. Multimodal functional mapping and neurovascular coupling at 0.55 T. [PITH_FULL_IMAGE:figures/full_fig_p004_6.png] view at source ↗
Figures from the paper (1 more)
Figure 5
Figure 5. Figure 5: Topographical distribution of SSVEP power modulation at 0.55 T. [PITH_FULL_IMAGE:figures/full_fig_p004_5.png]

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

30 extracted references · 1 canonical work pages

  1. [1]

    Simultaneous EEG-fMRI: What Have We Learned and What Does the Future Hold?

    T. Warbrick, “Simultaneous EEG-fMRI: What Have We Learned and What Does the Future Hold?”Sensors, vol. 22, no. 6, p. 2262, 2022

  2. [2]

    An open-access dataset of naturalistic viewing using simultaneous EEG-fMRI,

    Q. K. Telesford, et al., “An open-access dataset of naturalistic viewing using simultaneous EEG-fMRI,”Sci. Data, vol. 10, no. 1, p. 554, 2023

  3. [3]

    Clinical rationale for very-high-field (3.0 Tesla) functional magnetic resonance imaging,

    K. R. Thulborn, “Clinical rationale for very-high-field (3.0 Tesla) functional magnetic resonance imaging,”Top. Magn. Reson. Imaging, vol. 10, no. 1, pp. 37–50, 1999

  4. [4]

    The future of ultra-high field MRI and fMRI for study of the human brain,

    J. H. Duyn, “The future of ultra-high field MRI and fMRI for study of the human brain,”Neuroimage, vol. 62, no. 2, pp. 1241–1248, 2012

  5. [5]

    Imaging at ultrahigh magnetic fields: History, challenges, and solutions,

    K. Ugurbil, “Imaging at ultrahigh magnetic fields: History, challenges, and solutions,”Neuroimage, vol. 168, pp. 7–32, 2018

  6. [6]

    The impact of ultra-high field MRI on cognitive and computational neuroimaging,

    F. De Martino, et al., “The impact of ultra-high field MRI on cognitive and computational neuroimaging,”Neuroimage, vol. 168, pp. 366–382, 2018

  7. [7]

    Opportunities in interventional and diagnostic imaging by using high-performance low-field-strength MRI,

    A. E. Campbell-Washburn, et al., “Opportunities in interventional and diagnostic imaging by using high-performance low-field-strength MRI,” Radiology, vol. 293, no. 2, pp. 384–393, 2019

  8. [8]

    ICA-based procedures for removing ballisto- cardiogram artifacts from EEG data acquired in the MRI scanner,

    G. Srivastava, et al., “ICA-based procedures for removing ballisto- cardiogram artifacts from EEG data acquired in the MRI scanner,” Neuroimage, vol. 24, no. 1, pp. 50–60, 2005

Show all 30 references
  1. [9]

    Imaging in the presence of magnetic field in- homogeneities,

    B. P. Sutton and F. Lam, “Imaging in the presence of magnetic field in- homogeneities,”Magnetic Resonance Image Reconstruction, Academic Press, pp. 327–354, 2022

  2. [10]

    Low-cost high-performance MRI,

    M. Sarracanie, et al., “Low-cost high-performance MRI,”Sci. Rep., vol. 5, p. 15177, 2015

  3. [11]

    Low-field MRI: An MR physics perspective,

    J. P. Marques, et al., “Low-field MRI: An MR physics perspective,”J. Magn. Reson. Imaging, vol. 49, no. 6, pp. 1528–1542, 2019

  4. [12]

    Low-cost and portable MRI,

    L. L. Wald, et al., “Low-cost and portable MRI,”J. Magn. Reson. Imaging, vol. 52, no. 3, pp. 686–696, 2020

  5. [13]

    When less is more: the view of MRI vendors on low-field MRI,

    J. P. Marques, et al., “When less is more: the view of MRI vendors on low-field MRI,”MAGMA, vol. 34, no. 4, pp. 479–482, 2021

  6. [14]

    Optimization of Gradient-Echo Echo-Planar Imaging for T2* Contrast in the Brain at 0.5 T,

    A. Halder, et al., “Optimization of Gradient-Echo Echo-Planar Imaging for T2* Contrast in the Brain at 0.5 T,”Sensors, vol. 23, no. 20, p. 8428, Oct. 2023

  7. [15]

    Low-field MRI: Clinical promise and challenges,

    T. C. Arnold, et al., “Low-field MRI: Clinical promise and challenges,” J. Magn. Reson. Imaging, vol. 57, no. 1, pp. 25–44, 2023

  8. [16]

    Statistical parametric maps in functional imaging: a general linear approach,

    K. J. Friston, et al., “Statistical parametric maps in functional imaging: a general linear approach,”Human Brain Mapping, vol. 2, no. 4, pp. 189–210, 1994

  9. [17]

    Advancements in Prostate Luminal Water Imaging: Radial Turbo Spin-Echo Acquisition and Spatial Regularization,

    P. Razmara, et al., “Advancements in Prostate Luminal Water Imaging: Radial Turbo Spin-Echo Acquisition and Spatial Regularization,”Proc. Intl. Soc. Magn. Reson. Med., 2024

  10. [18]

    fMRI based on transition-band balanced SSFP in comparison with EPI on a high-performance 0.55 T scanner,

    Y . Wang, et al., “fMRI based on transition-band balanced SSFP in comparison with EPI on a high-performance 0.55 T scanner,”Magn. Reson. Med., vol. 85, no. 6, pp. 3196–3210, 2021

  11. [19]

    A feasibility study of task-based fMRI at 0.55 T,

    P. Razmara, et al., “A feasibility study of task-based fMRI at 0.55 T,” Proc. Intl. Soc. Mag. Reson. Med., vol. 33, p. 0491, 2025

  12. [20]

    Twenty years of functional MRI: The science and the stories,

    P. A. Bandettini, “Twenty years of functional MRI: The science and the stories,”Neuroimage, vol. 62, no. 2, pp. 575–588, 2012

  13. [21]

    A method for removing imaging artifact from continuous EEG recorded during functional MRI,

    P. J. Allen, et al., “A method for removing imaging artifact from continuous EEG recorded during functional MRI,”Neuroimage, vol. 12, no. 2, pp. 230–239, 2000

  14. [22]

    Functional analysis of human MT and related visual cortical areas using magnetic resonance imaging,

    R. B. Tootell, et al., “Functional analysis of human MT and related visual cortical areas using magnetic resonance imaging,”J. Neurosci., vol. 15, no. 4, pp. 3215–3230, 1995

  15. [23]

    New clinical opportunities of low-field MRI: heart, lung, body, and musculoskeletal,

    Y . Tian, et al., “New clinical opportunities of low-field MRI: heart, lung, body, and musculoskeletal,”MAGMA, vol. 37, no. 1, pp. 1–14, 2024

  16. [24]

    BFP: BrainSuite fMRI pipeline,

    A. A. Joshi, et al., “BFP: BrainSuite fMRI pipeline,”OHBM, 2018

  17. [25]

    BrainSuite: an automated cortical surface identification tool,

    D. W. Shattuck and R. M. Leahy, “BrainSuite: an automated cortical surface identification tool,”Med. Image Anal., vol. 6, no. 2, pp. 129– 142, 2002

  18. [26]

    BrainVision Analyzer (Version 2.2.0) [Software], Brain Products GmbH, Gilching, Germany, 2019

  19. [27]

    Brainstorm: A user-friendly application for MEG/EEG analysis,

    F. Tadel, et al., “Brainstorm: A user-friendly application for MEG/EEG analysis,”Comput. Intell. Neurosci., 2011

  20. [28]

    Neurovascular regulation in the normal brain and in Alzheimer’s disease,

    C. Iadecola, “Neurovascular regulation in the normal brain and in Alzheimer’s disease,”Nature Reviews Neuroscience, vol. 5, no. 5, pp. 347-360, 2004

  21. [29]

    Neurovascular coupling in the normal brain and in hypertension, stroke, and Alzheimer disease,

    H. Girouard and C. Iadecola, “Neurovascular coupling in the normal brain and in hypertension, stroke, and Alzheimer disease,”Journal of Applied Physiology, vol. 100, no. 1, pp. 328-335, 2006

  22. [30]

    Beyond subjective measures: Systematic review of deep learning in chronic pain: Modalities, methods, and applications,

    M. A. Ranjbar, A. Ghaleh, H. B. Dogaheh, P. Razmara, and M. Baghani, “Beyond subjective measures: Systematic review of deep learning in chronic pain: Modalities, methods, and applications,” https://doi.org/10.21203/rs.3.rs-6217737/v1, 2025

Pith tools

Reviewed August 4, 2026 · model on record in the stance chip above.