{"id":"a040273b-7a8f-4feb-a948-f242519215fc","arxiv_id":"2411.16360","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":5.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":3,"one_line_summary":"DCRs and ACEPs differ in the timing of the N1 onset (about 2 ms later for ACEP) and in the shape of responses after 40 ms, supporting distinct electrogenic mechanisms.","lead":"This paper compared two types of brain signals recorded during awake brain surgery: direct cortical responses and axono-cortical evoked potentials. It found small but measurable delays and different late signals that could help surgeons distinguish which pathway is being stimulated.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"The ACEP delay claim rests on t_zc1, a morphology-sensitive zero-crossing with p=0.0556 (two-tailed); no P0 onset latency is reported to show a true conduction delay.","rationale":"The reader's weakest assumption concerned the Euclidean distance proxy used to estimate conduction velocity, but the more load-bearing issue is the delay metric itself. t_zc1 is the only quantitative support for the 'delay at the onset of early components' leg of the central claim, yet it is a composite waveform landmark that can shift with P0/N1 morphology even in the absence of a true latency shift. The statistical reporting is also ambiguous: the table reports p=0.0556, while the text claims significance via a one-tailed test, and the table does not state the test direction. Because the central claim explicitly depends on two distinctions (delay plus late-shape difference), failure of the delay leg substantially weakens the headline conclusion even if the area-under-the-curve difference remains. The paper has genuine strengths (high sampling rate, artifact handling, same-site comparisons in three patients), so the appropriate outcome is a conditional acceptance requiring a re-analysis of the early waveform rather than outright rejection.","tokens_in":75,"tokens_out":6128,"duration_ms":120610,"concrete_test":"On the already-averaged per-patient DCR/ACEP traces, measure the P0 peak latency (or 10% rise point of P0) for each response and compare DCR vs. ACEP with a paired test. If P0 peak latencies are not significantly delayed in ACEP, the t_zc1 difference is driven by later waveform morphology rather than by axonal conduction delay. Also recompute the t_zc1 comparison with both one- and two-tailed paired t-tests and state clearly which p-value is reported in Table 1.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Table 1 reports t_zc1 (DCR 5.49±1.20 ms, ACEP 7.29±2.24 ms, p=0.0556, t=-2.29), yet §3.2 calls this a 'significant delay' based on a one-tailed t-test. The reported p=0.0556 is the two-tailed value; a one-tailed test would give p≈0.028. The table does not state the one-tailed decision, and the abstract claims a 'significant delay' as a main result. More fundamentally, t_zc1 is the P0–N1 zero-crossing, not an onset latency of the early component. Its timing is affected by P0 amplitude/width and N1 rise time, so it is not a pure measure of axonal conduction delay. The paper never reports P0 peak/onset latencies to show a uniform waveform shift. The conduction-velocity estimate (5.9±43.6 m/s) is accordingly unstable, with individual delays ranging from −1.71 to +5.15 ms. If the t_zc1 difference is not a bona fide latency shift, the first leg of the central claim ('delay at the onset of early components') is unsupported, leaving only the late-component shape difference, which was found by testing many metrics and reporting only the significant ones.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"The manuscript compares direct cortical responses (DCR) and axono-cortical evoked potentials (ACEP) recorded with ECoG in nine patients during awake brain surgery. It reports that ACEP exhibits a delayed first zero-crossing of the early P0/N1 complex (about 1.8–2.4 ms), a shorter N1 duration, a positive after-positivity between 40 and 100 ms versus a negative one for DCR, and higher gamma power for DCR at 20–40 ms. The authors interpret these differences as evidence for axonal conduction delay in ACEP and additional intracortical activation in DCR, and they argue that DCR and ACEP can be distinguished by a delay at the onset of early components and by the shape of later components.","tokens_in":14187,"tokens_out":4362,"duration_ms":38284,"significance":"If the reported distinctions replicate, they offer a practical way to infer the effects of direct electrical stimulation during brain tumor surgery, potentially improving intraoperative mapping. The study draws on direct human recordings at 19.2 kHz, within-patient paired comparisons, and shared recording sites for a subset of DCR and ACEP pairs, which are commendable strengths. However, the early-delay claim rests on a borderline zero-crossing metric rather than an onset latency, and the conduction-velocity estimate is extremely noisy; the late-component findings are also weakened by selective reporting. The paper explicitly acknowledges several of these limitations, which is to its credit, but the central claims need stronger statistical and interpretational support.","major_comments":[{"comment":"The primary delay metric t_zc1 is called 'significant' based on a one-tailed t-test, yet Table 1 reports only p=0.0556 without labeling it two-tailed or providing the one-tailed p-value. Because the central claim of a delay in ACEP hinges on this metric, the authors should report the one-tailed p-value, state the directional hypothesis as pre-specified, and provide P0 peak or onset latencies to demonstrate that the difference is a true latency shift rather than a morphology-induced change in the zero-crossing.","section":"Table 1 and §3.2"},{"comment":"The conduction-velocity estimate (mean 5.9 ± 43.6 m/s) is not meaningful as reported: the standard deviation is an order of magnitude larger than the mean, and the individual t_zc1 differences range from -1.71 to +5.15 ms. The paper acknowledges imprecise distance estimation, but the axonal-delay interpretation requires a positive correlation between distance and delay. Without such an analysis, or P0-based latencies, this estimate cannot support the conduction-delay explanation.","section":"§3.2"},{"comment":"The claim that DCR relaxation is 'monotonous and positive' is contradicted by the data: Min(dDCR/dt)[50:80]ms has a mean of -0.61 ± 1.42 with p=0.8819 for the test against >0, indicating no evidence of a positive monotonic slope. This inconsistency undermines the interpretation of the Area[40:100]ms difference and should be corrected or re-analyzed.","section":"§3.4 and Table 1"},{"comment":"The manuscript states that 'only those showing significant differences were presented here' after measuring many waveform parameters. This selective reporting, without any multiple-comparison correction, inflates the Type I error for the late-component findings (Area[40:100]ms, WHQN1, and gamma power). The authors should report all tested metrics or apply an appropriate correction.","section":"§2.4 and §3.1"}],"minor_comments":[{"comment":"The p-value for t_zc1 should be labeled with the test type and direction; the current presentation is ambiguous because a two-tailed p of 0.0556 is not significant at the conventional 0.05 level, while a one-tailed p would be.","section":"Table 1"},{"comment":"The phrase 'delay at the onset of early components' is imprecise; t_zc1 is the first zero-crossing between P0 and N1, not an onset latency. The wording should be adjusted to match the actual metric.","section":"Abstract and §3.2"},{"comment":"For Patient 8, the stimulation duration '[7.3:5.3]s' appears to be a typo, as the start exceeds the end; please correct or clarify.","section":"§2.3"},{"comment":"The gamma-band comparison is performed on five time windows but reported as a single pooled comparison; the number of comparisons and whether any correction was applied should be stated.","section":"§3.5"},{"comment":"The term 'phase delay' could be confused with a phase shift in steady-state signals; a more neutral term such as 'latency difference' would be clearer.","section":"Figure 3B inset"}],"recommendation":"major_revision","confidential_remarks":"The manuscript fits the journal's scope and the empirical data are valuable, but the early-delay claim needs strengthening: either report P0 latencies or clearly restrict the claim to the zero-crossing metric. The selective reporting and the internal inconsistency about the DCR slope are also fixable with additional analyses and re- wording. I see no reason to doubt the authors' good faith; the issues are statistical and interpretational rather than circularity."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Quick take: this is a direct extension of the group's own DCR/ACEP framework, and the most defensible new result is not the headline delay but the late-shape difference. The delay claim leans on a single zero-crossing metric that is only significant if you accept a one-tailed test, and the conduction-velocity estimate is basically noise. Still, the paper deserves refereeing; the clinical discriminator is plausible and the recording methodology is unusually careful.\n\nWhat's actually new: they systematically compare DCR and ACEP in the same patients (9 patients, 55 DCR vs 29 ACEP after amplitude threshold) and report t_zc1, N1 width, area 40-100 ms, and 50 Hz gamma. The late components are consistently different: ACEP shows an after-positivity and shorter N1, DCR stays negative and has a wider N1. The p-values for those are reasonable (0.002-0.04), and the shared-site comparisons in patients 1/5/7 support the direction. Credit also for 19.2 kHz sampling, no hardware filter, polarity alternation, and explicit artifact handling—that is the right way to record early components.\n\nSoft spots. The headline t_zc1 result is reported as p=0.0556 in Table 1 and then called significant in the text via a one-tailed test that isn't presented as such. That's not fatal—the hypothesis was directional and a one-tailed p~0.028 is defensible—but the paper should say so plainly. More importantly, t_zc1 is a zero-crossing between P0 and N1, not an onset latency. It shifts with P0 width and N1 rise time, so it is not a clean measure of axonal conduction delay. The paper never reports P0 peak/onset latencies, and the velocity estimate (5.9 ± 43.6 m/s) shows how unstable the distance-delay relation is. The 3D Euclidean distance is a weak proxy for the actual fiber path, and some ACEPs may be direct cortical stimulation from below, as the authors admit. Also, only significant metrics are reported; several time windows look post hoc. That tempers the gamma finding too (only 50 Hz, 20-40 ms). No data or code are provided, which is common for clinical intracranial recordings but still limits independent checking.\n\nBottom line: the late-component distinction is real enough to be clinically useful, but the conduction-delay interpretation needs better latency measures or a larger sample. For a reader working on intraoperative ECoG/DES, this is worth a careful look. I'd send it to peer review with a request for P0 latencies, one-tailed p-values stated as such, and the full metric set.","headline":"Useful clinical discriminator between DCR and ACEP, but the headline delay is a one-tailed zero-crossing effect with p=0.0556 and a noisy velocity estimate; the late-component shape difference is the more defensible finding.","tokens_in":14759,"tokens_out":2774,"would_cite":true,"duration_ms":25944,"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":"ACEP and DCR differ by a ~2 ms N1 delay and opposite late activity.","keywords":["Direct Electrical Stimulation","Evoked Potentials","Electrocorticography","Direct Cortical Response","Axono-Cortical Evoked Potential","Electrogenesis","Awake Brain Surgery","Intra-Operative Neural Monitoring"],"falsifier":"Measure ACEP and DCR from the same recording electrode while systematically moving the white-matter stimulation site closer to and farther from the cortex, and check whether the t_zc1 delay grows with distance by the expected amount. If the delay stays roughly constant, or if a short-distance ACEP shows no delay at all, the axonal-conduction interpretation is refuted.","tokens_in":56,"feed_emoji":"🧠","tokens_out":10446,"duration_ms":85864,"temperature":0.7,"pith_summary":"Direct electrical stimulation of the brain can elicit two kinds of evoked potentials: direct cortical responses (DCR), recorded at the same gyrus where the cortex is stimulated, and axono-cortical evoked potentials (ACEP), recorded after stimulating white matter fibers. The paper claims these two responses can be told apart by a short delay in the onset of the N1 component (about 1.8 to 2.4 ms, with ACEP later) and by the shape of the waveform 40 to 100 ms after the stimulus artifact, where ACEP turns positive and DCR remains negative. It interprets the early delay as axonal conduction time from the white-matter stimulation site and the late difference as extra intra-cortical activity recruited only by direct cortical stimulation. If correct, these markers give a practical electrophysiological signature for judging, in real time during tumor surgery, whether the stimulation is engaging cortex directly or through subcortical fibers, which matters for interpreting mapping results and for modeling the effects of DES.","feed_headline":"A 2 ms delay separates two brain-stimulation responses","feed_subtitle":"White-matter stimulation shows a delayed N1 and a positive late wave, while direct cortical stimulation leaves a negative one.","key_machinery":"The key machinery is the set of waveform metrics extracted from averaged electrocorticographic traces, chiefly the time of the first zero-crossing t_zc1 (the transition between the early positive P0 and the negative N1 component) and the signed area under the curve from 40 to 100 ms after the stimulation artifact. t_zc1 serves as a read-out of the moment the cortical population response begins, so its delay between ACEP and DCR provides an estimate of axonal conduction velocity; the late signed area serves as a read-out of whether a secondary cortical activity follows the main N1 deflection, which the paper associates with intra-cortical axon recruitment in DCR.","core_discovery":"The paper establishes that, even though DCR and ACEP share the same canonical P0–N1 waveform, they differ in two measurable respects: the first zero-crossing t_zc1, which marks the onset of N1, occurs significantly later for ACEP than for DCR, and the signed area under the curve between 40 and 100 ms after the stimulus artifact is positive for ACEP but negative for DCR. The observed t_zc1 delay averages about 1.8 ms (median 2.4 ms), and the N1 component is wider for DCR, whose relaxation after the N1 peak is monotonic, while ACEP shows an after-positivity. The investigators interpret the early delay as axonal conduction time from the deeper white-matter stimulation site to the cortical recording electrode, estimating a mean conduction velocity of about 5.9 m/s (median 8.38 m/s), and they interpret the late-wave divergence as evidence that direct cortical stimulation recruits additional small, slow intra-cortical axons that are not engaged when the stimulus reaches the cortex through white-matter pathways.","pith_inferences":["A natural extension is to test whether the t_zc1 delay grows linearly with stimulation-to-recording distance; if it does not, the axonal-conduction reading would need to be revised toward a local-network or synaptic-delay explanation.","Because the paper's velocity estimate uses straight-line distance rather than tract length, the true axonal conduction speed is likely lower than 5.9 m/s; tractography-based path length could correct this estimate.","The higher 50 Hz gamma power for DCR during the N1 period suggests a testable prediction that direct cortical stimulation produces stronger high-frequency cortical activation than white-matter stimulation, which could be probed with EEG or fMRI measurements.","If the ACEP after-positivity is inhibitory in origin, as the authors provisionally suggest, it could serve as a marker for the spatial extent of inhibition around a white-matter tract during awake surgery, testable with paired-pulse or pharmacological manipulations."],"forward_implications":["Within a single patient, an N1 onset delay near 2 ms can classify an evoked response as an ACEP rather than a DCR, turning a simple latency measurement into a connectivity check during resection.","The mean conduction velocity of about 5.9 m/s (median 8.38 m/s) is consistent with the longer latency reported for the arcuate fasciculus, so the same delay logic can be extended to monitor white-matter pathways at larger distances.","Because DCR's N1 is wider and followed by negative activity while ACEP shows a positive after-positivity, comparing these shape features can indicate whether a recording electrode is receiving input from subcortical fibers or is on a directly stimulated gyrus.","The results imply that computational models of direct electrical stimulation should treat DCR and ACEP as distinct electrogenic processes, with an explicit axonal conduction delay for white-matter stimulation and an additional intra-cortical recruitment term for cortical stimulation.","Resolving the ~2 ms delay requires high sampling rates and no hardware filter, so lower-sampling clinical acquisition systems may systematically miss the ACEP-DCR distinction."],"supporting_citations":[{"why":"Supplies the distribution of human white-matter axon diameters and the velocity-diameter relation used to predict a conduction delay near the observed 1.8–2.4 ms.","marker":"Liewald et al., 2014"},{"why":"Provides the long-range ACEP conduction latency (12.8 ms over ~10 cm) used to validate the study's estimated conduction velocity.","marker":"Yamao et al. (2014)"},{"why":"Defines the P0, N1 and after-positivity components and the canonical waveform of short-range ACEPs that this study compares with DCR.","marker":"Rossel et al., 2023"},{"why":"Establishes the intraoperative protocol for recording DCR and ACEP in the same patients, enabling the direct within-subject comparison made here.","marker":"Boyer et al., 2021a"},{"why":"Supports the claim that electrical stimulation activates large axons rather than cell bodies, grounding the interpretation of P0 as synchronized axonal spikes.","marker":"Nowak and Bullier, 1998a, b"},{"why":"Provides evidence that late positive components can reflect inhibitory processes, used to interpret the ACEP after-positivity.","marker":"Usami et al., 2015"}],"fun_headline_variants":["2 ms delay and reversed late wave distinguish brain stimulation types","Cortical vs. axonal stimulation: a 2-ms delay and opposite late potentials","Direct and indirect brain stimulation differ in early delay and late wave","Brain stimulation: DCR and ACEP differ in timing and late polarity","2 ms conduction delay and distinct late waves set DCR apart from ACEP"],"cache_read_input_tokens":16896,"weakest_assumption_plain":"The conduction-delay interpretation assumes the straight-line distance between the DES site and the ECoG electrode is a valid measure of the actual axonal path length; if the true fiber path is longer or the stimulation directly excites cortex from below, the estimated velocity and the delay explanation lose their basis.","fun_headline_variants_meta":{"raw":{"variants":["2 ms delay and reversed late wave distinguish brain stimulation types","Cortical vs. axonal stimulation: a 2-ms delay and opposite late potentials","Direct and indirect brain stimulation differ in early delay and late wave","Brain stimulation: DCR and ACEP differ in timing and late polarity","2 ms conduction delay and distinct late waves set DCR apart from ACEP"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000626,"raw_usage":{"total_tokens":2933,"prompt_tokens":1019,"completion_tokens":1914,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":635,"completion_tokens_details":{"reasoning_tokens":1818}},"tokens_in":635,"tokens_out":1914,"duration_ms":12577,"temperature":1.0,"reasoning_tokens":1818,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T13:12:14.513196+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Measure ACEP and DCR from the same recording electrode while systematically moving the white-matter stimulation site closer to and farther from the cortex, and check whether the t_zc1 delay grows with distance by the expected amount. If the delay stays roughly constant, or if a short-distance ACEP shows no delay at all, the axonal-conduction interpretation is refuted.","supporting_citations":[],"review_version":1}