{"id":"9a56c8d1-939b-4c7a-8aeb-3168cc8e6f71","arxiv_id":"2411.14669","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":4,"one_line_summary":"A breath-synchronized single-grid x-ray method produced the first in vivo 4D dark-field CT of mouse lungs, tracking alveolar-size changes through the breath cycle.","lead":"Researchers captured the first four-dimensional (3D over time) x-ray dark-field images of live mouse lungs by syncing single-exposure grid imaging with the breath cycle. The technique reveals how the lung's air sacs (alveoli) expand and contract region by region, information invisible in standard CT scans.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Breath-cycle repeatability is assumed but unvalidated in Section II-D; if breaths drift or jitter, the 8 phase-binned CTs mix lung states and the temporal dark-field curves are not attributable to alveolar dynamics.","rationale":"The central claim has two parts: first successful 4D dark-field lung CT, and the physiological interpretation that the dark-field change tracks alveolar expansion. The first can withstand some limitations, but the second requires the sorted projections to be self-consistent. The most fundamental numerical condition is breath repeatability, and the paper explicitly identifies it as an assumption in Section II-D, so my concern is not manufactured. The reader flagged the same weakest assumption. I also credit genuine strengths: the single-exposure retrieval algorithm was previously validated in phantom work, the qualitative agreement with post-mortem high-resolution CT and histology supports feasibility, and the manual ROI alignment is disclosed. Those strengths argue for a feasibility demonstration but do not settle quantitative temporal correctness. The proposed split-half reconstruction is a concrete, data-only check that would distinguish genuine phase-resolved signal from breath-to-breath mixing. Until it is performed, conditional acceptance is appropriate. My read therefore does not change the reader's verdict.","tokens_in":17030,"tokens_out":3903,"duration_ms":45330,"concrete_test":"From the raw projection sequence, reconstruct two independent 180° sinograms for each of the 8 phase bins, one using only odd-numbered breaths and one using only even-numbered breaths. If the voxel-wise RMS difference between the two µd volumes at any phase bin approaches or exceeds the observed change in µd between T0 and T2 (or between bins), then breath-to-breath variability is large enough to compromise the 4D signal, and the repeatability assumption in Section II-D fails. This single split-half consistency check directly tests whether the sorted projections support self-consistent CT reconstruction.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The paper's central claim—demonstrating time-resolved 3D dark-field CT of living lungs—depends on the Section II-D assumption 'that the lungs return to the same position at the same timepoint in each breath cycle to capture the next projection within the CT.' This is the load-bearing step: each of the 8 phase bins is assembled from projections taken across 1800 separate breaths, so any breath-to-breath variability (ventilator trigger jitter, tidal-volume fluctuation, slow drift of end-expiratory lung volume, mucus plugging, atelectasis) will place two different lung states into the same bin. The 62 ms projection time (55 ms exposure) in a 500 ms breath cycle means each image already averages over about 11% of the cycle; the paper does not report a measurement of gating jitter or consistency of repeated projections. Without such a check, the observed monotonic dark-field decrease to T2 and recovery during exhalation—the main physiological result—could be produced or exaggerated by phase misassignment and partial-volume mixing rather than by alveolar expansion and contraction. This concern is about correctness of the temporal reconstruction, not about the novelty of the single-grid acquisition, which is supported by earlier algorithm and phantom work.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper presents a proof-of-principle demonstration of time-resolved three-dimensional x-ray dark-field imaging of living mouse lungs. Using a single-grid setup at 25 keV, the authors synchronized image acquisition with a mechanical ventilator, binning projections from 1800 breaths into eight equally spaced breath-phase points, and reconstructed dark-field CT volumes via filtered backprojection of the retrieved visibility signal. Data are shown for one control mouse, one β-ENaC-Tg mouse (muco-obstructive disease/emphysema), and one KP lung-cancer mouse. The authors report that dark-field CT provides contrast between lung lobes not visible in attenuation CT, that dark-field signal decreases during inhalation and recovers during exhalation in the control, and that the disease models show weaker and flatter temporal dark-field responses. The central claim is that this is the first in vivo 4D x-ray dark-field lung imaging.","tokens_in":17308,"tokens_out":5760,"duration_ms":57876,"significance":"If the central claim holds, this is a significant technical advance: it combines single-exposure dark-field retrieval with gated acquisition to obtain 4D dark-field CT, something not previously demonstrated. The work is explicitly a proof-of-principle with one animal per model, and the authors are candid about the need for future cohort studies. The quantitative link to the linear diffusion coefficient µd and its proportionality to alveolar surface-area-to-volume ratio (from prior work) gives a plausible path to functional lung assessment. The main strengths are the clean experimental design, the use of a well-characterized retrieval algorithm, and the clear presentation of the temporal dark-field trends.","major_comments":[{"comment":"The temporal reconstruction assumes the lungs return to the same position and inflation state at the same time point in every breath cycle, yet no validation of this assumption is provided. The eight phase-binned CT datasets are each assembled from projections acquired over 1800 separate breaths, and the 62 ms projection time covers roughly 11% of the 500 ms breath cycle. Ventilator trigger jitter, tidal-volume fluctuations, or slow drift (e.g., atelectasis, mucus accumulation) would mix different lung states into a single bin and could artifactually flatten or shift the dark-field temporal curves in Figs. 4 and 5. Please add a quantitative check, for example: report the measured trigger jitter, compare repeated projections acquired at the same rotation angle and breath point, or plot the attenuation-derived lung volume per bin as a consistency diagnostic. Without such a check, the physiological interpretation of the temporal dark-field decrease at T2 is not fully supported.","section":"Section II-D"},{"comment":"The cross-model comparisons are based on one animal per group (n=1) and the plotted medians have no error bars or uncertainty estimates. The claim that the β-ENaC-Tg and cancer lungs generate weaker dark-field signals and smaller relative changes in magnitude compared to the control lung is therefore not statistically grounded. At minimum, please provide the spread of the dark-field values across the slices or voxels used in each ROI, or a noise estimate propagated from the retrieval algorithm, and state explicitly that the cross-model differences are illustrative single-animal observations.","section":"Section IV, Figs. 4 and 5"},{"comment":"The manual slice alignment and manual ROI masks are a source of potential bias in the measured temporal dark-field curves. The alignment is performed on the attenuation CT and then applied to the dark-field CT, but the two modalities have different effective spatial resolution (the dark-field analysis decimates by a factor of 10 in projection space) and may not co-register perfectly. Please quantify the sensitivity of the Fig. 4 and Fig. 5 curves to alignment choices (e.g., ±1 slice) and to mask boundaries, or replace the manual steps with automated registration.","section":"Section III-C"}],"minor_comments":[{"comment":"Please state explicitly that the CT reconstruction is applied to -ln(V) projection data (as is done for attenuation) in order to map µd; otherwise the reader may assume the visibility itself is being reconstructed.","section":"Section III-B, Eq. (1)"},{"comment":"The text and caption refer to the 'whole lung' measurement, but this is a median over a single 97-µm axial slice, not over the entire lung volume.","section":"Fig. 4 caption"},{"comment":"The phrase 'the first in vivo 4D x-ray dark-field lung imaging' appears as a categorical statement; adding 'to our knowledge' would be more precise given the rapid evolution of the field.","section":"Abstract and Introduction"},{"comment":"The statement that a 20% increase in µd implies a 20% increase in surface-area-to-volume ratio should reference the validity conditions of the proportionality derived in [67]; as written it could be read as a universal relation.","section":"Section V-A"}],"recommendation":"major_revision","confidential_remarks":"The manuscript is a good fit for physics.med-ph and the result is likely of interest to the biomedical imaging community. The main risk is the unvalidated breath-cycle repeatability assumption and the lack of uncertainty quantification, both of which can be addressed with additional analysis. The authors are well-known in the field; the high self-citation count is not unusual for a specialized technique. I would support publication after major revision."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Colleague,\n\nYou should know about this paper: it really is the first in vivo 4D dark-field CT of lungs. The combination of single-exposure grid retrieval with breath-synchronized acquisition across 1800 breaths to form 8 phase-binned CTs is a genuine step beyond prior 2D dynamic dark-field (Gradl 2018) and static 3D dark-field CT (Velroyen 2015, Burkhardt 2021). That claim holds up. The paper also does some things well: the experiment is described in enough detail to reproduce, the CNR jump from 2.4 to 5.5 between lobes is a solid demonstration of the added information, and the post-mortem high-resolution CT and histology provide some validation. The supplementary data are deposited with a DOI. The authors are also upfront about limitations and future work.\n\nThe soft spots are the ones you'd suspect, and they're real. Section II-D states the load-bearing assumption: lungs return to the same position at the same timepoint in each breath cycle. That assumption is never directly validated. No gating jitter measurement, no repeat-projection check. With a 62 ms exposure in a 500 ms cycle, each projection already averages ~11% of the breath, and if ventilation drifts or mucus plugs move, the phase bins blur or mix lung states. The temporal dark-field curves could then be partially a reconstruction artifact. The stress-test note makes this point and it lands. That said, the observed trend—decrease to T2, recovery at exhalation—is physiologically plausible, and the fact that they see a smooth consistent trend across control and disease models gives me some confidence the gating works. But 'some confidence' is not a measurement. The paper needs at least one check: e.g., comparing the attenuation CT at the same phase bin from repeated breaths, or reporting the spread of the ventilator trigger.\n\nOther limitations are proportionate for a proof-of-principle: one animal per model, no error bars, manual ROI tracing and slice alignment. These are addressable in a follow-up, but they should be stated as qualifications rather than as solid quantitative biology. The dark-field signal is also retrieved with a low grid visibility (0.08 in the horizontal direction) and no explicit correction for source-size blur; the authors argue it's fine for isotropic scattering, and that's plausible but worth a footnote.\n\nMy verdict: this is a worthwhile feasibility demonstration, not a definitive physiological study. The central method is novel and reproducible enough to deserve referee time. The main revision should be to validate the breath-cycle repeatability and tone down the temporal interpretation until that's done. I'd send it to peer review, and I'd cite it as the first 4D dark-field lung CT, with a caveat about the gating assumption.","headline":"First real 4D dark-field lung CT; the breath-repeatability assumption needs an explicit check before the temporal curves are taken at face value.","tokens_in":17949,"tokens_out":4153,"would_cite":true,"duration_ms":38604,"reading_group":"yes","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This paper claims the first in vivo 4D x-ray dark-field lung imaging in mice, achieved by synchronizing single-exposure grid-based imaging with the breath cycle, and shows that the time-resolved dark-field signal tracks regional alveolar…","keywords":["x-ray dark-field imaging","single-grid imaging","time-resolved computed tomography","lung imaging","in vivo mouse models","alveolar function","respiratory gating","phase-contrast x-ray imaging"],"falsifier":"Repeatedly scan the same mouse under identical ventilation and compare the eight reconstructed timepoints between scans, or deliberately change ventilation (for example, different PEEP or a single induced atelectasis in one lobe) and check whether the dark-field time course for the affected region flattens or misregisters. A direct quantitative test: at a fixed trigger point, measure the displacement of the diaphragm or a bronchial bifurcation across many breaths; if the lung position varies by more than one reconstructed voxel, the gating assumption underpinning the eight CT volumes is violated.","tokens_in":16859,"feed_emoji":"🫁","tokens_out":7824,"duration_ms":72792,"temperature":0.7,"pith_summary":"This paper reports the first time-resolved three-dimensional (4D) x-ray dark-field imaging of living mouse lungs. It combines single-exposure grid-based dark-field retrieval with ventilation-synchronized acquisition: projections taken at eight equally spaced points in the breath cycle, across 1800 breaths, are binned into eight CT volumes. The authors show that the dark-field signal, which arises from unresolved alveoli, drops as the lungs inflate and recovers as they deflate, and that diseased lungs (muco-obstructive disease/emphysema and lung cancer) show weaker signals and flatter temporal responses than a control. If the method holds up, it turns dark-field CT into a functional imaging tool that can map regional alveolar expansion, information absent from attenuation CT, and could assess lung disease locally rather than through whole-lung averages.","feed_headline":"First 4D dark-field CT tracks mouse lungs through each breath","feed_subtitle":"A single-exposure grid method synced to ventilation maps regional alveolar expansion across eight breath phases.","key_machinery":"The load-bearing mechanism is single-grid dark-field retrieval: a periodic attenuating grid placed upstream of the sample casts a shadow pattern on the detector, and comparing the sample-and-grid image with a grid-only image yields the visibility ratio $V=\\exp(-\\mu_d T)$, where $\\mu_d$ is the linear diffusion coefficient and $T$ is the thickness. Because the dark-field signal is extracted from a single exposure, 62 ms per projection, no grating stepping is needed. The second mechanism is ventilation-gated acquisition: a timing hub triggered by the ventilator records images at eight equally spaced breath points while the mouse rotates continuously, so 1800 breaths produce eight self-consistent 180-degree projection sets that are reconstructed by filtered back-projection into eight dark-field CT volumes. The $\\mu_d$ value is interpreted as proportional to the surface-area-to-volume ratio of the unresolved alveoli, which is what makes the temporal dark-field curve a proxy for alveolar expansion.","core_discovery":"The central claim is that 4D dark-field CT of breathing mouse lungs is feasible with a single-exposure grid-based setup, and that the reconstructed dark-field signal tracks alveolar size changes throughout the breath cycle. In the data, the median dark-field signal in a mid-lung slice falls to a minimum at the inhalation timepoint and returns during exhalation, consistent with alveoli expanding then contracting; the attenuation images over the same breath points look nearly identical. Spatial comparisons show dark-field contrast between lung lobes that is far weaker in attenuation CT (contrast-to-noise ratio 5.5 versus 2.4), and post-mortem high-resolution CT confirms that the lobe with weaker dark-field signal contains larger alveoli. In the β-ENaC-Tg and lung-cancer mice, dark-field signals are lower and change less through the breath than in the control, consistent with enlarged alveoli, tumour-obstructed airflow, or a reduced alveolar fraction. The paper presents these as proof-of-principle demonstrations rather than cohort statistics.","pith_inferences":["If breath-to-breath reproducibility holds in larger cohorts, the same eight-point gating could be extended to cardiac gating with a high-brightness source, removing the remaining motion artefact the authors note.","The parameter-free link between $\\mu_d$ and surface-area-to-volume ratio suggests a direct quantitative endpoint for preclinical drug trials: a treatment that restores alveolar expansion should raise the dark-field temporal swing back toward control values.","Moving to laboratory x-ray sources would require correcting beam-hardening and visibility-hardening biases that alter quantitative $\\mu_d$; without those corrections, lab-based 4D dark-field CT would likely be only qualitative.","The manual slice alignment used for ROI tracking could be automated with image registration or x-ray velocimetry, which would make regional dark-field measurements less operator-dependent and better suited to blinded studies."],"forward_implications":["Regional lung function can be read out voxel by voxel: dark-field CT maps where alveoli are larger or more loosely packed, and how much they expand during a breath, rather than averaging the whole lung into a single number.","Disease models show a measurable signature: compared with control lung, muco-obstructive/emphysema and lung-cancer lungs exhibit weaker dark-field signal and smaller breath-to-breath variation, suggesting the technique can characterize alveolar destruction or airflow obstruction.","Because $\\mu_d$ is proportional to surface-area-to-volume ratio, the eight timepoints give a local alveolar expansion curve, so the technique could quantify regional lung mechanics without contrast agents.","The same synchronized single-exposure protocol can in principle be applied to other periodic sample changes, such as metal-foam formation or dissolving microstructures, by collecting multiple rotations instead of one."],"supporting_citations":[{"why":"Supplies the single-grid dark-field retrieval algorithm used to extract the visibility signal from each 62 ms projection.","marker":"[26]"},{"why":"Provides the calibration relating dark-field signal to microstructure size and the formula for choosing the propagation distance, used for interpreting alveolar size.","marker":"[27]"},{"why":"Establishes the exponential visibility model $V=\\exp(-\\mu_d T)$ that the paper uses to reconstruct dark-field CT volumes.","marker":"[20]"},{"why":"Demonstrates dynamic in vivo dark-field imaging in mice with grating interferometry, the two-dimensional predecessor that this work extends to time-resolved CT.","marker":"[52]"},{"why":"Introduces the single-attenuation-grid single-exposure scheme that the paper's experimental setup directly uses.","marker":"[53]"},{"why":"Describes the ventilator-triggered timing hub and dynamic synchrotron respiratory imaging methods that make eight-point breath gating possible.","marker":"[62]"},{"why":"Provides the paraxial diffusion-field result that $\\mu_d$ is proportional to surface-area-to-volume ratio, the physical interpretation of the measured dark-field changes.","marker":"[67]"}],"fun_headline_variants":["4D dark-field X-ray tracks live mouse lungs in real time","Single-exposure dark-field CT images mouse lungs through breath","First 4D dark-field lung imaging in vivo catches alveoli in action","Breath-synced dark-field X-ray maps alveolar size changes in mice","Live 4D dark-field CT of mouse lungs reveals alveolar breathing cycles"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The approach assumes that at the same trigger point in every breath the lungs are in the same position and inflation state, so 1800 separate breaths can be pooled into eight consistent CT datasets; if breathing drifts or the lung does not return to the same state, the time-resolved dark-field curves blur.","fun_headline_variants_meta":{"raw":{"variants":["4D dark-field X-ray tracks live mouse lungs in real time","Single-exposure dark-field CT images mouse lungs through breath","First 4D dark-field lung imaging in vivo catches alveoli in action","Breath-synced dark-field X-ray maps alveolar size changes in mice","Live 4D dark-field CT of mouse lungs reveals alveolar breathing cycles"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000952,"raw_usage":{"total_tokens":4098,"prompt_tokens":1018,"completion_tokens":3080,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":634,"completion_tokens_details":{"reasoning_tokens":2986}},"tokens_in":634,"tokens_out":3080,"duration_ms":20677,"temperature":1.0,"reasoning_tokens":2986,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-12T15:01:32.565481+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Repeatedly scan the same mouse under identical ventilation and compare the eight reconstructed timepoints between scans, or deliberately change ventilation (for example, different PEEP or a single induced atelectasis in one lobe) and check whether the dark-field time course for the affected region flattens or misregisters. A direct quantitative test: at a fixed trigger point, measure the displacement of the diaphragm or a bronchial bifurcation across many breaths; if the lung position varies by more than one reconstructed voxel, the gating assumption underpinning the eight CT volumes is violated.","supporting_citations":[{"cited_title":"Quantifying the x-ray dark-field signal in single-grid imaging,","cited_arxiv_id":null,"evidence_quote":"Supplies the single-grid dark-field retrieval algorithm used to extract the visibility signal from each 62 ms projection."},{"cited_title":"On the quantification of sample microstructure using single-exposure x-ray dark-field imaging via a single-grid setup,","cited_arxiv_id":null,"evidence_quote":"Provides the calibration relating dark-field signal to microstructure size and the formula for choosing the propagation distance, used for interpreting alveolar size."},{"cited_title":"Quantitative x-ray dark-field computed tomography,","cited_arxiv_id":null,"evidence_quote":"Establishes the exponential visibility model $V=\\exp(-\\mu_d T)$ that the paper uses to reconstruct dark-field CT volumes."},{"cited_title":"Quantitative single- exposure x-ray phase contrast imaging using a single attenuation grid,","cited_arxiv_id":null,"evidence_quote":"Introduces the single-attenuation-grid single-exposure scheme that the paper's experimental setup directly uses."},{"cited_title":"Methods for dynamic synchrotron X-ray respira- tory imaging in live animals,","cited_arxiv_id":null,"evidence_quote":"Describes the ventilator-triggered timing hub and dynamic synchrotron respiratory imaging methods that make eight-point breath gating possible."},{"cited_title":"Paraxial diffusion-field retrieval,","cited_arxiv_id":null,"evidence_quote":"Provides the paraxial diffusion-field result that $\\mu_d$ is proportional to surface-area-to-volume ratio, the physical interpretation of the measured dark-field changes."}],"review_version":1}