{"id":"46aa426c-6fd5-448c-8b82-d9320768e1e6","arxiv_id":"2502.10322","paper_version":1,"verdict":"CONDITIONAL","confidence":"HIGH","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":0,"one_line_summary":"An offset rotation axis plus smooth redundancy weighting extends cone-beam micro-CT field-of-view by up to 2x with no pixel-size loss, shown on human lung tissue at two length scales.","lead":"This paper shows how to double the usable field of view of a lab X-ray micro-CT scanner by shifting the sample's rotation axis sideways, without shrinking the pixel size. The authors demonstrate it on a human lung lobectomy, producing multi-contrast images at 10.5 µm and high-resolution images at 450 nm.","discovery_kind":"new_application","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Eq. 11's 1D redundancy weight is validated only on the central slice; the 3D cone-beam case for off-plane points remains untested, so the no-resolution-loss claim rests on an unproven row-independence.","rationale":"The reader's weakest assumption correctly identifies the redundancy weight in Eq. 11 as the key unproven component. My stress-test sharpens this into a concrete geometric concern: the central-row simulation is only a 2D fan-beam check, whereas the actual reconstruction is 3D cone-beam, and for off-plane points the complementary rays used by the redundancy condition are not guaranteed to be symmetric in the single detector column coordinate W(x) depends on. This is a real soft spot because the headline claim, doubling the FOV without sacrificing resolution and with multi-contrast capability, depends on the weight producing quantitatively correct volumes over the full 3D field of view, not just on the central plane. The experimental results are visually impressive and the public reconstruction code is a positive independent resource, but they do not settle the quantitative 3D question. The concern is testable with a relatively simple simulation and does not by itself overturn the experimental demonstration, so it reinforces rather than changes the reader's CONDITIONAL verdict. I therefore recommend UNCHANGED.","tokens_in":15383,"tokens_out":14411,"duration_ms":164221,"concrete_test":"Run a 3D cone-beam simulation with the published geometries (beam-tracking: zsm+zmo = 870 mm, zod = 170 mm, deltaCOR = 7.6 mm, 2424 columns; and FSP: deltaCOR = 630 um) using a phantom with high-contrast features at several off-plane heights, e.g., a Defrise disc phantom or a Shepp-Logan volume with added spheres at z = +/- half the detector height. Reconstruct with Eq. 11 and the Astra vector backprojection, and compare per-slice MSE at z != 0 against the large-detector full-FOV benchmark. Additionally, ray-trace each off-plane voxel to find its two complementary source positions and detector coordinates; if the sum of the two weights deviates from 2 by more than a few percent at the cone edges, the 1D weight is insufficient for the claimed 3D multi-contrast performance.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The load-bearing step is the 1D redundancy weight W(x) of Eq. 11, applied uniformly to all detector rows and all contrast channels. The simulation in Sec. 2.4 validates this weight only on the central detector row, i.e., as a 2D fan-beam problem, with simple line integrals and no phase effects. In the actual 3D cone-beam geometry, a ray through an off-plane point and its complementary ray, the same line measured from the other source position, generally strike different detector rows. A weight depending only on the detector column x therefore cannot, in general, satisfy the line-based redundancy normalization for off-plane points. The required condition W(DCOR + beta) + W(DCOR - beta) = 2 is asserted but not derived, and it is adapted from a two-short-scan combination [37] rather than derived for a single full 360-degree scan. If row-dependent errors are non-negligible at the experimental cone angles, the 'without sacrificing spatial resolution' claim would fail away from the midplane, and the multi-contrast channels would inherit the artifacts after phase retrieval. The experimental images are compelling but qualitative and provide no numerical 3D ground-truth check.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"The paper describes an offset center-of-rotation (offset-COR) acquisition geometry for cone-beam X-ray microtomography, combined with a smooth redundancy weighting function and a full 360° scan, to extend the field of view without detector rebinning. The method is demonstrated numerically on the central detector row with a Shepp-Logan phantom and experimentally in two configurations: a beam-tracking multi-contrast scan of a resected human lung lobectomy achieving a 4.3 cm horizontal FOV at 10.5 µm voxels, and a free-space-propagation phase-contrast scan of a 2.7 mm specimen segment at 450 nm voxels. The authors claim the approach doubles the achievable FOV without sacrificing spatial resolution and is compatible with multi-contrast and multi-scale X-ray phase-contrast imaging.","tokens_in":15648,"tokens_out":3726,"duration_ms":37784,"significance":"If the full 3D cone-beam validity of the weighting is established, the contribution is practically valuable: it offers a simple, hardware-compatible way to increase FOV for lab-based multi-contrast micro-CT, with public reconstruction code and two well-executed experimental demonstrations that show plausible image quality. The simulations demonstrate that the weighting works for the central fan-beam row, and the experimental images support feasibility. However, the central claim of 'no resolution loss' and the multi-contrast applicability depend on a row-independent weighting whose off-plane behavior is not yet demonstrated; the significance is therefore conditional on additional 3D validation.","major_comments":[{"comment":"The redundancy condition for the offset-COR geometry is stated as W(D_COR + β) + W(D_COR − β) = 2, but the paper does not derive Eq. (11) or verify analytically that its piecewise definition satisfies this condition over the full redundant region. The weight is adapted from the two complementary short-scan work of Belotti et al. [37], which is a different acquisition scheme, while the present method uses a single full 360° scan. Since the weighting is the mechanism that makes truncated offset-COR projections quantitatively reconstructable, an analytic check or derivation of the redundancy property for Eq. (11) is load-bearing and currently missing.","section":"§2.3, Eq. (11)"},{"comment":"The numerical validation is performed only on the central detector row, i.e., as a 2D fan-beam problem, using a Shepp-Logan phantom and simple line integrals. In a true cone-beam geometry, a ray through an off-plane point and its complementary ray from the opposite source position generally strike different detector rows, so a weight W(x) that depends only on the detector column cannot in general satisfy a line-based redundancy normalization for off-plane points. The paper provides no 3D simulation or off-plane numerical test to show that the row-independent weighting is adequate at the experimental cone angles. This is a direct gap in the support for the claim that the FOV is doubled without sacrificing spatial resolution.","section":"§2.4, simulations"},{"comment":"The two experimental demonstrations are qualitative and do not provide a numerical check on the 3D reconstruction accuracy of the offset weighting. For instance, the line profile in Fig. 4h reports a 45 µm FWHM through an arterial lumen, but this quantity mixes the anatomical lumen width with the system spatial resolution and is not a resolution or accuracy measurement of the offset-COR reconstruction. To support the 'no loss of spatial resolution' claim, the authors should report a quantitative resolution or accuracy metric (e.g., edge response, point-like feature spread, or comparison against a full-FOV reference scan) for both geometries.","section":"§3.2, experiments"}],"minor_comments":[{"comment":"The symbol 'sng' appears to denote the sign function; please define it explicitly. Also, the piecewise conditions use 'D0 < x ≤ Dend', which omits the point x = D0; please clarify the value of W at that boundary and whether continuity is intended.","section":"§2.3, Eq. (11)"},{"comment":"The relationship between the detector-coordinate x and the angle β used in the redundancy condition is not stated explicitly; adding the conversion would help readers verify the symmetry condition against the plot in Fig. 2.","section":"§2.3"},{"comment":"The text reads 'interoperative imaging' in the paragraph on rapid reconstructions; this should be 'intraoperative imaging'.","section":"§4, Discussion"},{"comment":"For the free-space-propagation experiment, the offset is given as ΔCOR ≈ 630 µm and the FOV increase as 1.85×, but the exact source-to-detector distance and magnification are not specified in the same detail as for the beam-tracking setup; please provide the complete geometry parameters for reproducibility.","section":"§2.5, experiments"},{"comment":"The simulation uses 2701 projections at 12.5 µm pixel size and a detector of 2424 columns, but the corresponding angular sampling step and the exact phantom placement relative to the COR are not given; listing these parameters would make the simulation reproducible.","section":"§3.1, simulations"},{"comment":"The beam-tracking experiment reports a total exposure time of 18 hours and the FSP experiment 'just under 7 hours'; it would be useful to state whether these include overhead (e.g., mask dithering steps and flat-field acquisition) or only integration time.","section":"§3.2, experiments"}],"recommendation":"major_revision","confidential_remarks":"The central concern is the unvalidated extension of the 1D redundancy weight to off-plane rays in cone-beam geometry. This is fixable within the scope of the paper by adding a 3D simulation with a 3D phantom (or at least a multi-row fan-beam test) that reports MSE per slice or across the reconstructed volume, and by adding a quantitative resolution/accuracy measurement in at least one experimental geometry. If those additions are made, the paper would be a solid methods contribution. I would not reject on the basis of the current gap, but I would not accept without the 3D validation."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Good paper to know about if you work in lab-based phase-contrast CT. The core idea is not new — offset detector and offset-COR with redundancy weighting have been around for a while, and they cite the right prior work. What is new here is the adaptation to a full 360° single scan with multi-contrast beam-tracking and free-space propagation, demonstrated on real human lung lobectomy specimens at two length scales. That demonstration is the paper's real value: 1.7× and 1.85× FOV extension at 10.5 µm and 450 nm voxels, with clean images of vessels and alveoli. They ship code for the weighting and reconstruction, which is more than most papers in this area do.\n\nSoft spots: Eq. 11 is asserted rather than derived. They borrow a smooth weighting from Belotti et al. and apply it to the full 360° case, but the redundancy condition W(DCOR+β)+W(DCOR−β)=2 is stated, not proven. The simulation validates the weight only on the central detector row — effectively a 2D fan-beam test — with line integrals and no phase retrieval. The stress-test concern about off-plane points is real: in cone-beam, complementary rays for off-plane points do not strike the same detector row, and a weighting that depends only on the column cannot exactly cancel the redundancy away from the midplane. At the small cone angles used here the error is likely modest, but the paper's claim of 'without sacrificing spatial resolution' is stronger than what the evidence supports. A 3D simulation with a structured phantom, or a quantitative comparison against a full-width detector scan on the same object, would settle this. The experimental validation is qualitative — nice pictures, no numerical ground-truth comparison. That is acceptable for a methods demonstration, but it leaves the multi-contrast off-plane behaviour untested.\n\nOverall: the paper is honest, cites relevant prior work, and the method is plausibly useful. It deserves a serious referee. I would ask for a derivation or at least a careful statement of the conditions under which the 1D weight is valid in cone-beam, plus a 3D simulation. Not a desk reject.","headline":"A practical, well-demonstrated FOV-extension trick for lab phase-contrast micro-CT, with a real but fixable gap in the cone-beam weighting justification.","tokens_in":16197,"tokens_out":3565,"would_cite":true,"duration_ms":32040,"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":"Offsetting the rotation axis of a cone-beam micro-CT system can double the horizontal field of view at unchanged voxel size, and the paper demonstrates it for multi-contrast, multi-scale imaging of resected human lung tissue.","keywords":["X-ray microtomography","phase-contrast imaging","offset centre-of-rotation","field-of-view extension","beam-tracking","free-space propagation","virtual histology","cone-beam CT"],"falsifier":"Take a 3D phantom with known attenuation, phase, and dark-field distributions, simulate or measure offset-COR projections over 360°, reconstruct each channel with the proposed weighting, and compare the redundant crescent region against a full-field reference reconstruction; if the weighting is only valid for the central row, the error in the crescent will exceed the error in the always-visible disc. A physical version would scan a phantom small enough to fit both native and extended field-of-view and compare line profiles across the seam.","tokens_in":15220,"feed_emoji":"🫁","tokens_out":7163,"duration_ms":67264,"temperature":0.7,"pith_summary":"The paper claims that simply offsetting the rotation axis of a cone-beam X-ray microtomography system—rather than offsetting the detector—extends the horizontal field of view by up to a factor of two at unchanged voxel size and without losing X-ray flux. The key is a smooth redundancy weighting that lets a full 360° scan combine rays that pass through the sample more than once, replacing the need for a larger detector, rebinned geometry, or iterative reconstruction. On a resected human lung lobectomy specimen the method produced a 4.3 cm field of view at 10.5 µm voxels using beam-tracking multi-contrast imaging (attenuation, phase, dark-field), and a 2.7 mm field of view at 450 nm voxels using free-space propagation phase retrieval on the same instrument. If it holds, lab-based virtual histology can image whole tissue samples at microscopic resolution without custom hardware.","feed_headline":"Offset rotation axis doubles lab micro-CT field of view","feed_subtitle":"The same instrument images a 4.3 cm lung specimen at 10.5 µm and a 2.7 mm segment at 450 nm.","key_machinery":"The central mechanism is the offset centre-of-rotation geometry: the sample rotates about an axis displaced laterally from the source-detector line, so the full cone beam still hits the detector and no flux is lost. The load-bearing object is the smooth redundancy weighting function $W(x)$ of Eq. (11), adapted from complementary short-scan weights, which assigns full weight to rays seen once and combines rays seen twice by enforcing $W(D_{\\mathrm{COR}}+\\beta)+W(D_{\\mathrm{COR}}-\\beta)=2$ in angle about the projected centre of rotation. It is applied after ramp filtering in Eq. (12), and backprojection uses a vector cone-beam projector, so the reconstruction needs no rebinning and no iterative refinement.","core_discovery":"The paper establishes that a cone-beam micro-CT system with a fixed source and detector can image samples nearly twice as wide as its native field of view, at the same spatial resolution, by translating the rotation stage so the centre of rotation is offset from the source-detector axis and scanning a full 360°. In the offset geometry, rays in the central circular region are measured twice and rays in the outer crescent are measured once; the proposed weighting function $W(x)$ (Eq. 11) smooths between these cases in angle about the projected centre of rotation, satisfying $W(D_{\\mathrm{COR}}+\\beta)+W(D_{\\mathrm{COR}}-\\beta)=2$, and is applied to ramp-filtered projections before vector backprojection. The authors demonstrate the method in two experimental regimes on a human lung lobectomy specimen: a beam-tracking scan with 10.5 µm voxels over a 4.3 cm horizontal field of view giving quantitative attenuation, phase, and dark-field volumes that resolve vessels of tens of micrometres and emphysematous air spaces, and a free-space propagation scan with 450 nm voxels over 2.7 mm resolving alveolar septa and vessels of roughly 8 µm. They argue the same recipe extends to grating, edge-illumination, and speckle-tracking systems and to conventional cone-beam CT.","pith_inferences":["If the weighting condition holds in full 3D, the same offset-COR principle could be stacked with tiled gratings or scanning-based field-of-view extension to push beyond 2x; the paper notes compatibility but does not test the combination.","The angular-symmetry argument suggests the method should generalize to non-circular trajectories or partial 360° arcs, where the redundancy condition would take a different angular interval; that is a natural next test.","A quantitative agreement study comparing offset-COR reconstructions with a large-detector reference on the same physical phantom would separate weighting artefacts from sample-preparation effects, which the current experimental images cannot do."],"forward_implications":["A single lab instrument can first image an entire resected tissue sample at 10.5 µm voxels and then zoom into a 2.7 mm region at 450 nm voxels, giving context plus cellular detail without changing detectors or resolution.","The field-of-view extension is achieved at no loss of flux density per detector element and without rebinning or iterative reconstruction, so the extra coverage costs little in scan or computation time.","The method preserves quantitative multi-contrast channels: attenuation, integrated phase, and dark-field, because the same redundancy weighting is applied to each retrieved projection.","Because only a translation of the rotation stage is required, the approach applies directly to grating, edge-illumination, speckle-tracking, free-space propagation, and conventional cone-beam CT systems.","The demonstrated extension factors of 1.7x and 1.85x indicate the near-2x limit is reachable when the sample fills the full redundant region."],"supporting_citations":[{"why":"supplies the complementary short-scan weight form that the paper adapts to a full 360° offset-COR scan for the smooth redundancy weighting in Eq. 11.","marker":"[37]"},{"why":"establishes the redundancy condition for points symmetric about the projected centre of rotation and the sinogram-padding step used to reduce truncation artefacts.","marker":"[56]"},{"why":"provides the vector cone-beam backprojection that reconstructs the custom offset geometry without rebinning.","marker":"[55]"},{"why":"gives the beam-tracking retrieval that separates attenuation, refraction, and dark-field signals from one exposure for the multi-contrast demonstration.","marker":"[14]"},{"why":"supplies the single-distance phase retrieval used to turn the high-resolution free-space propagation projections into phase maps for reconstruction.","marker":"[52]"}],"fun_headline_variants":["Offset axis doubles micro-CT field of view","Offset rotation axis doubles micro-CT FOV","Lung micro-CT: offset axis doubles FOV, same resolution","Offset geometry doubles micro-CT FOV, retains resolution"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The paper's load-bearing assumption is that the smooth weighting function, which is validated on a central-row attenuation-only Shepp-Logan simulation and on qualitative experimental images, also correctly handles redundant rays for full 3D reconstructions and for phase and dark-field contrast channels; the redundancy property that makes it work is not derived for those cases.","fun_headline_variants_meta":{"raw":{"variants":["Offset axis doubles micro-CT field of view","Offset rotation axis doubles micro-CT FOV","Lung micro-CT: offset axis doubles FOV, same resolution","Offset geometry doubles micro-CT FOV, retains resolution"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000714,"raw_usage":{"total_tokens":3277,"prompt_tokens":1077,"completion_tokens":2200,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":693,"completion_tokens_details":{"reasoning_tokens":2138}},"tokens_in":693,"tokens_out":2200,"duration_ms":14738,"temperature":1.0,"reasoning_tokens":2138,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-07T18:35:19.165965+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Take a 3D phantom with known attenuation, phase, and dark-field distributions, simulate or measure offset-COR projections over 360°, reconstruct each channel with the proposed weighting, and compare the redundant crescent region against a full-field reference reconstruction; if the weighting is only valid for the central row, the error in the crescent will exceed the error in the always-visible disc. A physical version would scan a phantom small enough to fit both native and extended field-of-view and compare line profiles across the seam.","supporting_citations":[{"cited_title":"Extension of the cone-beam ct field-of-view using two com- plementary short scans,","cited_arxiv_id":null,"evidence_quote":"supplies the complementary short-scan weight form that the paper adapts to a full 360° offset-COR scan for the smooth redundancy weighting in Eq. 11."},{"cited_title":"Cone-beam ct from width-truncated projections,","cited_arxiv_id":null,"evidence_quote":"establishes the redundancy condition for points symmetric about the projected centre of rotation and the sinogram-padding step used to reduce truncation artefacts."},{"cited_title":"Fast and flexible x-ray tomography using the astra toolbox,","cited_arxiv_id":null,"evidence_quote":"provides the vector cone-beam backprojection that reconstructs the custom offset geometry without rebinning."},{"cited_title":"Beam tracking approach for single–shot retrieval of absorption, refraction, and dark–field signals with laboratory x–ray sources,","cited_arxiv_id":null,"evidence_quote":"gives the beam-tracking retrieval that separates attenuation, refraction, and dark-field signals from one exposure for the multi-contrast demonstration."},{"cited_title":"Simultaneous phase and amplitude extraction from a single defocused image of a homogeneous object,","cited_arxiv_id":null,"evidence_quote":"supplies the single-distance phase retrieval used to turn the high-resolution free-space propagation projections into phase maps for reconstruction."}],"review_version":1}