{"id":"0ad44acb-c0a4-4f48-9f53-d6eb0022e793","arxiv_id":"2608.07687","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Adding a short 20 ms T2-preparation pulse to 3D-QALAS improves short-T2 (myelin water) estimation and enables vendor-agnostic, motion-corrected joint T1/T2/MWF mapping in under 5 minutes.","lead":"This paper presents a vendor-neutral MRI sequence that maps T1, T2, and myelin water fraction across the whole brain in about 4.7 minutes, with built-in correction for head motion. It could make quantitative brain imaging practical for large multi-site studies, including pediatric populations where motion is common.","discovery_kind":"new_method","skeptic_critique":{"model":"deepseek-v4-flash","headline":"MWF estimates rest on fixed three-compartment no-exchange model assumptions, acknowledged in the Discussion, and the 3D-GRASE comparison is not an independent validation.","rationale":"The reader's weakest assumption identifies the same concern: the three-compartment no-exchange model with fixed relaxation parameters. This is the most load-bearing uncertainty for the MWF component of the central claim. Other candidate concerns—small in vivo cohort, missing NRMSE variance, imperfect phantom slopes, and limited vendor coverage—affect precision, generalizability, or completeness of validation, but they do not underdetermine the core quantity being measured. In contrast, the no-exchange fixed-parameter model is baked into the dictionary that produces the MWF maps; if it is wrong for the intended pediatric or pathological population, the MWF values are biased in a way that no acquisition or reconstruction improvement can correct. The Discussion already concedes this, which is honest but also makes it the central limitation. The proposed test targets the two model axes directly (myelin-water T2 and exchange rate) and observes whether the resulting MWF estimates move enough to matter. Because the paper's verdict is already CONDITIONAL on this class of limitation, my analysis does not change the verdict; it sharpens the condition under which the claim should be accepted.","tokens_in":22089,"tokens_out":5284,"duration_ms":51909,"concrete_test":"Re-run the in vivo MWF fitting on the same MWF-QALAS reconstructions (Section 3.3) with a family of dictionaries: vary myelin-water T2 from 10 to 25 ms while keeping T1 fixed at 150 ms, and add Bloch-McConnell exchange between myelin and intra/extracellular pools with rates k=0, 1, 5, and 10 s^-1. Compare resulting structure-wise MWF values against the reported 3D-GRASE values and against the fixed-model maps. If mean white-matter MWF shifts by more than roughly 2 percentage points, or if the ICC/Pearson correlation against 3D-GRASE drops materially, the fixed-parameter no-exchange assumption is load-bearing and MWF should be reported with this uncertainty. If MWF is stable across this range, the concern is mitigated.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The most load-bearing assumption in the MWF portion of the central claim is the fixed-parameter, non-exchanging three-compartment model in Section 2.3: myelin water (T1=150 ms, T2=15 ms), intra/extracellular water (T1=1100/1300 ms, T2=50/70 ms), and free water (T1=4500 ms, T2=500 ms). The claim that the sequence produces meaningful MWF maps depends on this model being accurate in the studied population. If myelin-water T2 differs from 15 ms in developing or pathological tissue, or if water exchange transfers magnetization between pools over the 20 ms T2-prep and 5.8 ms echo spacing, the dictionary fit will reallocate signal among compartments and bias MWF even with perfect acquisition, reconstruction, and motion correction. The Discussion (Section 4) explicitly concedes that MWF is not a direct measure of myelin, that accuracy depends on model assumptions that can vary across populations and pathologies, and that no histological validation was performed. The in vivo comparison to 3D-GRASE is reassuring but is not independent: 3D-GRASE MWF is itself a model-based estimate, and the comparison was performed in only three healthy adults. Thus the external validity of the MWF numbers in the intended pediatric and clinical applications is underdetermined by the evidence presented.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents MWF-QALAS, a Pulseq-based vendor-agnostic 3T quantitative MRI sequence that extends 3D-QALAS with a second short T2-preparation (TE=20 ms) and a sixth FLASH readout, enabling joint T1, T2, and myelin water fraction (MWF) mapping in about 4.7 minutes at 1 mm isotropic resolution. A center-out Cartesian spiral ordering supports self-navigated retrospective rigid motion correction in a regularized temporal subspace reconstruction. A turbo-AFI scan provides B1+ and coil sensitivity calibration, and dictionary-based Bloch simulation is used for T1/T2 and three-compartment MWF estimation. Validation includes numerical simulations, a NIST/ISMRM phantom, three healthy adult volunteers on Siemens and GE scanners, controlled motion experiments, and illustrative pediatric cases.","tokens_in":22252,"tokens_out":8107,"duration_ms":77249,"significance":"If the results hold, this is a practical contribution to multi-site and pediatric neuroimaging: the sequence is open-source, expressed in vendor-neutral Pulseq, includes integrated motion correction and harmonized calibration, and improves short-T2 estimation relative to standard 3D-QALAS (short-T2 simulation bias -0.1 ms vs -3.1 ms; phantom low-T2 slope 0.91 vs 0.78). The multi-layer validation and public code release are genuine strengths. The principal caveat is that MWF is a model-derived estimate under a fixed, non-exchanging three-compartment model, and the only in vivo MWF comparison is against 3D-GRASE in three adults; external validity in pediatric and pathological tissue is therefore not yet established.","major_comments":[{"comment":"The central claim of myelin water fraction mapping rests on the fixed three-compartment, no-exchange model in Section 2.3, which fixes myelin-water T2 at 15 ms and intra/extracellular T2 at 50/70 ms. The Discussion acknowledges that MWF is not a direct myelin measure and that these assumptions may vary across populations and pathologies, but the in vivo validation in Section 3.3 compares MWF-QALAS to 3D-GRASE in only three healthy adults, and 3D-GRASE is itself a model-based estimate. With ICC=0.732 and Pearson r=0.757, the agreement is moderate rather than strong. This evidence is insufficient to support unbiased MWF mapping in the pediatric and pathological populations emphasized in the Introduction. Please add a sensitivity analysis varying compartment relaxation times and/or exchange, or restrict the claims to 'short-T2 water fraction under the assumed model' throughout the title, abstract, and results.","section":"Section 2.3, 3.3, 4"},{"comment":"The numerical simulation is a self-consistency check rather than an independent accuracy validation: the data are generated with the same Bloch signal model that builds the dictionary. The phantom experiments provide the non-circular accuracy check, but the phantom T2 regression slope is 0.86 overall and 0.91 for T2 <= 70 ms, so absolute accuracy is still imperfect. The text should describe the simulation results as demonstrating relative improvement and short-T2 sensitivity, not standalone accuracy.","section":"Section 2.5, 3.1"},{"comment":"Motion robustness is a core contribution, but the quantitative motion experiments are single realizations without variance estimates, and in Experiment 6 the corrected NRMSE remains 20.47%. The Discussion acknowledges residual error, but the abstract and results should more clearly state the empirical motion range and residual errors rather than implying general motion immunity.","section":"Section 3.5"}],"minor_comments":[{"comment":"The second T2-preparation is given as TE=80 ms 'as in the original 3D-QALAS', but standard QALAS is described elsewhere as TE=100 ms; please reconcile this inconsistency.","section":"Section 2.1, 3.1"},{"comment":"The dictionary grid step sizes and the wavelet regularization parameter are said to be empirically selected; please report the actual values used so that the short-T2 estimates are reproducible.","section":"Section 2.3"},{"comment":"Cross-vendor MWF agreement is moderate (ICC=0.771; pairwise Pearson 0.706-0.902); the sentence comparing these results to within-vendor MRF repeatability is not an apples-to-apples comparison and should be rephrased.","section":"Section 3.4"},{"comment":"The pediatric cases are illustrative; the text should label them as such and not as validation of MWF accuracy in children, particularly since no reference standard was acquired in those subjects.","section":"Section 3.3, 3.6"},{"comment":"The abstract mentions B1- mapping, while the methods describe coil sensitivity estimation from the ACS region; clarify the relationship between B1- and receive sensitivity.","section":"Section 2.4, abstract"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is within scope for a quantitative MRI methods journal and the open-source package is genuinely useful. The main risk is overclaiming MWF validity; the authors should either provide a sensitivity analysis or soften the language in the title and abstract. The comparison to MRF repeatability should also be corrected. If these points are addressed, I would support publication."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The genuinely new piece is the second, shorter T2-prep (TE=20 ms) added to 3D-QALAS, and the simulations and phantom data support the claim that it reduces short-T2 bias. The fact that the whole acquisition and reconstruction are released as open-source Pulseq code is a real plus, and the paper is honest about its limitations.\n\nThe vendor-agnostic framing is mostly credible: the same sequence definition was run on Siemens and GE scanners, with a tAFI calibration scan integrated. The self-navigation motion correction works at TR resolution (~5.4 s), which is reasonable for the pediatric use case, and the retrospective simulations show reduced NRMSE even for 4 mm/1° motion.\n\nThe soft spots are real but proportionate. The MWF estimates rest on a fixed three-compartment no-exchange model with fixed relaxation parameters; the paper acknowledges this. That means MWF is a model-derived estimate, not a direct measurement. The 3D-GRASE comparison is not an independent validation—GRASE MWF is itself model-based and carries its own assumptions. With three healthy adults, the agreement statistics (ICC 0.73, Pearson 0.76) demonstrate feasibility, not clinical accuracy. The phantom T2 slopes, while improved (0.78→0.91 for T2≤70 ms), are still below unity, so residual bias remains.\n\nThe motion correction NRMSE values come from single runs with no variance estimates, which tempers the strength of that claim but is not a fatal flaw. The pediatric cases are explicitly illustrative. The numerical simulation in Section 2.5 is a self-consistency check rather than independent evidence, but the phantom data carries the real validation load.\n\nOverall, the central claim—improved short-T2 sensitivity with dual T2-prep, inside a vendor-agnostic package—holds up. The MWF specificity is conditional, and the authors say so. This paper deserves a serious referee. After revision (larger cohort, error bars on motion metrics, and framing GRASE as a comparison rather than ground truth), it would be a useful addition to the qMRI literature.","headline":"Dual T2-prep QALAS is a real, well-validated sequence modification that improves short-T2 estimation, and the vendor-agnostic open-source package is a practical contribution—but the MWF numbers rest on model assumptions and a small in vivo cohort, so the paper is conditionally acceptable after revision.","tokens_in":23058,"tokens_out":1543,"would_cite":true,"duration_ms":16409,"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":"The paper proposes a vendor-agnostic MRI sequence that jointly maps T1, T2, and myelin water fraction from one motion-corrected 4.7-minute scan, validated in simulations, phantoms, volunteers, and pediatric cases.","keywords":["quantitative MRI","relaxometry","myelin water fraction","3D-QALAS","motion correction","B1 mapping","Pulseq","vendor-agnostic imaging"],"falsifier":"Acquire the proposed sequence on a multi-compartment phantom with known short-$T_2$ volume fractions and no water exchange, or on ex vivo brain tissue with histologically stained myelin; if the fitted myelin water fraction deviates from the known fraction beyond the reported agreement limits (roughly $\\pm 0.05$ absolute in vivo), the three-compartment model or the sequence's short-$T_2$ sensitivity is falsified.","tokens_in":21771,"feed_emoji":"🧠","tokens_out":13058,"duration_ms":108653,"temperature":0.7,"pith_summary":"MWF-QALAS is a modified MRI acquisition that adds a short $T_2$-preparation module (echo time 20 ms) to the existing QALAS (Look-Locker acquisition with $T_2$ preparation) scheme and reads out six FLASH images per repetition, targeting the fast-decay signal of myelin water that standard long-echo preparations miss. The paper's claim is that this single scan, lasting 4.7 minutes at 1 mm isotropic resolution at 3T, can jointly produce accurate $T_1$, $T_2$, and myelin water fraction maps on scanners from two major vendors. Supporting evidence includes simulations in which mean short-$T_2$ bias improved from $-3.1$ ms with standard QALAS to $-0.1$ ms with the modified sequence, and phantom measurements in which the short-range $T_2$ slope improved from 0.78 to 0.91. The package also includes a 31-second harmonized $B_1^+$ and $B_1^-$ calibration scan and self-navigation-based retrospective motion correction that works without external hardware. If correct, this addresses a practical gap: pediatric and multi-site studies need myelin-sensitive relaxometry that is fast, motion-robust, and reproducible across scanners.","feed_headline":"A 4.7-minute MRI scan maps T1, T2, and myelin","feed_subtitle":"Vendor-neutral implementation with built-in motion correction makes fast relaxometry practical for multi-site pediatric studies.","key_machinery":"The load-bearing object is the MWF-QALAS pulse sequence itself, a myelin-sensitive variant of the QALAS scheme: two $T_2$-preparation modules (20 ms and 80 ms), one inversion pulse, and six FLASH readouts per repetition that sample the recovery curve at effective inversion times of 110, 1010, 1910, 2810, and 3710 ms. This design is coupled to a temporal subspace reconstruction in which the signal evolution is represented by a low-dimensional basis and a motion-aware non-uniform fast Fourier transform operator applies per-repetition rigid transforms estimated from self-navigation images. Parameter maps come from Bloch-simulated dictionary matching for $T_1$ and $T_2$, and from a partial-volume dictionary for three water compartments with fixed relaxation values: myelin water, intra/extra-cellular water, and free water. The short 20 ms preparation is what gives the sequence sensitivity to the fast-decaying myelin component that the standard 100 ms preparation cannot see.","core_discovery":"The central discovery is that the myelin-water short-$T_2$ component, which decays too quickly for the standard 100 ms $T_2$-preparation module, becomes measurable when a second preparation with 20 ms echo time is inserted before the standard one, with both followed by FLASH readouts within the same repetition. The authors show this dual-preparation design keeps $T_1$ mapping accuracy essentially unchanged while substantially reducing short-$T_2$ bias, and they integrate it with dictionary-based Bloch-simulation matching and a three-compartment partial-volume model to extract myelin water fraction. They further claim that the whole workflow, including motion estimation from undersampled self-navigation volumes and retrospective motion-corrected subspace reconstruction, can be expressed in an open, hardware-independent sequence format and run on scanners from two vendors without vendor-specific sequence code.","pith_inferences":["An untested implication is that improved short-$T_2$ sensitivity may also help quantify short-$T_2$ compartments in disease, but the paper only demonstrates normal adult and pediatric brains.","The motion-correction result at 5.4 s temporal resolution implies a practical limit: intra-repetition motion, or motion larger than about one $B_1$ voxel relative to calibration, could still bias maps because a single $B_1$ map is assumed.","The model-derived myelin water fraction is not histologically validated; if exchange or population-specific relaxation times matter, estimates in developing or diseased tissue could diverge from true myelin content even though the acquisition is sound.","The 13.02% mean percent difference from the reference myelin-water method suggests the two techniques measure overlapping but not identical quantities; a direct calibration study would need to decide which reference is closer to true myelin."],"forward_implications":["A single 4.7-minute whole-brain acquisition replaces separate long scans for $T_1$, $T_2$, and myelin water fraction mapping, at 1 mm isotropic resolution and 3T.","Short-$T_2$ accuracy improves enough to lower mean bias in the below-30 ms range from $-3.1$ ms to $-0.1$ ms, directly benefiting myelin-relevant measurements.","Self-navigation motion correction removes the need for external tracking hardware; retrospective simulations reduce normalized root-mean-square error against a motion-free reference from 29.53% to 13.16% using the estimated trajectory.","Because both sequence and reconstruction are expressed in an open, hardware-independent format, the same protocol can run on scanners from two major vendors, with cross-vendor intraclass correlations of 0.987 for $T_2$ and 0.771 for myelin water fraction.","Open-sourced sequences and code permit other sites to run the protocol without vendor-specific re-implementation, subject only to an interpreter for the sequence format and local safety checks."],"supporting_citations":[{"why":"Introduces the original 3D-QALAS sequence that MWF-QALAS modifies by adding a short $T_2$-preparation module.","marker":"(Kvernby et al., 2014)"},{"why":"Shows cross-vendor QALAS feasibility and provides the harmonization baseline this work extends to myelin mapping.","marker":"(Fujita et al., 2024)"},{"why":"Provides the dictionary-based $T_1$/$T_2$ fitting approach used for Bloch-simulated parameter estimation.","marker":"(Cho et al., 2024)"},{"why":"Supplies the multi-compartment model assumptions for myelin, intra/extra-cellular, and free water used in MWF estimation.","marker":"(Chen et al., 2019)"},{"why":"Contributes the partial-volume dictionary method that solves for the water-fraction compartments.","marker":"(Deshmane et al., 2019)"},{"why":"Original actual flip-angle imaging method that the rapid turbo AFI $B_1^+$ calibration extends.","marker":"(Yarnykh, 2007)"},{"why":"Provides the retrospective k-space motion-correction framework with phase shifts and rotated k-space coordinates.","marker":"(Gallichan et al., 2016)"},{"why":"Establishes the temporal-subspace representation used to reconstruct the six-contrast signal evolution.","marker":"(Tamir et al., 2017)"},{"why":"Fast multi-echo 3D-GRASE myelin water imaging used as the in vivo reference for MWF comparison.","marker":"(Piredda et al., 2021)"},{"why":"Supplies the hardware-independent sequence description format that underpins the vendor-agnostic implementation.","marker":"(Layton et al., 2017)"}],"fun_headline_variants":["One scan, any scanner: T1, T2, and myelin maps","Motion-robust MRI maps myelin and relaxation without hardware","Dual-prep MRI captures fast-decaying myelin signal","Vendor-agnostic relaxometry with built-in motion correction"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The entire myelin water fraction estimate assumes each voxel is a weighted sum of exactly three non-exchanging water pools with fixed relaxation values (myelin water with $T_2$ = 15 ms, intra/extra-cellular water, and free water), so if real tissue has different relaxation times or water exchange, the fraction will be biased even with perfect acquisition and reconstruction.","fun_headline_variants_meta":{"raw":{"variants":["One scan, any scanner: T1, T2, and myelin maps","Motion-robust MRI maps myelin and relaxation without hardware","Dual-prep MRI captures fast-decaying myelin signal","Vendor-agnostic relaxometry with built-in motion correction"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000202,"raw_usage":{"total_tokens":1357,"prompt_tokens":897,"completion_tokens":460,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":513,"completion_tokens_details":{"reasoning_tokens":386}},"tokens_in":513,"tokens_out":460,"duration_ms":4394,"temperature":1.0,"reasoning_tokens":386,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-11T00:24:20.705950+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Acquire the proposed sequence on a multi-compartment phantom with known short-$T_2$ volume fractions and no water exchange, or on ex vivo brain tissue with histologically stained myelin; if the fitted myelin water fraction deviates from the known fraction beyond the reported agreement limits (roughly $\\pm 0.05$ absolute in vivo), the three-compartment model or the sequence's short-$T_2$ sensitivity is falsified.","supporting_citations":[],"review_version":1}