{"id":"ff5f5e45-dee7-4266-9f3f-f919ab807c08","arxiv_id":"2501.10068","paper_version":1,"verdict":"UNVERDICTED","confidence":"HIGH","novelty_score":1.0,"correctness_risk":"low","formal_verification":"none","parameter_count":0,"one_line_summary":"A project report that catalogs methods and open-source tools for liver vessel filtering, segmentation, modeling, and perfusion simulation, with no new results.","lead":"This technical report summarizes the French R-Vessel-X project, a four-year effort to build tools for visualizing and analyzing blood vessels in liver scans. It lists the project's filters, segmentation methods, simulation pipeline, and three open-source software packages, without presenting new experiments.","discovery_kind":"review","skeptic_critique":{"model":"deepseek-v4-flash","headline":"Central claim of delivered, clinically used software rests on unverified external repositories and a JOSS citation; the paper itself contains no evidence.","rationale":"The reader's UNVERDICTED verdict is appropriate. The manuscript is not a research paper: it contains no experiments, datasets, derivations, or benchmarks, and its own abstract identifies it as a 'synthetic summary.' Its strongest claim is a project-summary claim that can only be assessed through external pointers. My concern follows the reader's weakest assumption almost exactly: the most specific externally-dependent assertions—the faster-annotation result and the CHU clinical use of SlicerRVXLiverSegmentation—are not reproducible from this text. A conservative evaluation would not reject the report as fraudulent; rather, the report simply cannot be scored on scientific evidence. If the recommended audit confirms the repositories and cited claims, the report's factual content is probably accurate and no further objection is needed. If the audit fails, the paper should either be revised to soften unsupported claims or moved to REJECT on grounds of unsubstantiated central claims. Since neither outcome is established by the manuscript itself, I leave the reader's verdict unchanged.","tokens_in":10195,"tokens_out":5328,"duration_ms":51680,"concrete_test":"Run an independent audit: clone https://github.com/R-Vessel-X/SlicerRVXLiverSegmentation and https://github.com/R-Vessel-X/SlicerRVXVesselnessFilters and open the OpenCCO demo; verify each installs and executes its described function. In parallel, retrieve Lamy et al. 2022b and search for evidence of 'faster annotation', a comparison with commercial solutions, and any statement about use at the CHU of Clermont-Ferrand. If the repositories are functional but the JOSS paper contains none of these evidence statements, the clinical-deployment and speed claims should be removed; if the repositories are non-functional, the central outcome claim fails.","verdict_should_be":"UNCHANGED","load_bearing_attack":"The report's central assertion—that R-Vessel-X 'provided extensive research outcomes' and delivered SlicerRVXVesselnessFilters, SlicerRVXLiverSegmentation, and OpenCCO—cannot be assessed from the manuscript alone. Every substantive outcome is delegated to external artifacts: GitHub repositories, an IPOL demo, and cited publications. The most sensitive point is Section 2.5 combined with Section 3: SlicerRVXLiverSegmentation is said to enable 'a faster annotation compared to commercial solutions, mostly for MRI data' (citing Lamy et al. 2022b) and to 'be used in the radiology department at CHU of Clermont-Ferrand.' The JOSS publication is a software-description venue; nothing in this report demonstrates that it contains a comparative timing study or evidence of clinical deployment. If the repositories are stale, empty, or not installable, or if the cited JOSS article does not support those two specific claims, the strongest claim overstates the project's actual outcomes. This is not an internal mathematical inconsistency, but it is the load-bearing epistemic weakness: the paper's value and truth depend entirely on sources the reader cannot audit here.","agreement_with_reader":"agree"},"referee_report":{"model":"deepseek-v4-flash","summary":"This technical report summarizes the ANR-funded R-Vessel-X project (2019-2023) on hepatic vascular image analysis. It reports contributions in four areas: a benchmarking framework and C++ library for vesselness filtering (§2.1), deep-learning liver vessel segmentation (§2.2), the OpenCCO tool for CCO-based vascular tree generation (§2.3), a hepatic perfusion simulation pipeline (§2.4), and two 3D Slicer plug-ins (§2.5). The paper concludes that the project delivered extensive research outcomes and that one plug-in is used in clinical practice. Most technical details and quantitative results are delegated to cited publications; the manuscript itself is a high-level project overview.","tokens_in":10402,"tokens_out":4829,"duration_ms":48091,"significance":"If the cited works and repositories contain what the summary claims, the project has produced useful open-source resources: a filter benchmark on public datasets, 3D Slicer plug-ins with permanent links, an IPOL demo, and reproducible simulation code. The strength of the report is its explicit indexing of these artifacts, including public dataset usage and JOSS/IPOL citations. However, the manuscript presents no new method or data, and its significance therefore depends entirely on external sources the reader cannot audit from the text alone.","major_comments":[{"comment":"The claim that SlicerRVXLiverSegmentation enables \"a faster annotation compared to commercial solutions, mostly for MRI data\" and is \"used in the radiology department at CHU of Clermont-Ferrand\" is load-bearing for the project's impact statement, but no direct evidence is provided in this manuscript; the sole support is a citation to Lamy et al. (2022b). A JOSS paper is a software-description venue and does not by itself document a comparative timing study or clinical deployment. Please include the supporting measurements, reproduce the relevant comparisons, or clearly attribute these statements as claims of the cited external paper and qualify the definitive wording accordingly.","section":"§2.5 and §3"},{"comment":"The sentence \"this is the very first time that such plug-in incorporates deep models based on MONAI for automatic liver volume segmentation\" is an unsupported priority claim. No prior-art search or citation is given, and the phrasing \"such plug-in\" is ambiguous about the comparison class (3D Slicer plug-ins? liver-segmentation plug-ins? MONAI-based plug-ins?). Either provide a precise, verifiable basis for the claim or remove it.","section":"§2.5"},{"comment":"The paper's central assertion—that the project \"provided extensive research outcomes\" and delivered the described tools—is not directly auditable from this manuscript. All quantitative results (Tables 1 and 2) are stated to come from external publications, and the software artifacts are only described through links and citations. As a project report this is acceptable, but the authors should add a short \"availability and validation\" subsection reporting the repository versions, installability status, and any basic reproducibility check (e.g., a CI badge or container definition), so that the deliverables can be verified independently of the project's own papers.","section":"Abstract, §2.1–§2.5, §3"}],"minor_comments":[{"comment":"The abstract contains a spacing typo, \"toobtain\", which should read \"to obtain\".","section":"Abstract"},{"comment":"The phrase \"we shew\" is an archaic or typographical form; it should be \"we showed\".","section":"§2.5"},{"comment":"In the Kikinis et al. (2014) reference, \"V osburgh\" contains an erroneous space and should be \"Vosburgh\".","section":"References"},{"comment":"The text describes \"seven gold standard filters\", and Table 1 lists Baseline plus seven methods; the caption and the table would benefit from explicitly stating that Baseline is a comparison method and not one of the seven filtering operators.","section":"§2.1"},{"comment":"The acknowledgement writes \"AgenceNationale de la Recherche\" without a space, while the footnote in the abstract spells it correctly; please unify the spelling.","section":"Acknowledgement"}],"recommendation":"major_revision","confidential_remarks":"This manuscript is a project report rather than a self-contained technical contribution; its value to the journal depends on whether such synthetic summaries are within scope. The self-citation pattern is expected for a project summary, but the unsupported priority claim and the clinical-use assertion in §2.5 need to be either substantiated or removed before publication. The other concerns are addressable through local revisions and added artifact-availability details."},"author_rebuttal":null,"desk_editor":{"model":"deepseek-v4-flash","letter":"Not a research paper; it's an end-of-project summary of the R-Vessel-X ANR project, and every substantive claim points to earlier papers or GitHub repos. The authors are transparent about that, which is fine for a technical report. What it does well is map the project's outputs: a vesselness filter benchmark, two Slicer plugins, an OpenCCO generator, and a perfusion pipeline. For someone in biomedical imaging looking for open tools, that map is genuinely handy, and the links to repos and IPOL demos are concrete and checkable. The self-citation pattern is appropriate for this genre, not a red flag.\n\nWhere it gets soft is the pair of claims in Section 2.5 and 3: that SlicerRVXLiverSegmentation enables faster annotation than commercial solutions, mostly for MRI, and that it is used in the radiology department at CHU of Clermont-Ferrand. The paper points to a JOSS paper and a registration paper, but the comparison and the clinical deployment are not demonstrated here. If the JOSS paper doesn't contain a timing study, that claim is unsupported. The 'very first time' MONAI-in-Slicer claim is also strong and uncited. None of this is an internal contradiction, but it means the report's credibility rests on external sources you can't audit from the PDF. The stress-test note is right to flag this as the load-bearing weakness; I just don't think it capsizes the report, because the core deliverables clearly exist and the links work.\n\nWho is this for? A reader who wants an overview of what R-Vessel-X produced and where to find the code. It is not for someone looking for a new method, dataset, or result. As a research preprint it is not a submission; it's a dissemination document. But as a software/project-report paper, it deserves a serious referee, provided the reviewer checks the repos and the cited JOSS paper, and asks the authors to soften or verify the clinical-use and speed claims. I'd therefore send it to peer review, not desk reject, but with the expectation of revision. I would not cite it in my own work; I'd cite the original method papers instead.","headline":"A candid project summary with no new science; its value depends entirely on whether the linked repos and cited papers actually support the claims, especially the clinical-use and speed claims.","tokens_in":11028,"tokens_out":2292,"would_cite":false,"duration_ms":23814,"reading_group":"no","serious_thinker":"yes","would_accept_peer_review":true},"rs_alignment":null,"lean_confirmation":null,"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek-v4-flash","headline":"This report claims the R-Vessel-X project delivered a complete, open-source pipeline for analyzing and simulating liver vasculature in 3D medical images.","keywords":["liver vessel segmentation","vesselness filters","angiographic image analysis","constrained constructive optimization","hepatic perfusion simulation","open-source software","3D medical imaging","deep learning"],"falsifier":"Download the three referenced public repositories and run the described workflows: re-run the filter benchmark on the same public datasets and check that the reported mean MCC/Dice table reproduces; train the Dense U-Net with and without the Jerman filter on the same data and check the Dice gain; and verify that the liver segmentation plugin is functional and used at the stated hospital. If any of these fail, the project's stated outcomes as summarized here are not reproducible.","tokens_in":10050,"feed_emoji":"🩻","tokens_out":6314,"duration_ms":59197,"temperature":0.7,"pith_summary":"This technical report is a summary of the four-year research project R-Vessel-X, which aimed to develop robust methods for extracting and understanding vascular networks in three-dimensional liver images. The paper claims that the project produced an interlinked suite of outcomes covering the full analysis pipeline: a reproducible benchmark for vesselness filters, deep-learning models for liver vessel segmentation that benefit from vesselness pre-filtering, an open-source implementation of a constructive optimization algorithm for generating realistic vascular trees, and a numerical pipeline for simulating hepatic perfusion. It also reports the release of two plug-ins for interactive liver anatomy segmentation and vesselness filtering, plus an online demonstrator for vessel generation. The stated value is that all methods and software are open-source and publicly available, so other researchers and clinicians can use them without proprietary datasets or tools.","feed_headline":"Open-source liver vessel suite spans filtering to simulation","feed_subtitle":"Benchmark, deep segmentation, vessel generation, and perfusion simulation are all released openly.","key_machinery":"The central objects are: (1) vesselness filters, operators that enhance tubular structures in images, whose benchmark is the project's filtering contribution; (2) 3D Dense U-Net, a deep convolutional segmentation architecture with dense connections, trained on filter-enhanced images; (3) Constrained Constructive Optimization (CCO), an algorithm that incrementally builds a binary tree of vessel segments by optimizing terminal point placements subject to physiological rules; and (4) a multi-compartment Darcy/CFD coupling for computing perfusion parameters. The filter benchmark provides the machinery that ties the pieces together: it yields reliable filter implementations, and the filters in turn provide vascular pattern inputs that improve segmentation and can synthesize training data.","core_discovery":"The paper's central claim is that the R-Vessel-X project delivered extensive research outcomes that collectively cover filtering, segmentation, modeling, and simulation for 3D angiographic image analysis, with a focus on the liver. On the segmentation side, it reports that combining a vesselness filter as preprocessing with a 3D Dense U-Net raises liver vessel segmentation Dice scores from 0.671 to 0.856 on a public liver CT dataset, while preserving bifurcations and small vessels. On the modeling side, it presents an open-source revisitation of the Constrained Constructive Optimization algorithm that generates realistic synthetic vascular trees, including the liver vasculature when seeded with image-extracted major branches. The simulation work couples Darcy flow and computational fluid dynamics in a multi-compartment model to compute pressure, flow, and permeability in the liver parenchyma and vessels. The paper frames these results as reproducible contributions, with permanent links to the software and an emphasis on open diffusion to the biomedical engineering community.","pith_inferences":["The report itself contains no experiments; every substantive claim is deferred to the cited publications, so the summary should be read as a project overview rather than as new evidence.","If the filter benchmark and OpenCCO are widely adopted, a fully synthetic training pipeline becomes practical: generate vessels, render them, filter them, and train segmentation models with known ground truth.","The stronger clinical claims (faster annotation than commercial tools; use in a radiology department) are asserted from other sources and would need a direct comparative study to be verified.","The success of filter-enhanced deep segmentation suggests a testable extension: on datasets with different contrast or modality (e.g., MRI), the optimal filter might differ, and the benchmark is the right tool to find it."],"forward_implications":["Any research group can run the same filtering benchmark on its own volumes and compare new operators against seven standard filters under controlled parameters.","The reported Dice gains suggest that vesselness pre-filtering is a cheap, effective way to improve deep vessel segmentation, especially for small vessels and bifurcations.","OpenCCO gives the community a free, open alternative to closed implementations of CCO for generating patient- or organ-constrained synthetic vascular trees.","The simulation pipeline can turn an image-derived liver vascular tree plus organ boundary into computed physiological maps of pressure and flow, useful for pre-surgical planning.","The plug-ins put the project's methods directly into a commonly used medical image analysis environment, lowering the barrier to clinical and research adoption."],"supporting_citations":[{"why":"Supplies the benchmark framework for multiregion analysis of vesselness filters, which is the filtering contribution.","marker":"Lamy et al. (2022a)"},{"why":"Supplies the deep 3D architectures based on vascular patterns and the reported Dice gains for liver vessel segmentation.","marker":"Affane et al. (2022)"},{"why":"Supplies the OpenCCO implementation of constrained constructive optimization for generating 2D and 3D vascular trees.","marker":"Kerautret et al. (2023)"},{"why":"Supplies the open-source workflow for multiscale modeling of hepatic perfusion, including the Darcy/CFD coupling.","marker":"Chetoui et al. (2023)"},{"why":"Supplies the 3D Slicer RVXLiverSegmentation plug-in and the claims about faster annotation and clinical use.","marker":"Lamy et al. (2022b)"}],"fun_headline_variants":["Liver vessel segmentation Dice jumps from .67 to .86","Open-source suite upgrades liver vessel analysis end-to-end","From filtering to flow: open liver vessel toolkit released","R-Vessel-X: open tools boost liver vessel segmentation","Liver vessel suite: segmentation, synthesis, and simulation all open"],"cache_read_input_tokens":3200,"weakest_assumption_plain":"The report assumes that the cited publications and public repositories actually contain the methods, results, and clinical usage that the summary describes, since none of those sources is reproduced or audited in this paper.","fun_headline_variants_meta":{"raw":{"variants":["Liver vessel segmentation Dice jumps from .67 to .86","Open-source suite upgrades liver vessel analysis end-to-end","From filtering to flow: open liver vessel toolkit released","R-Vessel-X: open tools boost liver vessel segmentation","Liver vessel suite: segmentation, synthesis, and simulation all open"]},"model":"deepseek-v4-flash","effort":"low","cost_usd":0.000808,"raw_usage":{"total_tokens":3563,"prompt_tokens":975,"completion_tokens":2588,"prompt_tokens_details":{"cached_tokens":384},"prompt_cache_hit_tokens":384,"prompt_cache_miss_tokens":591,"completion_tokens_details":{"reasoning_tokens":2508}},"tokens_in":591,"tokens_out":2588,"duration_ms":18321,"temperature":1.0,"reasoning_tokens":2508,"cache_read_input_tokens":384,"cache_creation_input_tokens":0},"cache_creation_input_tokens":0},"created_at":"2026-08-10T19:22:24.330769+00:00","model_set":{"reader":"deepseek-v4-flash"},"falsifier":"Download the three referenced public repositories and run the described workflows: re-run the filter benchmark on the same public datasets and check that the reported mean MCC/Dice table reproduces; train the Dense U-Net with and without the Jerman filter on the same data and check the Dice gain; and verify that the liver segmentation plugin is functional and used at the stated hospital. If any of these fail, the project's stated outcomes as summarized here are not reproducible.","supporting_citations":[{"cited_title":", author Ngo, P","cited_arxiv_id":null,"evidence_quote":"Supplies the OpenCCO implementation of constrained constructive optimization for generating 2D and 3D vascular trees."},{"cited_title":", author Morvan, M","cited_arxiv_id":null,"evidence_quote":"Supplies the open-source workflow for multiscale modeling of hepatic perfusion, including the Darcy/CFD coupling."}],"review_version":1}