{"id":"84f5410c-4650-4f9a-ae66-7e88f3f345aa","arxiv_id":"2506.11634","paper_version":1,"verdict":"REJECT","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"high","formal_verification":"none","parameter_count":4,"one_line_summary":"In a small resting-state EEG-fNIRS sample, patients with major depressive disorder showed altered neurovascular coupling, with higher coupling strength in the maintenance phase and shorter recovery intervals in the acute phase compared with healthy controls.","lead":"This study used simultaneous EEG and fNIRS recordings to compare how brain electrical activity and blood oxygen responses are coupled in people with major depressive disorder at different stages. It reports that coupling is stronger in maintenance-phase patients and that a recovery timing metric is shorter in acute patients, suggesting a possible wearable biomarker for monitoring depression.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The MG-vs-HC difference in mNVC_R may be driven by uncorrected 0.1 Hz Mayer-wave oscillations in HbT rather than by neurovascular coupling; Section 4.1 acknowledges these oscillations but does not control them.","rationale":"The reader's stated weakest assumption is that mNVC_R isolates neurovascular coupling, with Mayer waves noted but not regressed. My independent reading of Sections 2.5, 2.7, 3.1, and 4.1 confirms this as the single most load-bearing vulnerability. The paper's own display of ~0.1 Hz oscillations in the averaged NVC_R curves is evidence that the slow oscillation is present in the very signal used for group comparisons, yet no control comparison, surrogate analysis, or systemic-signal regression is offered. The central discovery claim—that maintenance-phase patients have stronger resting NVC—would be materially weakened if the effect survived only because of phase-locked or amplitude-modulated slow hemodynamic fluctuations. The concrete test I propose is a minimal check that any reviewer could require: low-frequency regression plus a time-shift null on mNVC_R. Because the reader already assigned REJECT and my concern directly supports that assessment rather than changing it, I recommend UNCHANGED. I agree with the reader's framing and do not see a more decisive or more specific attack elsewhere in the paper; the data-sharing limitations and small sample size are real but secondary to the measurement-validity issue.","tokens_in":12612,"tokens_out":2761,"duration_ms":29706,"concrete_test":"Recompute the primary group statistics after regressing the 0.05–0.15 Hz component out of each HbT channel, for example by using short-separation-channel regression or by fitting and subtracting a common low-frequency component derived from the fNIRS array. In parallel, build a null distribution for mNVC_R by time-shifting each GFP peak by a random 10–20 s offset relative to the HbT signal, re-extracting the maximum Spearman correlation in the 2–8 s window, and repeating this over many shuffles. If the MG-vs-HC eyes-open difference in mNVC_R (Section 3.1) is no longer significant after low-frequency removal, or if group differences comparable to the observed ones appear in the time-shifted null data, the central claim is not supported.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The central finding—maintenance-phase patients show higher mNVC_R (Section 3.1)—depends on mNVC_R being a valid, selective measure of neurovascular coupling. Section 2.7 defines mNVC_R as the maximum Spearman correlation between EEG-derived activation and HbT in a hand-set 2–8 s window after each GFP peak. The fNIRS preprocessing band-pass filter in Section 2.5 is 0.05–0.7 Hz, which explicitly retains ~0.1 Hz oscillations. Section 4.1 and Figure 6 acknowledge 'a slow variation of about 0.1 Hz. This is consistent with the Mayer Wave,' but no regression, short-separation correction, or global-signal removal is reported. Because mNVC_R is a maximum over a 6-s window, it is a selection statistic: high values can arise from chance phase alignment between GFP peaks and ongoing slow HbT fluctuations, especially with small group sizes. The group differences could therefore reflect group differences in systemic vascular oscillations or noise levels rather than coupling strength. The arbitrary mNVC_R > 0.1 inclusion threshold (Section 2.7) and the multiple testing across eyes-open/eyes-closed and several derived metrics do not remove this confound. The recovery-timing claim built on mNVC-T (Section 4.3) inherits the same problem because mNVC-T is constructed from two selected extrema, mID_RT and mNVC_RT. Thus the load-bearing assumption—that mNVC_R isolates neurovascular coupling from systemic hemodynamic oscillations—is not secured by the reported analyses.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a simultaneous resting-state EEG-fNIRS study of neurovascular coupling (NVC) in major depressive disorder. The authors define a metric NVC_R as the Spearman correlation between EEG-derived activation (global field power peaks) and total hemoglobin (HbT) in an 8-second window following each peak, and extract the maximum correlation mNVC_R in a 2–8 s window as the NVC strength. They report that maintenance-phase MDD patients have significantly higher mNVC_R than healthy controls, that acute-phase patients show a shorter interval (mNVC-T) from the initial dip to the NVC peak, and that this interval increases with recovery. They also report correlations between mNVC-T and clinical scales, proposing it as a potential biomarker.","tokens_in":12939,"tokens_out":6258,"duration_ms":53266,"significance":"If the findings were robust, this would be a novel and clinically useful multimodal biomarker for MDD phase and recovery, and the portable EEG-fNIRS system is a practical contribution. However, the study is small (final n=28), the primary endpoint is a hand-crafted selection statistic, and the analysis does not control for systemic physiological oscillations. The strengths of the paper are the hardware design and the ecological resting-state paradigm, but the statistical and methodological issues currently preclude drawing reliable conclusions about NVC differences.","major_comments":[{"comment":"The fNIRS band-pass filter (0.05–0.7 Hz) preserves ~0.1 Hz oscillations that the authors identify as Mayer waves in Section 4.1 and Figure 6. Because mNVC_R (Section 2.7) is the maximum Spearman correlation between EEG-derived activation and HbT in a 2–8 s window after each GFP peak, it is a selection statistic over a ~6 s window; chance alignment of GFP peaks with ongoing 0.1 Hz HbT fluctuations can produce large correlations. Without regressing out systemic oscillations or using short-separation channels, the group difference in mNVC_R (Section 3.1) may reflect differences in systemic vascular physiology rather than neurovascular coupling. This confound directly undermines the central claim.","section":"Section 2.5 and 4.1"},{"comment":"The inclusion criterion for the initial dip analysis (mNVC_R > 0.1) is arbitrary and applied at the individual level. The manuscript does not report how many GFP peaks or epochs were excluded in each group under this threshold. If exclusion rates differ across groups, the subsequent comparisons of mID_R, mID_RT, and mNVC-T (Sections 3.2–3.3) are biased; the reported group differences may be an artifact of differential selection rather than a physiological effect.","section":"Section 2.7"},{"comment":"The study tests a large number of derived metrics (mNVC_R, mNVC_RT, mID_R, mID_RT, mNVC-T, ΔmNVC-R) across eyes-open and eyes-closed conditions, yet reports uncorrected p-values and several 'marginally significant' results (0.05<p<0.1) as evidence. Table 2 claims Bonferroni correction, but the text reports raw p-values; no corrected p-values are given. With a healthy-control group of n=6, the ANOVA F(2,25)=3.5775 and the post hoc MG vs HC comparison (p<0.01) are fragile and not robust to outliers or non-normality. The abstract's 'significantly higher' claim is not supported after accounting for multiple comparisons.","section":"Section 3.1 and Table 2"},{"comment":"Of 55 recruited participants, only 28 were analyzed (~49% exclusion), predominantly due to fNIRS signal quality. The final groups are severely unbalanced (12, 10, and 6). The exclusion process is likely non-random with respect to group (e.g., hair characteristics, age, medication), but no comparison of included vs excluded participants is provided. This selection bias, combined with the small healthy-control group, makes the reported group differences unreliable.","section":"Section 3 and 2.2"},{"comment":"The correlation coefficients between mNVC-T_EC and clinical scales (HAMD -0.4084, HAMA -0.4168, BDI -0.3266, SAS -0.2745) are reported without p-values, confidence intervals, or correction for multiple testing. These correlations underpin the biomarker claim, but with n=28 across three groups they are not statistically established. Moreover, mNVC-T is derived from the same selection-based measures criticized above, so the correlations inherit their confounds.","section":"Section 4.3"}],"minor_comments":[{"comment":"The phrase 'at least 1 months of training' should be 'at least 1 month of training'.","section":"Section 2.1"},{"comment":"The p-value range '0.5<p<0.1' appears to be a typo for '0.05<p<0.1'; the subsequent '0.1<p<0.15' is an unusual notation and should be clarified.","section":"Section 3.3"},{"comment":"The sentence 'our study did observe such differences in EEG amplitude in this group' is confusing; given the context, it likely should read 'did not observe'.","section":"Section 4.1"},{"comment":"The caption uses 'eye-close' where 'eyes-closed' would be more consistent with the text.","section":"Figure 6"}],"recommendation":"reject","confidential_remarks":"The manuscript describes an interesting hardware prototype and a novel resting-state paradigm, but the analysis is not yet at the standard required for a serious journal. The main concerns are the unaddressed Mayer-wave confound, the arbitrary thresholding, and the unreliable statistics with a final n=28 and unbalanced groups. The authors over-interpret marginally significant results. Substantial new analyses and likely new data are needed before the claims can be supported."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"You asked for my read on arXiv:2506.11634. Short version: the question is good, the novelty is real, but the central claim doesn't hold up as presented.\n\nWhat's new: this is the first resting-state EEG-fNIRS coupling analysis anchored on GFP peaks in MDD, and it introduces mNVC-T as a recovery-timing construct. The portable hardware is practical, and the idea of using spontaneous neural events instead of oddball cues is sensible for this patient group. I also credit the authors for being open about the ~50% data loss and for explicitly noting where their own results did not reach significance.\n\nThe soft spots are serious, though. The main result—maintenance-phase patients having higher mNVC_R—depends entirely on a metric that is not adequately described. Section 2.7 says Spearman correlation between EEG activation values and HbT over an 8-second window, but it is not clear whether that correlation is across channels at each time point, across time within the window, or something else. That ambiguity alone is enough to make interpretation difficult. And because mNVC_R is the maximum of this correlation over a 2–8 s window, it is a selection statistic: chance alignments with ongoing slow oscillations can inflate it. The fNIRS filter keeps 0.1 Hz activity, and the authors themselves note the Mayer wave in Section 4.1 but never regress it out. With only 12, 10, and 6 subjects after exclusion, and multiple comparisons across eyes-open/eyes-closed and several derived measures, the marginal p-values just don't support the abstract's confident summary.\n\nThat said, I don't think this is a sloppy paper. The logic is clear, the limitations are acknowledged, and the methodological weaknesses are fixable in a future study. But the current evidence is not enough to claim a phase-dependent NVC biomarker.\n\nFor you: if you work on multimodal biomarkers or NVC methodology, it's worth a skim to see the pitfalls. I wouldn't cite it yet. I would send it to peer review—the idea deserves referee time—but with the expectation that the central metric needs rigorous re-analysis and, ideally, code/data release before any of its conclusions could be accepted.","headline":"A worthwhile pilot question with a real first-of-kind method, but the central mNVC_R finding is not yet trustworthy because the metric is under-specified and the 0.1 Hz Mayer wave is left in the data.","tokens_in":13485,"tokens_out":3351,"would_cite":false,"duration_ms":34213,"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":"In resting-state EEG-fNIRS, maintenance-phase MDD patients show stronger neurovascular coupling than healthy controls, while acute patients show an abbreviated dip-to-peak interval that lengthens with recovery.","keywords":["neurovascular coupling","major depressive disorder","resting-state EEG","functional near-infrared spectroscopy","global field power","initial dip","prefrontal cortex","recovery biomarker"],"falsifier":"Recompute the group comparisons after removing the ~0.1 Hz Mayer-wave component from each hemoglobin channel, or after repeating the coupling calculation using time-shifted EEG peaks as controls; if the maintenance-versus-healthy difference in coupling strength or the acute-versus-healthy difference in dip-to-peak timing vanishes, the reported effect is a slow-oscillation artifact rather than neurovascular coupling.","tokens_in":12387,"feed_emoji":"🧠","tokens_out":8006,"duration_ms":79292,"temperature":0.7,"pith_summary":"Neurovascular coupling is the chain by which local neural firing raises regional blood flow to meet energy demand. This paper asks whether that chain is altered in major depressive disorder and whether it normalizes during recovery, using a wearable cap that records EEG and fNIRS simultaneously during rest. The authors find that coupling strength is highest in maintenance-phase patients, significantly above healthy controls in the eyes-open condition, and that the time from the initial oxygen-dip to the peak of coupling is compressed in acute depression and gradually lengthens toward healthy values as patients recover. If correct, these resting-state measures offer an objective, low-cost way to monitor depression stage and recovery without task compliance.","feed_headline":"Acute depression compresses a brain blood-flow coupling interval","feed_subtitle":"In resting EEG-fNIRS, the dip-to-peak interval lengthens from acute MDD to maintenance to healthy controls.","key_machinery":"The machinery is a resting-state neurovascular coupling assay built from simultaneous EEG-fNIRS. Global field power (GFP), the spatial spread of EEG voltage, is computed only over the prefrontal and temporal regions covered by fNIRS, and the top five GFP peaks in each resting epoch are treated as spontaneous neural events. For each peak, the Spearman correlation between interpolated EEG activation and total hemoglobin (HbT) over the following eight seconds defines the coupling response; the maximum within a hand-set 2–8 second window is mNVC_R, and the preceding negative excursion is called the initial dip (mID_R). The time from the initial dip to the coupling peak, mNVC-T, is the study's central timing measure. This design replaces stimulus-locked averaging with endogenous neural peaks, aiming to avoid habituation and attentional confounds that plague oddball paradigms in depressed patients.","core_discovery":"The central claim is that neurovascular coupling consistency is not uniformly reduced in depression but is phase-dependent: maintenance-phase patients show the strongest EEG-to-hemoglobin correlation, while acute-phase patients show an abnormally short interval from the initial dip of oxygen consumption to the neurovascular coupling peak. Concretely, maintenance patients had significantly higher maximum coupling response coefficients than healthy controls in eyes-open rest (0.2678 versus 0.1566, p<0.05), acute and maintenance patients showed deeper initial dips than healthy controls in eyes-closed rest, and the eyes-closed dip-to-peak interval was shortest in acute patients, intermediate in maintenance patients, and longest in healthy controls. This interval correlated negatively with depression and anxiety scale scores, and the authors interpret the pattern as an over-responsive neurovascular feedback loop in acute depression that gradually normalizes, with longer-term illness possibly leading to vascular aging or reduced hemoglobin-carrying capacity.","pith_inferences":["Beyond the paper, a within-subject longitudinal design that follows the same acute patients into maintenance would separate the reported timing recovery from chronic medication effects, since SSRIs themselves can alter endothelial function; the present cross-sectional comparison cannot make that separation.","Beyond the paper, a sensitivity analysis varying the 2–8 second window and the 0.1 correlation threshold would test whether the group differences are stable or artifacts of those choices; the paper does not report such a sweep.","Beyond the paper, if the shortened dip-to-peak interval really reflects an over-responsive neurovascular loop, a breath-hold or carbon-dioxide vascular reactivity test combined with the same EEG-fNIRS setup should show correspondingly faster or larger vascular responses in acute patients."],"forward_implications":["If the timing metric mNVC-T reflects the state of the neurovascular feedback loop, acute depression can be characterized as an over-responsive, prematurely peaking vascular response, with remission returning the interval toward healthy values.","The eyes-open specificity of the coupling-strength difference suggests that externally oriented attention states reveal neurovascular alterations that eyes-closed rest masks, so recording both conditions is necessary for monitoring.","Because mNVC-T correlates negatively with HAMD, HAMA, BDI, and SAS scores, combining this timing metric with clinical scales could improve tracking of recovery progress in outpatient settings.","A wearable, reusable EEG-fNIRS system capable of these measurements could carry neurovascular coupling assessment outside the scanner, making repeated longitudinal monitoring feasible."],"supporting_citations":[{"why":"Defines neurovascular coupling and the physiological cascade from neural activity to cerebral blood flow, the framework the study tests.","marker":"[11]"},{"why":"Reports the 4–6 s hemodynamic lag for event-related responses, which motivates the 2–8 s window used for the maximum coupling coefficient.","marker":"[12]"},{"why":"Documents hair color, cleanliness, light, and motion effects on fNIRS signal quality, used to explain the high participant exclusion rate.","marker":"[17]"},{"why":"Provides the temporal derivative distribution repair (TDDR) motion-correction method applied to fNIRS before hemoglobin conversion.","marker":"[18]"},{"why":"Introduces the 'initial dip' concept that the study formalizes as mID_R and uses for timing comparisons.","marker":"[21]"},{"why":"Characterizes ~0.1 Hz Mayer-wave oscillations in fNIRS, the slow systemic signal the authors note in their waveforms.","marker":"[22]"},{"why":"Reports that SSRIs improve endothelial function, an alternative explanation for altered coupling in maintenance-phase patients on medication.","marker":"[24]"},{"why":"Links depression to vascular aging, supporting the paper's interpretation of reduced vascular elasticity and hemoglobin replenishment.","marker":"[25]"}],"fun_headline_variants":["Depression's acute phase shortens brain blood-flow coupling dip-to-peak time","Maintenance-phase MDD shows strongest EEG-fNIRS coupling, not acute","EEG-fNIRS reveals phase-dependent neurovascular coupling in depression","Acute MDD compresses coupling interval; recovery lengthens it","Resting-state EEG-fNIRS tracks depression phase via neurovascular coupling"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The analysis assumes that a spontaneous burst of coordinated brain electrical activity behaves like an evoked neural event, so that the strongest brain-signal-to-blood-flow correlation in a fixed two-to-eight-second window isolates neurovascular coupling rather than reflecting slow blood-pressure oscillations or the arbitrary choice of a 0.1 correlation cutoff.","fun_headline_variants_meta":{"raw":{"variants":["Depression's acute phase shortens brain blood-flow coupling dip-to-peak time","Maintenance-phase MDD shows strongest EEG-fNIRS coupling, not acute","EEG-fNIRS reveals phase-dependent neurovascular coupling in depression","Acute MDD compresses coupling interval; recovery lengthens it","Resting-state EEG-fNIRS tracks depression phase via neurovascular coupling"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000212,"raw_usage":{"total_tokens":1437,"prompt_tokens":984,"completion_tokens":453,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":600,"completion_tokens_details":{"reasoning_tokens":356}},"tokens_in":600,"tokens_out":453,"duration_ms":4933,"temperature":1.0,"reasoning_tokens":356,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T04:04:09.010225+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Recompute the group comparisons after removing the ~0.1 Hz Mayer-wave component from each hemoglobin channel, or after repeating the coupling calculation using time-shifted EEG peaks as controls; if the maintenance-versus-healthy difference in coupling strength or the acute-versus-healthy difference in dip-to-peak timing vanishes, the reported effect is a slow-oscillation artifact rather than neurovascular coupling.","supporting_citations":[{"cited_title":"B., & Davidson, R","cited_arxiv_id":null,"evidence_quote":"Defines neurovascular coupling and the physiological cascade from neural activity to cerebral blood flow, the framework the study tests."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports the 4–6 s hemodynamic lag for event-related responses, which motivates the 2–8 s window used for the maximum coupling coefficient."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Documents hair color, cleanliness, light, and motion effects on fNIRS signal quality, used to explain the high participant exclusion rate."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Provides the temporal derivative distribution repair (TDDR) motion-correction method applied to fNIRS before hemoglobin conversion."},{"cited_title":"P ., Horovitz, S","cited_arxiv_id":null,"evidence_quote":"Introduces the 'initial dip' concept that the study formalizes as mID_R and uses for timing comparisons."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Characterizes ~0.1 Hz Mayer-wave oscillations in fNIRS, the slow systemic signal the authors note in their waveforms."},{"cited_title":null,"cited_arxiv_id":null,"evidence_quote":"Reports that SSRIs improve endothelial function, an alternative explanation for altered coupling in maintenance-phase patients on medication."}],"review_version":1}