{"id":"41840cf5-4c88-4218-af15-a71579ca13a1","arxiv_id":"2412.13354","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"For MDEIT, a small array of single-axis optically pumped magnetometers oriented normal to the scalp provides the best image quality, and larger vapour cells improve performance.","lead":"This modelling study tested how the number, size, and orientation of optically pumped magnetometers affect image quality in Magnetic Detection Electrical Impedance Tomography, a proposed method for non-invasively imaging fast brain activity. It finds that a small array of single-axis sensors measuring the field perpendicular to the scalp is best, and that larger vapour cells improve performance, guiding future sensor design.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Table 2's total noise omits the sqrt(bandwidth) factor used in Section 2.2, so the cell-volume reconstructions were run with underestimated noise; the size conclusion needs to be re-tested with corrected values.","rationale":"The reader's conditional verdict is supported. The most load-bearing concern is the numerical inconsistency in Table 2 because it directly underpins the paper's headline design recommendation about cell volume. I considered the physical assumption that density and bandwidth remain constant as cell size grows (Eq. 3); however, this is a standard fundamental-sensitivity scaling and the paper cites literature supporting larger cells, so it is not internally inconsistent. The Table 2 error, by contrast, is demonstrable from the paper's own Section 2.2 calculation and affects the exact noise values used in the simulations. Correcting it may or may not change the qualitative ordering, but the paper's most significant finding cannot be considered robust until this is checked. The sensing-axis and sensor-number results are less affected, since they use the consistent Section 2.2 noise, so the needed revision is targeted at the cell-volume section.","tokens_in":14303,"tokens_out":14951,"duration_ms":124673,"concrete_test":"Re-run the five cell-volume cases in Section 2.4 using the corrected total noise: N_tot = sqrt((S × sqrt(1000) / sqrt(232))^2 + (0.2 × sqrt(1000) / sqrt(232))^2) plus the current-source term, for each cell size, and re-compute the WSV for all four perturbations. Compare the corrected WSV trend against Figure 5B; if larger cells no longer yield significantly lower WSV for perturbations 1 and 2, the claim that vapour cell size need not be constrained to current commercial sizes should be withdrawn or weakened.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central claim that image quality increases with OPM vapour cell volume (Section 4.6) rests on the noise values in Table 2. Those values are internally inconsistent with Section 2.2: for a sensor with sensitivity S (fT/rtHz) and a measurement bandwidth of ±500 Hz with 232 averages, Section 2.2 converts S to total noise via S × sqrt(1000) / sqrt(232), e.g., 10 fT/rtHz × sqrt(1000) / sqrt(232) ≈ 20.75 fT. Table 2 instead gives S / sqrt(232), omitting the sqrt(1000) factor. For the 1 mm cell, 261 fT/rtHz becomes 17.1 fT instead of ≈542 fT; for the 18 mm cell, 3.42 fT/rtHz becomes 0.225 fT instead of ≈7.1 fT. The cell-volume simulations in Section 3.5 were run with these incorrect total-noise values. Because the error is a common multiplicative factor on the intrinsic-noise term, the relative ordering across cell sizes is preserved, so the qualitative trend may survive; however, the absolute SNR and the balance against the current-source noise term change, and the WSV improvements reported in Figure 5 could be different or even reversed. The conclusion therefore needs re-testing with the corrected noise before it can be accepted.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This computational modelling study optimises the sensing axis, sensor number, and vapour cell volume of optically pumped magnetometers (OPMs) for Magnetic Detection Electrical Impedance Tomography (MDEIT). Using an anatomically realistic finite-element head model, the authors simulate forward and inverse problems with four neural perturbations at different depths, add realistic noise, and evaluate reconstructed image quality with weighted spatial variance (WSV). They report that single-axis measurements normal to the scalp give the best image quality, that image quality improves with increasing sensor number but with diminishing returns, and that larger OPM vapour cell volumes improve image quality because of the assumed 1/sqrt(V) sensitivity scaling. The paper concludes that future OPMs for MDEIT should be single-axis, highly sensitive, and not necessarily constrained to the sizes of current commercial sensors.","tokens_in":14607,"tokens_out":6869,"duration_ms":58384,"significance":"The paper addresses a practically important hardware-design question for a novel neuroimaging technique, and the computational framework is careful in several respects: the forward model is realistic, the inverse crime is avoided by using different forward and reconstruction meshes, four perturbations at different depths are considered, and statistical testing is performed over 100 noise realisations. The sensing-axis and sensor-number results appear consistent with the described methods and provide useful design guidance. The cell-volume finding, if it survives re-analysis with corrected noise values, would be a significant driver for OPM development. However, the cell-volume result is a direct consequence of the assumed fundamental-sensitivity scaling rather than an independent empirical discovery; the main non-trivial aspect is the demonstration that spatial averaging and standoff effects do not negate the sensitivity gain, as evidenced by the constant Jacobian rank.","major_comments":[{"comment":"","section":"Section 2.4, Table 2"},{"comment":"","section":"Section 4.6 and Eq. (3)"}],"minor_comments":[{"comment":"","section":"References"},{"comment":"","section":"Throughout text"},{"comment":"","section":"Section 2.4"},{"comment":"","section":"Section 4.4"}],"recommendation":"major_revision","confidential_remarks":"The main issue is the inconsistency in the noise conversion for the cell-volume simulations, which directly affects the paper's most significant claim. The qualitative trend of improved image quality with larger cell volume may survive a corrected re-analysis because the error is a common multiplicative factor on the intrinsic-noise term and the current-source noise may dominate for the larger cells, but the quantitative results must be re-computed and the figures redrawn before the manuscript can be accepted. I would also encourage the authors to add a sensitivity analysis on the Eq. (3) scaling parameters, as the current treatment assumes an idealised constant-density, constant-bandwidth extrapolation that may not be achievable in practice."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely useful results are the sensing-axis and sensor-count findings. Single-axis normal-to-scalp measurement beats tangential and multi-axis configurations, and image quality plateaus as the array grows past roughly 64–128 sensors, with cortical targets needing fewer. These come from a sensible FEM pipeline—separate forward and reconstruction meshes to avoid inverse crime, realistic noise cases, and repeated-measures ANOVA with Tukey corrections. I take those conclusions as solid.\n\nThe cell-volume half is where I have problems. Table 2's total-noise column is internally inconsistent with Section 2.2. In Section 2.2 a 10 fT/rtHz sensor at ±500 Hz with 232 averages becomes 21.2 fT additive noise—that's 10×sqrt(1000)/sqrt(232). Table 2 instead computes S/sqrt(232), dropping the sqrt(1000) factor entirely. So the 1 mm cell at 261 fT/rtHz is listed at 17.1 fT instead of roughly 542 fT, and the 18 mm cell at 0.225 fT instead of roughly 7.1 fT. The cell-volume reconstructions were run with these too-low values, so the balance between intrinsic sensor noise and current-source noise is wrong. The relative ordering across cell sizes is preserved, so the qualitative trend may survive, but the absolute SNR and WSV numbers, and possibly the size of the benefit, are not trustworthy until the authors correct and re-run.\n\nThere's a second, more conceptual caveat: the 'bigger cell helps' result is largely entailed by the assumed 1/sqrt(V) sensitivity scaling in Eq. 3, not independently discovered. The simulations propagate that assumption. The single-injection-protocol limitation is acknowledged in 4.2, which is honest but narrows generality. And data/code are 'available upon request' only—for a purely computational paper, that's a bit weak.\n\nBottom line: this is a useful design study for MDEIT hardware, not a breakthrough. The axis and number results deserve to be used. The cell-volume conclusion needs a corrected Table 2 and re-run before it can support the design recommendation. I'd send it to peer review, but with a required revision, not as-is.","headline":"Useful parameter sweep for OPM-based MDEIT, but the cell-volume headline rests on a real arithmetic error in Table 2 and should be re-run.","tokens_in":15143,"tokens_out":4840,"would_cite":true,"duration_ms":42642,"reading_group":"maybe","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"For MDEIT brain imaging, the best magnetometer configuration is a single axis normal to the scalp, and larger sensors improve image quality.","keywords":["magnetic detection electrical impedance tomography","optically pumped magnetometer","image reconstruction","sensor optimisation","vapour cell volume","neural activity imaging","computational modelling","weighted spatial variance"],"falsifier":"Measure the intrinsic noise of OPMs with different vapour-cell volumes while holding atomic density and bandwidth constant; if the noise does not fall as $1/\\sqrt{V}$, the predicted image-quality gain from larger cells will not appear. Alternatively, build a phantom MDEIT system with two OPM sizes and check whether reconstructed image quality actually improves with volume as the simulations predict.","tokens_in":14110,"feed_emoji":"🧠","tokens_out":5129,"duration_ms":42961,"temperature":0.7,"pith_summary":"The paper sets out to determine the optimal arrangement of optically pumped magnetometers for Magnetic Detection Electrical Impedance Tomography, a proposed technique for non-invasively imaging fast neural activity. Using computational models of the human head, it compares sensing axes, array sizes, and vapour-cell volumes. It finds that measuring only the magnetic field component normal to the scalp gives the best reconstructed images, and that image quality increases as the magnetometer's vapour cell grows. This matters because it suggests future MDEIT-specific OPMs need not copy the small cell sizes of current commercial sensors, and can instead use larger, more sensitive cells to meet the demanding bandwidth and sensitivity requirements.","feed_headline":"One magnetometer axis and bigger cells win for MDEIT brain imaging","feed_subtitle":"Simulations show radial single-axis sensing and larger vapour cells improve reconstructed image quality.","key_machinery":"The argument runs on a simulation pipeline, but the load-bearing physical identity is the fundamental sensitivity limit of an OPM: $\\delta B_{\\mathrm{opt}} = \\frac{\\sqrt{2\\,BW}}{\\gamma \\sqrt{nV}}$, which says that for fixed atomic density $n$, bandwidth $BW$, and gyromagnetic ratio $\\gamma$, the sensor noise floor falls as the inverse square root of the vapour-cell volume $V$. The paper scales the intrinsic noise for five cell sizes using this formula, adds environmental and current-source noise, embeds the sensors in a finite-element head model, and reconstructs images. Image quality is scored with the weighted spatial variance, a metric that measures how closely the reconstructed conductivity blob matches the true perturbation's location and extent.","core_discovery":"On the paper's own terms, the central discovery is a design rule: for MDEIT, the optimal OPM configuration is a single sensing axis normal to the scalp, and larger vapour cells improve reconstructed image quality. This conclusion comes from finite-element simulations of a seven-tissue human head with four neural perturbations, image reconstruction using 0th-order Tikhonov regularisation with uncorrelated noise-based correction, and assessment using the weighted spatial variance. The authors report that single-axis normal measurements produced the largest signal-to-noise ratio and best image quality, that adding tangential axes introduced noisy measurements that degraded quality, and that increasing vapour-cell side length from 1 mm to 18 mm reduced sensor noise and improved image quality while the rank of the Jacobian stayed constant.","pith_inferences":["If the $1/\\sqrt{V}$ noise scaling holds experimentally, the cell-volume benefit could extend to other on-scalp magnetic sensing applications, not just MDEIT, wherever spatial averaging is acceptable.","The paper shows the Jacobian rank is unchanged by cell size, but real larger cells may also increase standoff distance and power draw; whether those practical penalties offset the sensitivity gain is a question for hardware prototypes.","One testable extension is to verify the exact noise values in Table 2 against direct measurements of OPMs with different cell volumes, since the table's numbers appear to derive from a single 3 mm reference cell and may not reflect practical operation at constant atomic density and bandwidth."],"forward_implications":["Future OPMs built for MDEIT should prioritise a single high-sensitivity axis normal to the scalp over multi-axis capability.","Arrays of roughly 48 to 96 single-axis sensors are practically suitable; going beyond about 128 gives little additional benefit, and around 32 may suffice for cortical activity.","Manufacturers can increase vapour-cell volume beyond current commercial sizes to gain sensitivity and bandwidth, potentially approaching SQUID-level performance without cryogenic cooling.","For a fixed total number of measurements, using more magnetometers is more beneficial than using multi-axis measurements on fewer magnetometers."],"supporting_citations":[{"why":"Establishes the MDEIT forward model, injection protocol, and noise cases that this study reuses and extends.","marker":"[5]"},{"why":"Supplies the fundamental OPM sensitivity formula and the operating parameters used to scale vapour-cell noise.","marker":"[22]"},{"why":"Provides the MEG-based precedent for optimising OPM sensing volume, which this work adapts to MDEIT.","marker":"[24]"},{"why":"Sets the baseline commercial OPM sensitivity and the context for what is practically achievable.","marker":"[25]"},{"why":"Supplies the environmental noise floor used in the total noise model.","marker":"[14]"},{"why":"Gives the uncorrelated noise-based correction algorithm used for all image reconstructions.","marker":"[20]"},{"why":"Defines the weighted spatial variance figure of merit used to compare reconstructed image quality.","marker":"[59]"}],"fun_headline_variants":["Single-axis OPMs optimize MDEIT brain imaging","Larger vapor cells improve MDEIT reconstruction","Radial single-axis sensing best for MDEIT","OPM design rule: one axis, bigger cells","MDEIT favored by single-axis, large OPMs"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper assumes that a larger OPM vapour cell can be operated with the same atomic density and bandwidth, so that its intrinsic noise drops as the inverse square root of the volume; if real larger cells cannot maintain those parameters, the predicted size benefit disappears.","fun_headline_variants_meta":{"raw":{"variants":["Single-axis OPMs optimize MDEIT brain imaging","Larger vapor cells improve MDEIT reconstruction","Radial single-axis sensing best for MDEIT","OPM design rule: one axis, bigger cells","MDEIT favored by single-axis, large OPMs"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.00024,"raw_usage":{"total_tokens":1472,"prompt_tokens":856,"completion_tokens":616,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":472,"completion_tokens_details":{"reasoning_tokens":538}},"tokens_in":472,"tokens_out":616,"duration_ms":5431,"temperature":1.0,"reasoning_tokens":538,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T13:13:52.791202+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure the intrinsic noise of OPMs with different vapour-cell volumes while holding atomic density and bandwidth constant; if the noise does not fall as $1/\\sqrt{V}$, the predicted image-quality gain from larger cells will not appear. Alternatively, build a phantom MDEIT system with two OPM sizes and check whether reconstructed image quality actually improves with volume as the simulations predict.","supporting_citations":[{"cited_title":"Non-invasive imaging of neural activity with magnetic detection electrical impedance tomography (MDEIT): a modelling study,","cited_arxiv_id":null,"evidence_quote":"Establishes the MDEIT forward model, injection protocol, and noise cases that this study reuses and extends."},{"cited_title":"Optically pumped magnetometers: From quantum origins to multi-channel magnetoencephalography,","cited_arxiv_id":null,"evidence_quote":"Supplies the fundamental OPM sensitivity formula and the operating parameters used to scale vapour-cell noise."},{"cited_title":"Optimising the sensing volume of OPM sensors for MEG source reconstruction,","cited_arxiv_id":null,"evidence_quote":"Provides the MEG-based precedent for optimising OPM sensing volume, which this work adapts to MDEIT."},{"cited_title":"QZFM, QuSpin,","cited_arxiv_id":null,"evidence_quote":"Sets the baseline commercial OPM sensitivity and the context for what is practically achievable."},{"cited_title":"An ultra-sensitive and wideband magnetometer based on a superconducting quantum interference device,","cited_arxiv_id":null,"evidence_quote":"Supplies the environmental noise floor used in the total noise model."},{"cited_title":"Noise-based correction for electrical impedance tomography,","cited_arxiv_id":null,"evidence_quote":"Gives the uncorrelated noise-based correction algorithm used for all image reconstructions."},{"cited_title":"An exhaustive criterion for estimating quality of images in electrical impedance tomography with application to clinical imaging,","cited_arxiv_id":null,"evidence_quote":"Defines the weighted spatial variance figure of merit used to compare reconstructed image quality."}],"review_version":1}