{"id":"8c27f814-e342-4629-9a39-cddb406c4c46","arxiv_id":"1908.08782","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":4.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"The paper releases and documents eight 3D cone-beam CT sinogram datasets of a ladybug, with full measurement geometry, for dynamic tomography research.","lead":"This paper documents an open dataset of 3D dynamic X-ray CT scans of a ladybug, available on Zenodo. It provides 360-projection cone-beam sinograms from eight different measurement positions for use in tomography research.","discovery_kind":"extension","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Stated geometry is self-contradictory: 1024 columns × 48 µm = 49.2 mm, not Fig. 7's W=102.4 mm; users cannot know which detector geometry to enter in ASTRA.","rationale":"This is a data descriptor, not a new scientific method, so the central claim is that an accessible, usable benchmark exists and that the documentation is sufficient to reconstruct from it. The archive link and file list are concrete, and the stated 8×1024×360×186 structure is directly checkable, so the data-availability part is credible. The weakest link is exactly the acquisition geometry needed for quantitative reconstruction, which the reader also flagged. Our review sharpens that concern: the manuscript's own numbers disagree with one another, making the geometry ambiguous without any calibration to resolve it. This is an internal inconsistency, not a disagreement with consensus, so it is a genuine correctness risk in the documentation rather than a matter of opinion. The tube-current error is real but does not affect reconstruction; it only reinforces the need for a corrected specification. A one-line clarification from the authors, or a calibration scan, would resolve the issue. Therefore the reader's CONDITIONAL verdict is appropriate and no change is needed.","tokens_in":2861,"tokens_out":10003,"duration_ms":110720,"concrete_test":"Download sinogram1.mat from Zenodo; verify shape 1024×360×186. Reconstruct the central selected row with ASTRA cone-beam FBP using exactly two geometries: (A) detector pixel size 48 µm, FOD=111 mm, FDD=241 mm, 186 central rows; (B) detector width 102.4 mm and height 96.7 mm (pixel pitch 100 µm) with the same FOD/FDD and row crop. Then measure the lateral extent of the reconstructed ladybug, for example by thresholding at 50% of the maximum. Geometry (B) should yield a physically plausible ladybug width (about 5–8 mm) and match Figure 2; geometry (A) should compress the object by roughly a factor of 2.08. The variant that reproduces Figure 2 and a plausible size identifies which reported value is the detector pixel pitch; if neither does, the geometry is mis-specified and a calibration scan is needed.","verdict_should_be":"UNCHANGED","load_bearing_attack":"Section 3 states the detector pixel size is 48 µm and images are 1024×967, while Figure 7 gives detector height H=96.7 mm and width W=102.4 mm. These are mutually incompatible: 1024×48 µm = 49.15 mm and 967×48 µm = 46.42 mm. If 48 µm is actually the effective pixel size in the object plane (FDD/FOD=241/111≈2.17, so 48 µm ≈ 100 µm/2.17), then the detector pixel pitch is about 100 µm and the H/W values are consistent, but the text never says this. The same paragraph also reports 400 mA at 50 kV with a 50 W maximum tube output, which is physically impossible and reduces confidence in the parameter list. The load-bearing issue is not whether the files exist but that the one number users must pass to ASTRA or opTomo to reconstruct quantitatively from the 1024-column sinograms is ambiguous. The paper lacks any calibration scan or validation object to disambiguate.","agreement_with_reader":"partial"},"referee_report":{"model":"deepseek-v4-flash","summary":"This paper documents an open X-ray computed tomography dataset of a ladybug, deposited on Zenodo, consisting of eight 3D cone-beam sinograms of dimensions 1024 x 360 x 186, acquired with 360 projections using a Procon X-ray CT portable device. The authors specify the measurement geometry (FOD=111 mm, FDD=241 mm, detector height H=96.7 mm, width W=102.4 mm), describe the sinogram organization, and give the linear model Ax=m for reconstruction with toolboxes such as ASTRA and opTomo. The paper also reports acquisition parameters including 50 kV voltage, 400 ms exposure, and a tube current of 400 mA, and states that the sinograms were logarithmically transformed and normalized.","tokens_in":3052,"tokens_out":3403,"duration_ms":32620,"significance":"If the stated geometry is correct, this dataset is a potentially valuable open benchmark for dynamic and sparse-view cone-beam tomography, complementing existing public CT data with a biological specimen and repeat measurements from eight starting positions. The paper makes no derived predictions or fitted parameters, so circularity is not an issue; its value lies entirely in the released data and its documentation. The explicit file naming, stated array dimensions, and the intended use with standard toolboxes are strengths. However, the dataset's usability for quantitative reconstruction hinges on the accuracy and unambiguous reporting of the acquisition geometry, and the current manuscript contains a geometry inconsistency and an implausible tube-current value that must be corrected before the dataset can be used reliably.","major_comments":[{"comment":"The detector pixel size is stated as 48 µm, but the detector dimensions given in Figure 7 (W=102.4 mm, H=96.7 mm) imply a pixel pitch of about 100 µm: 1024 columns × 100 µm = 102.4 mm and 967 rows × 100 µm = 96.7 mm. With the stated 48 µm pixel size, 1024 columns correspond to 49.15 mm, a factor of two discrepancy. If the 48 µm value is the effective pixel size in the object plane rather than on the detector, the paper should say so explicitly and give the detector pixel pitch (approximately 100 µm) that users must enter into ASTRA or opTomo. Without this clarification, the geometry required for quantitative reconstruction from the 1024-column sinograms is ambiguous.","section":"Section 3, Figure 7"},{"comment":"The reported tube current of 400 mA at 50 kV implies an electrical input power of 20 kW, which is inconsistent with the stated 50 W maximum tube output. This is likely a typographical error (for instance, 400 µA), but as written it reduces confidence in the reliability of the entire parameter list. The authors should correct the value and verify that all acquisition parameters are mutually consistent.","section":"Section 3, exposure parameters"},{"comment":"The paper provides no validation or calibration measurements, such as a scan of a known phantom or a reconstruction quality assessment, that would confirm the stated geometry. For a dataset intended to support quantitative tomographic reconstruction, at least one reconstruction performed with the reported geometry should be shown, along with a quantitative measure of data consistency or a calibration scan. This would help resolve the pixel-size ambiguity and demonstrate that the released sinograms can be accurately reconstructed with the documented parameters.","section":"Sections 2 and 3"}],"minor_comments":[{"comment":"The caption contains the informal phrase \"Please mention\" and should be rewritten as an indicative sentence, for example: \"These reconstructions are shown only for visualization of the different starting positions.\"","section":"Figure 2 caption"},{"comment":"The abstract says the data is \"available here\" without a visible hyperlink; the full Zenodo URL should be given in the text.","section":"Abstract"},{"comment":"The grammar in \"It can be also possible that the scanning angle is limited\" is awkward; consider rephrasing to \"It is also possible that the scanning angle is limited.\"","section":"Section 1, last paragraph"},{"comment":"The paper does not state whether dark-field and flat-field corrections were applied before the logarithmic transform and normalization; the normalization procedure should be described explicitly so that users can interpret the sinogram values.","section":"Section 3"},{"comment":"The dimensions in Figure 7 use a comma as a decimal separator (96,7 mm, 102,4 mm), which is inconsistent with the decimal point used in the text; please use a uniform notation.","section":"Figure 7"},{"comment":"Reference [4] has an extra comma between the author list and \"and J. Sijbers\"; this should be cleaned up.","section":"References"}],"recommendation":"major_revision","confidential_remarks":"This manuscript is a data descriptor rather than a full research article, which is appropriate for the dataset it documents. The geometry inconsistency regarding the pixel size is the main substantive issue; it appears to be a correctable error rather than a fundamental flaw. The implausible tube current also needs a simple correction. If the authors provide a corrected parameter list and a basic validation reconstruction, the dataset would become a solid public benchmark. I do not see grounds for rejection, but the current version cannot be used as-is for quantitative reconstruction, so a major revision is warranted."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"The one thing you should know: this is a data descriptor for eight 3D cone-beam CT sinograms of a ladybug, openly archived on Zenodo. That is a real resource. Dynamic CT benchmarks with measured data are scarce, and having eight time-indexed sinograms of the same object with 360 projections each gives algorithm developers something concrete to test on. The paper is short, plain, and gives file dimensions, angle step, and a pointer to ASTRA/opTomo. It does not pretend to be a theory paper. For what it is, it is fine in intent.\n\nThe problem is that the one number users need most, the detector geometry, is internally inconsistent. The text says the detector pixel size is 48 µm and the images are 1024×967. Figure 7 gives detector width W = 102.4 mm and height H = 96.7 mm. Do the arithmetic: 1024×48 µm = 49.2 mm, not 102.4 mm; 967×48 µm = 46.4 mm, not 96.7 mm. The figure's dimensions match a 100 µm pixel pitch exactly. So either the pixel size or the figure is wrong, and a user who follows the text will enter the wrong geometry into ASTRA and get a volume scaled by about a factor of two. The paper also reports 50 kV, 400 mA with a 50 W maximum tube output, which is physically impossible (that would be 20 kW). These are not minor typos in a footnote; they sit in the specification section that the paper itself says is needed to build the measurement matrices.\n\nI also note there is no calibration scan, no validation object, and no quantitative check that the geometry is correct. A single 2D slice for visualization is not validation. None of this means the dataset is fake or useless; it probably contains the eight sinograms it claims. But the current documentation does not meet the standard for a usable benchmark. The core claim of the paper is not just that files exist, but that they can be freely used for quantitative reconstruction. That claim is not supported until the contradiction is resolved.\n\nWho this is for: people working on dynamic or limited-data cone-beam CT who need measured test data. As a data note, it deserves peer review, but only conditionally. A serious referee would catch the geometry issue in ten minutes and ask for a corrected specification plus a simple validation, for example a reconstruction of a known phantom or at least a consistent example showing the declared geometry reproduces a clean image. I would not cite it in its current form, and I would not bring it to reading group until the numbers are cleaned up. If the authors fix the spec, it becomes a modest but legitimate contribution.\n\nRecommendation: send to peer review, but with clear expectation that the geometry and tube-current errors are corrected before publication.","headline":"A genuinely open dynamic CT dataset that is useful as a benchmark, but its documentation has a self-contradictory geometry section that must be corrected before anyone can use the data quantitatively.","tokens_in":3516,"tokens_out":2653,"would_cite":false,"duration_ms":28864,"reading_group":"no","serious_thinker":"no","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":["87.57.Q-"],"model":"deepseek-v4-flash","headline":"This paper documents an open 3D cone-beam CT dataset of a ladybug: eight 360-projection sinograms with shared acquisition geometry, intended for dynamic tomography research.","keywords":["computed tomography","cone-beam CT","dynamic tomography","open data","sinogram","ladybug","X-ray imaging","3D reconstruction"],"falsifier":"Reconstruct all eight sinograms with filtered backprojection using the stated geometry and compare the recovered physical size of the ladybug's shell with the specimen's known size; a clear scale mismatch would show that the documented FOD, FDD, or pixel size is incorrect.","tokens_in":2696,"feed_emoji":"🐞","tokens_out":9681,"duration_ms":82452,"temperature":0.7,"pith_summary":"This documentation paper presents an open 3D dynamic X-ray computed tomography dataset of a ladybug. The dataset consists of eight sinogram files, each a cone-beam sinogram of size $1024\\times360\\times186$, measured with 360 projections and the same angular steps but different starting positions. The authors' aim is to give researchers a real biological test object for three-dimensional CT reconstruction, including dynamic and limited-data settings. If the data and geometry are as described, the set provides a public benchmark on which reconstruction algorithms can be compared directly.","feed_headline":"Eight 3D X-ray scans of a ladybug are now open data","feed_subtitle":"Each scan holds 360 projections and shared geometry, letting dynamic 3D reconstruction be tested on real data.","key_machinery":"The central object is the three-dimensional cone-beam sinogram array: for each scan, a $1024\\times360\\times186$ matrix whose dimensions record detector width, number of projection angles, and selected central detector rows. It carries the argument because every reconstruction depends on this arrangement, on the log-normalization, and on the stated geometry (FOD, FDD, pixel size, row selection). The eight arrays, all taken with the same angular sampling from different starting positions, are what let a user build either individual measurement matrices or a stacked dynamic model.","core_discovery":"The paper's central claim is that the eight released files sinogram1.mat through sinogram8.mat each hold a log-normalized 3D cone-beam sinogram of a ladybug, acquired at 50 kV with 360 projections, a focus-to-object distance of 111 mm, a focus-to-detector distance of 241 mm, and 48 µm detector pixels, and that only the central 186 of the 967 detector rows were kept. These eight measurements of the same object at different starting positions are intended to be enough to assemble a dynamic CT model $Ax=m$, or a stacked dynamic version, using standard tomography toolboxes. The documentation is thus the specification that makes the data usable for quantitative reconstruction.","pith_inferences":["A natural extension not stated in the paper: the eight scans of the same specimen can also serve as a repeatability set for registration and alignment methods, since scan-to-scan variation includes real positioning differences.","Because the sinograms keep only 186 of 967 detector rows, an indirect test of the data is to reconstruct with two independent toolboxes and compare the central slices; large discrepancies would point to a row-selection or geometry mismatch.","One could reinterpret the eight starting positions as quasi-static frames of an object being reoriented, giving a surrogate time-lapse sequence for benchmarking motion-aware reconstruction, though the paper does not claim the ladybug moved during a single scan."],"forward_implications":["Each sinogram can be used as the right-hand side of the CT model $Ax=m$, giving a 3D reconstruction from 360 cone-beam projections.","Taking subsets of the 360 projections yields sparse-view data, which the paper notes is useful for studying reduced radiation dose and measurement time.","Because all eight scans use the same projection angles, the data can be stacked into a dynamic system with eight entries, following the paper's dynamic-model suggestion.","The explicit geometry and detector-row selection allow the same data to be reconstructed with different toolboxes, making the set a shared comparison point."],"supporting_citations":[{"why":"Supplies the tomography platform the paper says can build the 3D measurement matrix once the geometry is specified.","marker":"[4]"},{"why":"Companion toolbox reference for constructing measurement matrices and reconstructions from the sinograms.","marker":"[5]"},{"why":"The space-time tomography formulation cited for stacking measurement matrices into a dynamic model.","marker":"[9]"}],"fun_headline_variants":["Ladybug CT dataset: 8 dynamic 3D scans open to all","Open 3D X-ray benchmark: ladybug in 8 dynamic scans","Eight ladybug CT scans set stage for dynamic tomography","New open dataset: 3D+time ladybug X-ray for reconstruction","Ladybug tomographic scans: free data for dynamic CT tests"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The load-bearing premise is that the written geometry (FOD=111 mm, FDD=241 mm, 48 µm pixel size, central 186 of 967 rows) is the true acquisition geometry, because every quantitative reconstruction from the data depends on these numbers.","fun_headline_variants_meta":{"raw":{"variants":["Ladybug CT dataset: 8 dynamic 3D scans open to all","Open 3D X-ray benchmark: ladybug in 8 dynamic scans","Eight ladybug CT scans set stage for dynamic tomography","New open dataset: 3D+time ladybug X-ray for reconstruction","Ladybug tomographic scans: free data for dynamic CT tests"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000279,"raw_usage":{"total_tokens":1551,"prompt_tokens":733,"completion_tokens":818,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":349,"completion_tokens_details":{"reasoning_tokens":725}},"tokens_in":349,"tokens_out":818,"duration_ms":6726,"temperature":1.0,"reasoning_tokens":725,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-14T11:28:31.600940+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Reconstruct all eight sinograms with filtered backprojection using the stated geometry and compare the recovered physical size of the ladybug's shell with the specimen's known size; a clear scale mismatch would show that the documented FOD, FDD, or pixel size is incorrect.","supporting_citations":[{"cited_title":"The astra toolbox: A platform for advanced algo- rithm development in electron tomogra- phy,","cited_arxiv_id":null,"evidence_quote":"Supplies the tomography platform the paper says can build the 3D measurement matrix once the geometry is specified."},{"cited_title":"Fast and flexible x-ray tomography using the astra toolbox,","cited_arxiv_id":null,"evidence_quote":"Companion toolbox reference for constructing measurement matrices and reconstructions from the sinograms."},{"cited_title":"Space-time tomography for continu- ously deforming objects,","cited_arxiv_id":null,"evidence_quote":"The space-time tomography formulation cited for stacking measurement matrices into a dynamic model."}],"review_version":1}