arxiv: 2604.06671 · v1 · submitted 2026-04-08 · 📡 eess.IV · cs.CV· physics.med-ph

Recognition: no theorem link

4D Vessel Reconstruction for Benchtop Thrombectomy Analysis

Ethan Nguyen , Javier Carmona , Arisa Matsuzaki , Naoki Kaneko , Katsushi Arisaka

Authors on Pith no claims yet

Pith reviewed 2026-05-10 18:11 UTC · model grok-4.3

classification 📡 eess.IV cs.CVphysics.med-ph

keywords thrombectomy4D reconstructionGaussian Splattingvessel kinematicsbenchtop phantomstress proxymulti-view imagingsilicone model

0 comments

The pith

A nine-camera multi-view workflow with 4D Gaussian Splatting delivers standardized time-resolved vessel surface kinematics and comparative stress proxies for thrombectomy benchtop studies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper introduces a protocol that records thrombectomy procedures on silicone middle cerebral artery phantoms using nine synchronized cameras at 2160p and 20 fps, then reconstructs the changing 3D surface with 4D Gaussian Splatting. Reconstructed point clouds are turned into fixed edge graphs so that region displacements can be tracked and a relative surface stress proxy can be computed from edge stretches via a Neo-Hookean mapping. This matters for benchtop device testing because prior methods have supplied only limited or qualitative motion information, making it difficult to compare how different catheter placements affect vessel deformation in a consistent way. If the protocol performs as described, researchers gain repeatable quantitative metrics that support direct comparisons across experimental conditions without claiming absolute wall-stress values. Validation on synthetic deformations with known ground truth precedes preliminary real-phantom trials that already show measurable differences between placement sites.

Core claim

The paper establishes that its nine-camera multi-view imaging and 4D Gaussian Splatting reconstruction protocol provides standardized, time-resolved surface kinematics together with comparative relative displacement and stress-proxy measurements suitable for thrombectomy benchtop studies. In synthetic bulk-translation tests the stress proxy stays near zero for most edges, while synthetic pulling deformations of 1-5 mm produce Chamfer distances of 1.714-1.815 mm and high precision at the 1 mm threshold. Early benchtop trials indicate higher maximum-median ROI displacement and stress-proxy values for cervical aspiration-catheter placement than for internal-carotid-artery-terminus placement.

What carries the argument

4D Gaussian Splatting reconstruction of calibrated multi-view videos into point clouds that are converted to fixed-connectivity edge graphs; the graphs support ROI displacement tracking and Neo-Hookean mapping of edge stretches to produce comparative surface stress proxies.

If this is right

The protocol supports direct quantitative comparisons between different thrombectomy device placements or conditions in the same benchtop setup.
It supplies time-resolved data at 20 frames per second that can be used for temporal analysis of deformation events.
All reported metrics remain comparative proxies rather than absolute wall-stress estimates.
A synthetic Blender validation pipeline confirms geometric and temporal accuracy before real-phantom use.
Open code and example data allow other groups to replicate or extend the measurements.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same multi-view reconstruction approach could be adapted to study deformation in other soft-tissue benchtop models beyond cerebral vessels.
Combining the surface kinematics with computational fluid dynamics simulations might reveal coupled mechanical and flow effects during thrombectomy.
The fixed edge-graph representation could facilitate direct comparison with finite-element vessel models for hybrid experimental-computational studies.
Increasing camera count or frame rate could extend the method to faster dynamic events in future device testing.

Load-bearing premise

The 4D Gaussian Splatting reconstruction accurately recovers true surface deformations without substantial multi-view artifacts or segmentation errors, and the Neo-Hookean mapping applied to edge stretches yields a reliable comparative stress proxy.

What would settle it

Independent strain measurements from fiducial markers embedded in the phantom surface under controlled loads that deviate substantially from the edge-stretch values and derived stress proxies obtained by the reconstruction pipeline.

Figures

Figures reproduced from arXiv: 2604.06671 by Arisa Matsuzaki, Ethan Nguyen, Javier Carmona, Katsushi Arisaka, Naoki Kaneko.

**Figure 1.** Figure 1: Overview of the experimental and computational pipeline. Multi-camera chessboard calibration yields camera parameters for nine-view vessel recordings. After SAM2-based segmentation, cropping, and dataset preparation, the deforming vessel is reconstructed with 4D Gaussian Splatting and filtered to obtain a time-varying point cloud. The point cloud is converted into a fixed-connectivity edge graph using clus… view at source ↗

**Figure 2.** Figure 2: Experimental setup and major stages of the reconstruction and analysis pipeline. (A) Nine-camera acquisition rig used for multi-view imaging, shown during chessboard-based calibration; long arrows (shown in blue) indicate the cameras’ viewing directions toward the working volume. (B) Example multi-view raw frames acquired during benchtop thrombectomy imaging under UV illumination, showing the fluorescent-b… view at source ↗

**Figure 3.** Figure 3: Synthetic validation workflow (schematic; no single frame/timepoint metric shown). Ground-truth vessel geometries were deformed in Blender using controlled bulk-translation and localized-pulling conditions, rendered with the same camera geometry as the physical setup, and passed through the same camera geometry and downstream reconstruction pipeline, with segmentation bypassed because the renders use trans… view at source ↗

**Figure 4.** Figure 4: Regions of the ICA/MCA surface used for regional aggregation and analysis. R1: M1 mid segment; R2: M1/M2 bifurcation; R3: proximal inferior M2; R4: proximal superior M2; R5: ICA terminus/M1 origin. 2.2 Synthetic validation design and metric definitions Synthetic experiments were used to check geometric and temporal behavior against known motion before interpreting benchtop comparisons. To enable quantitati… view at source ↗

**Figure 5.** Figure 5: Rigid-body translation control (synthetic vessel), evaluated at the final frame of the bulk-translation sequence. (A) Displacement magnitude (mm) for ground truth (GT) and reconstruction (Ours). (B) Edge-based stress-proxy magnitude |σe| (MPa); GT remains near zero under rigid motion, and Ours is near zero for most edges with sparse outliers. Synthetic deformation fidelity Under localized pulling, reconstr… view at source ↗

**Figure 6.** Figure 6: Synthetic pulling deformation (1–5 mm), evaluated at each condition’s final frame. Rows compare ground truth (GT) and reconstruction (Ours). (A) Displacement magnitude (mm). (B) Edge-based stress-proxy magnitude |σe| (MPa). Benchtop comparative application Using the same validated metrics, the benchtop comparison (one trial per condition) showed broader and larger displacement/stress-proxy fields for AC ce… view at source ↗

**Figure 7.** Figure 7: GT-versus-reconstruction agreement across synthetic pull magnitudes (condition-level comparison; not a single-frame plot). Pairwise heatmaps report (A) Chamfer distance (CD, mm), (B) temporal Chamfer delta (∆CD, mm), and (C) F-score; metric definitions follow Sec. 2.2 (Eqs. 12, 16, and 19). Best agreement occurs along the diagonal (matched pull conditions) [PITH_FULL_IMAGE:figures/full_fig_p010_7.png] view at source ↗

**Figure 8.** Figure 8: Agreement analyses for max-median ROI metrics between reconstruction and ground truth (condition-level summaries, not single-frame values): correlation and Bland–Altman for displacement magnitude (A–B, mm) and relative surface-based stress-proxy magnitude (C–D, MPa). A Cervical Terminal 3cm B Cervical Terminal [PITH_FULL_IMAGE:figures/full_fig_p011_8.png] view at source ↗

**Figure 9.** Figure 9: Benchtop application demonstration at peak deformation for a single representative trial per AC placement condition (cervical and terminal). (A) Reconstructed displacement magnitude (mm) relative to the initial frame. (B) Corresponding edge-based stress-proxy magnitude |σe| (MPa) at the same peak-deformation timepoint. 11 [PITH_FULL_IMAGE:figures/full_fig_p011_9.png] view at source ↗

**Figure 10.** Figure 10: Regional max-median displacement and stress-proxy during thrombectomy for a single representative trial per AC placement condition. Bars report the maximum over time of the regional median displacement magnitude (A, mm) and edge-based stress-proxy magnitude (B, MPa) across distal M1 segment, MCA bifurcation (M1/M2), and inferior M2 division. 4 Discussion This methods/validation study defines and tests a m… view at source ↗

read the original abstract

Introduction: Mechanical thrombectomy can cause vessel deformation and procedure-related injury. Benchtop models are widely used for device testing, but time-resolved, full-field 3D vessel-motion measurements remain limited. Methods: We developed a nine-camera, low-cost multi-view workflow for benchtop thrombectomy in silicone middle cerebral artery phantoms (2160p, 20 fps). Multi-view videos were calibrated, segmented, and reconstructed with 4D Gaussian Splatting. Reconstructed point clouds were converted to fixed-connectivity edge graphs for region-of-interest (ROI) displacement tracking and a relative surface-based stress proxy. Stress-proxy values were derived from edge stretch using a Neo-Hookean mapping and reported as comparative surface metrics. A synthetic Blender pipeline with known deformation provided geometric and temporal validation. Results: In synthetic bulk translation, the stress proxy remained near zero for most edges (median $\approx$ 0 MPa; 90th percentile 0.028 MPa), with sparse outliers. In synthetic pulling (1-5 mm), reconstruction showed close geometric and temporal agreement with ground truth, with symmetric Chamfer distance of 1.714-1.815 mm and precision of 0.964-0.972 at $\tau = 1$ mm. In preliminary benchtop comparative trials (one trial per condition), cervical aspiration catheter placement showed higher max-median ROI displacement and stress-proxy values than internal carotid artery terminus placement. Conclusion: The proposed protocol provides standardized, time-resolved surface kinematics and comparative relative displacement and stress proxy measurements for thrombectomy benchtop studies. The framework supports condition-to-condition comparisons and methods validation, while remaining distinct from absolute wall-stress estimation. Implementation code and example data are available at https://ethanuser.github.io/vessel4D

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper gives a practical nine-camera 4D Gaussian Splatting pipeline for tracking vessel phantom motion with solid synthetic checks, but single-trial real runs leave the condition comparisons on shaky ground.

read the letter

The core of this paper is a pipeline that captures multi-view video of deforming silicone vessels with nine cameras, reconstructs them in 4D with Gaussian Splatting, and then tracks ROI displacements plus a relative stress proxy from edge stretches. They did solid work on the synthetic side. The Blender validation gives concrete numbers on geometric and temporal accuracy, and the code is public. That part is reproducible and useful. The nine-camera low-cost approach is practical for a lab setting. The real experiments are preliminary though. One trial per condition means we can't tell how much the reported differences between catheter placements come from the actual setup versus run-to-run variation or reconstruction artifacts. The stress proxy is comparative and based on a standard material model without specific validation for their phantoms, so it stays in the relative category as they note. The synthetic tests don't fully simulate the lighting and catheter interaction in the real runs. This is for researchers running benchtop thrombectomy models who want time-resolved 3D surface data. It gives a concrete workflow and validation that others can build on. The thinking is straightforward and the methods are detailed enough that it should go to peer review. I'd accept it for review but expect the authors to add repeatability tests and perhaps some checks on how sensitive the proxy is to the material constants.

Referee Report

2 major / 2 minor

Summary. The paper introduces a nine-camera multi-view workflow for time-resolved 3D reconstruction of silicone vessel phantoms during benchtop thrombectomy using 4D Gaussian Splatting. Reconstructed surfaces are converted to fixed-connectivity edge graphs to track ROI displacements and compute a relative stress proxy from edge stretches via a fixed Neo-Hookean mapping. Synthetic validation with Blender ground truth reports Chamfer distances of 1.714-1.815 mm and precision 0.964-0.972; preliminary real-data trials (one per condition) indicate higher max-median displacement and stress-proxy values for cervical versus ICA-terminus catheter placement. The work provides open code and positions the output as standardized comparative metrics rather than absolute wall stress.

Significance. If the reconstruction fidelity and comparative reliability hold under repeated real acquisitions, the protocol would address a clear gap in benchtop thrombectomy analysis by delivering accessible, time-resolved surface kinematics and relative deformation metrics. The concrete quantitative synthetic validation (Chamfer/precision numbers) and public implementation are notable strengths that support reproducibility and methods validation.

major comments (2)

[Results (benchtop experiments)] Results (benchtop comparative trials): Only a single trial is reported per catheter-placement condition, with no within-condition standard deviations, repeatability metrics, or error propagation from the 4D Gaussian Splatting/edge-graph pipeline. This is load-bearing for the central claim of providing reliable standardized comparative metrics, because the reported elevation in max-median ROI displacement and stress-proxy values for cervical vs. ICA-terminus placement cannot yet be distinguished from reconstruction artifacts, lighting variability, or single-run stochastic effects (synthetic validation does not replicate these real-acquisition factors).
[Methods (stress proxy)] Methods (stress-proxy derivation): The relative surface-based stress proxy is obtained by applying a fixed Neo-Hookean mapping directly to reconstructed edge stretches without sensitivity analysis on the material constants or independent validation against known stress fields. While the paper correctly labels it comparative rather than absolute, the lack of quantification of how constant choice propagates into the reported differences weakens the robustness of condition-to-condition comparisons.

minor comments (2)

[Abstract] Abstract and Conclusion: The phrasing 'provides standardized... measurements' is stronger than the 'preliminary' qualifier and single-trial description used in the real-data results; harmonizing this language would improve consistency.
[Figures] Figure clarity: The edge-graph visualization and ROI definition would benefit from explicit annotation of how fixed connectivity is enforced across time frames to avoid reader confusion about topology changes.

Simulated Author's Rebuttal

2 responses · 1 unresolved

We thank the referee for the constructive review and for recognizing the strengths of our synthetic validation and open-source code. We address each major comment below with honest revisions to the manuscript where feasible.

read point-by-point responses

Referee: [Results (benchtop experiments)] Results (benchtop comparative trials): Only a single trial is reported per catheter-placement condition, with no within-condition standard deviations, repeatability metrics, or error propagation from the 4D Gaussian Splatting/edge-graph pipeline. This is load-bearing for the central claim of providing reliable standardized comparative metrics, because the reported elevation in max-median ROI displacement and stress-proxy values for cervical vs. ICA-terminus placement cannot yet be distinguished from reconstruction artifacts, lighting variability, or single-run stochastic effects (synthetic validation does not replicate these real-acquisition factors).

Authors: We agree that the benchtop results rely on single trials per condition and therefore lack within-condition standard deviations or repeatability metrics. The manuscript already describes these as preliminary trials, with the core contribution being a reproducible protocol for standardized comparative metrics rather than statistically powered condition comparisons. In the revised manuscript we will add an explicit limitations paragraph in the Discussion that (i) enumerates real-acquisition variability sources (reconstruction artifacts, lighting, phantom positioning), (ii) propagates uncertainty bounds derived from the synthetic Chamfer/precision figures to the reported real-data displacements and stress-proxy values, and (iii) states that future multi-trial studies are required to confirm the observed trends. This textual clarification addresses the referee’s concern without overstating current statistical reliability. revision: partial
Referee: [Methods (stress proxy)] Methods (stress-proxy derivation): The relative surface-based stress proxy is obtained by applying a fixed Neo-Hookean mapping directly to reconstructed edge stretches without sensitivity analysis on the material constants or independent validation against known stress fields. While the paper correctly labels it comparative rather than absolute, the lack of quantification of how constant choice propagates into the reported differences weakens the robustness of condition-to-condition comparisons.

Authors: We accept that a sensitivity analysis on the Neo-Hookean parameters would strengthen the presentation. Although the proxy is labeled comparative and not absolute, quantifying parameter influence on the reported differences is warranted. In the revised manuscript we will add a sensitivity study that varies the shear modulus and bulk modulus over a range drawn from silicone-phantom literature, recomputes the max-median stress-proxy values for both catheter-placement conditions, and shows that the relative ordering and magnitude of differences remain stable. The new analysis will be inserted in the Methods section with a brief Results paragraph summarizing the outcome. revision: yes

standing simulated objections not resolved

The absence of multiple independent real-acquisition trials prevents reporting of within-condition standard deviations or formal repeatability statistics for the benchtop comparative results.

Circularity Check

0 steps flagged

No significant circularity in the derivation chain

full rationale

The paper's workflow proceeds from multi-view video acquisition through standard calibration and segmentation, 4D Gaussian Splatting reconstruction, fixed-connectivity edge-graph conversion, ROI displacement tracking, and a direct Neo-Hookean mapping applied to measured edge stretches to obtain the stress proxy. None of these steps reduce to their own inputs by construction, nor are any load-bearing premises justified solely by self-citation. Synthetic validation employs an independent Blender pipeline with prescribed ground-truth deformations, supplying external geometric and temporal benchmarks rather than self-referential checks. The reported comparative benchtop metrics are straightforward outputs of the pipeline; no parameters are fitted to target data and then re-presented as predictions. The derivation therefore remains self-contained against the stated external validation.

Axiom & Free-Parameter Ledger

1 free parameters · 2 axioms · 1 invented entities

The method depends on standard computer-vision calibration axioms and a domain assumption that the silicone material stretch can be mapped to a comparative stress proxy via Neo-Hookean elasticity; no new physical entities are introduced beyond the defined proxy metric.

free parameters (1)

Neo-Hookean material constants
Constants used in the stretch-to-stress-proxy mapping; values are not reported as fitted to the experimental data in the abstract.

axioms (2)

standard math Multi-view camera calibration accurately recovers 3D positions from synchronized 2D images.
Invoked for the initial reconstruction step in the nine-camera workflow.
domain assumption Silicone phantom deformation can be approximated by a Neo-Hookean hyperelastic model for relative stress comparison.
Used to convert edge stretch into the reported stress-proxy values.

invented entities (1)

relative surface-based stress proxy no independent evidence
purpose: To provide a comparative metric of surface stress derived from reconstructed edge stretch.
Defined within the paper as a proxy rather than absolute wall stress; no independent falsifiable prediction is supplied.

pith-pipeline@v0.9.0 · 5641 in / 1704 out tokens · 67707 ms · 2026-05-10T18:11:39.340916+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

45 extracted references · 39 canonical work pages · 2 internal anchors

[1]

Ishiguro, Y

T. Ishiguro, Y. Omura, Y. Takano, T. Arai, K. Karagiozov, K. Fukuda, Y. Kikuta, N. Momozaki, H. Eguchi, M. Nishitani, M. Sakaguchi, M. Uchida, M. Suzuki, T. Kawamata, Low radial and axial force stent retriever reduces symptomatic subarachnoid hemorrhage after mechanical thrombectomy for acute middle cerebral artery and medium vessel occlusion: A prospecti...

work page doi:10.3389/fneur.2025.1723287 2025
[2]

A. A. Dmytriw, B. Musmar, H. Salim, S. Ghozy, J. E. Siegler, H. Kobeissi, H. Shaikh, J. Khalife, M. Abdalkader, P. Klein, T. N. Nguyen, J. J. Heit, R. W. Regenhardt, N. M. Cancelliere, J. D. Bernstock, K. E. Naamani, A. Amllay, L. Meyer, A. Dusart, F. Bellante, G. Forestier, A. Rouchaud, S. Saleme, C. Mounayer, J. Fiehler, A. L. Kühn, A. S. Puri, C. Dyzma...

work page doi:10.1177/23969873231219412 2024
[3]

Y. Liu, M. Abbasi, J. L. Arturo Larco, R. Kadirvel, D. F. Kallmes, W. Brinjikji, L. Savastano, Preclinical testing platforms for mechanical thrombectomy in stroke: A review on phantoms, in-vivo animal, and cadaveric models, Journal of Neurointerventional Surgery 13 (2021) 816–822. doi:10.1136/neurintsurg-2020-017133. 1, 12

work page doi:10.1136/neurintsurg-2020-017133 2021
[4]

Johnson, A

S. Johnson, A. Dwivedi, M. Mirza, R. McCarthy, M. Gilvarry, A Review of the Advancements in the in-vitro Modelling of Acute Ischemic Stroke and Its Treatment, Frontiers in Medical Technology 4 (Jun. 2022).doi:10.3389/fmedt.2022.879074. 1, 12

work page doi:10.3389/fmedt.2022.879074 2022
[5]

D. A. Poulos, M. T. Froehler, B. C. Good, Investigation of stent retriever removal forces in an experimental model of acute ischemic stroke, Frontiers in Neurology 15 (Oct. 2024). doi:10.3389/fneur.2024.1486738. 1, 12

work page doi:10.3389/fneur.2024.1486738 2024
[6]

Nagargoje, The role of friction forces in arterial mechanical thrombectomy: A review, Journal of Biomechanics 192 (2025) 112966.doi:10.1016/j.jbiomech.2025.112966

M. Nagargoje, The role of friction forces in arterial mechanical thrombectomy: A review, Journal of Biomechanics 192 (2025) 112966.doi:10.1016/j.jbiomech.2025.112966. 1, 12

work page doi:10.1016/j.jbiomech.2025.112966 2025
[7]

Ernst, F

M. Ernst, F. Sommer, M. Bartl, C. H. Riedel, P. Langer, Making the invisible visible: Integrated visualization and automated quantification of thrombus deformation during mechanical thrombectomy, Frontiers in Medical Technology 8 (Jan. 2026).doi:10.3389/fmedt.2026.1704010. 1

work page doi:10.3389/fmedt.2026.1704010 2026
[8]

Solav, K

D. Solav, K. M. Moerman, A. M. Jaeger, K. Genovese, H. M. Herr, MultiDIC: An Open-Source Toolbox for Multi-View 3D Digital Image Correlation, IEEE Access 6 (2018) 30520–30535. doi:10.1109/ACCESS.2018.2843725. 2

work page doi:10.1109/access.2018.2843725 2018
[9]

Palanca, G

M. Palanca, G. Tozzi, L. Cristofolini, The use of digital image correlation in the biomechanical area: A review, International Biomechanics 3 (2016) 1–21.doi:10.1080/23335432.2015.1117395. 2

work page doi:10.1080/23335432.2015.1117395 2016
[10]

A. C. Sanchez, D.-K. Kim, Assessment of the Film-Free Water Decal Method for Speckle Pattern Application in Digital Image Correlation, Sensors (Basel, Switzerland) 24 (2024) 5657. doi:10.3390/s24175657. 2

work page doi:10.3390/s24175657 2024
[11]

Mustafa, M

A. Mustafa, M. Volino, H. Kim, J.-Y. Guillemaut, A. Hilton, Temporally Coherent General Dynamic Scene Reconstruction, International Journal of Computer Vision 129 (2021) 123–141. doi:10.1007/s11263-020-01367-2. 2

work page doi:10.1007/s11263-020-01367-2 2021
[12]

G. Wu, T. Yi, J. Fang, L. Xie, X. Zhang, W. Wei, W. Liu, Q. Tian, X. Wang, 4D Gaussian Splatting for Real-Time Dynamic Scene Rendering (Jul. 2024).doi:10.48550/arXiv.2310.08528. 2, 4

work page doi:10.48550/arxiv.2310.08528 2024
[13]

Kortman, E

H. Kortman, E. Smit, M. Oei, R. Manniesing, M. Prokop, F. Meijer, 4D-CTA in Neurovascular Disease: A Review, AJNR: American Journal of Neuroradiology 36 (2015) 1026–1033.doi:10.3174/ajnr.A4162. 2 14

work page doi:10.3174/ajnr.a4162 2015
[14]

Kaneko, T

N. Kaneko, T. Mashiko, T. Ohnishi, M. Ohta, K. Namba, E. Watanabe, K. Kawai, Manufacture of patient-specific vascular replicas for endovascular simulation using fast, low-cost method, Scientific Reports 6 (2016) 39168.doi:10.1038/srep39168. 2

work page doi:10.1038/srep39168 2016
[15]

Zhang, A flexible new technique for camera calibration, IEEE Transactions on Pattern Analysis and Machine Intelligence 22 (2000) 1330–1334

Z. Zhang, A flexible new technique for camera calibration, IEEE Transactions on Pattern Analysis and Machine Intelligence 22 (2000) 1330–1334.doi:10.1109/34.888718. 3

work page doi:10.1109/34.888718 2000
[16]

Bradski, The OpenCV library, Dr

G. Bradski, The OpenCV library, Dr. Dobb’s Journal of Software Tools (2000). 3

2000
[17]

N. Ravi, V. Gabeur, Y.-T. Hu, R. Hu, C. Ryali, T. Ma, H. Khedr, R. Rädle, C. Rolland, L. Gustafson, E. Mintun, J. Pan, K. V. Alwala, N. Carion, C.-Y. Wu, R. Girshick, P. Dollár, C. Feichtenhofer, SAM 2: Segment Anything in Images and Videos (Oct. 2024).doi:10.48550/arXiv.2408.00714. 3

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2408.00714 2024
[18]

J. MacQueen, Some methods for classification and analysis of multivariate observations, in: Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability, Volume 1: Statistics, Vol. 5.1, University of California Press, 1967, pp. 281–298. 4

1967
[19]

Ester, H.-P

M. Ester, H.-P. Kriegel, J. Sander, X. Xu, A density-based algorithm for discovering clusters in large spatial databases with noise, in: Proceedings of the Second International Conference on Knowledge Discovery and Data Mining, KDD’96, AAAI Press, Portland, Oregon, 1996, pp. 226–231. 4

1996
[20]

Scikit-learn: Machine Learning in Python

F. Pedregosa, G. Varoquaux, A. Gramfort, V. Michel, B. Thirion, O. Grisel, M. Blondel, A. Müller, J. Nothman, G. Louppe, P. Prettenhofer, R. Weiss, V. Dubourg, J. Vanderplas, A. Passos, D. Cournapeau, M. Brucher, M. Perrot, É. Duchesnay, Scikit-learn: Machine Learning in Python, https://arxiv.org/abs/1201.0490v4 (Jan. 2012). 4

work page Pith review arXiv 2012
[21]

E., et al

P. Virtanen, R. Gommers, T. E. Oliphant, M. Haberland, T. Reddy, D. Cournapeau, E. Burovski, P. Peterson, W. Weckesser, J. Bright, S. J. van der Walt, M. Brett, J. Wilson, K. J. Millman, N. Mayorov, A. R. J. Nelson, E. Jones, R. Kern, E. Larson, C. J. Carey, İ. Polat, Y. Feng, E. W. Moore, J. VanderPlas, D. Laxalde, J. Perktold, R. Cimrman, I. Henriksen, ...

work page doi:10.1038/s41592-019-0686-2 2020
[22]

G. Taubin, A signal processing approach to fair surface design, in: Proceedings of the 22nd Annual Conference on Computer Graphics and Interactive Techniques, SIGGRAPH ’95, Association for Computing Machinery, New York, NY, USA, 1995, pp. 351–358.doi:10.1145/218380.218473. 4

work page doi:10.1145/218380.218473 1995
[23]

Perona, J

P. Perona, J. Malik, Scale-space and edge detection using anisotropic diffusion, IEEE Transactions on Pattern Analysis and Machine Intelligence 12 (1990) 629–639.doi:10.1109/34.56205. 4

work page doi:10.1109/34.56205 1990
[24]

Black, G

M. Black, G. Sapiro, D. Marimont, D. Heeger, Robust anisotropic diffusion, IEEE Transactions on Image Processing 7 (1998) 421–432.doi:10.1109/83.661192. 4

work page doi:10.1109/83.661192 1998
[25]

M. J. Black, P. Anandan, The Robust Estimation of Multiple Motions: Parametric and Piecewise-Smooth Flow Fields, Computer Vision and Image Understanding 63 (1996) 75–104. doi:10.1006/cviu.1996.0006. 4

work page doi:10.1006/cviu.1996.0006 1996
[26]

P. J. Huber, Robust Estimation of a Location Parameter, The Annals of Mathematical Statistics 35 (1964) 73–101.doi:10.1214/aoms/1177703732. 5

work page doi:10.1214/aoms/1177703732 1964
[27]

P. J. Rousseeuw, C. Croux, Alternatives to the Median Absolute Deviation, Journal of the American Statistical Association 88 (1993) 1273–1283.doi:10.1080/01621459.1993.10476408. 5

work page doi:10.1080/01621459.1993.10476408 1993
[28]

G. A. Holzapfel, Nonlinear Solid Mechanics: A Continuum Approach for Engineering, Kluwer Academic Publishers Dordrecht, 2000. 5

2000
[29]

J. Y. Chueh, A. K. Wakhloo, M. J. Gounis, Neurovascular Modeling: Small-Batch Manufacturing of Silicone Vascular Replicas, American Journal of Neuroradiology 30 (2009) 1159–1164. doi:10.3174/ajnr.A1543. 5

work page doi:10.3174/ajnr.a1543 2009
[30]

K. Gong, Z. Wen, M. Xu, W. He, Q. Wang, N. Zhang, Z. Li, C. Li, D. Lian, W. Zhao, X. He, M. Zhang, SWiT-4D: Sliding-Window Transformer for Lossless and Parameter-Free Temporal 4D Generation (Dec. 2025).doi:10.48550/arXiv.2512.10860. 7

work page doi:10.48550/arxiv.2512.10860 2025
[31]

J. M. Bland, D. G. Altman, Statistical methods for assessing agreement between two methods of clinical measurement, Lancet 1 (1986) 307–310. 8

1986
[32]

Nariai, T

Y. Nariai, T. Takigawa, A. Hyodo, K. Suzuki, Modification by an aspiration catheter for vessel stretching in thrombectomy using a stent retriever in vitro, Journal of Stroke and Cerebrovascular Diseases 32 (Mar. 2023).doi:10.1016/j.jstrokecerebrovasdis.2022.106948. 12

work page doi:10.1016/j.jstrokecerebrovasdis.2022.106948 2023
[33]

S. H. Baek, S. Kim, M. Kang, J.-H. Choi, H. J. Kwon, D. W. Kim, Effect of distal access catheter tip position on angiographic and clinical outcomes following thrombectomy using the combined stent-retriever and aspiration approach, PLoS ONE 16 (2021) e0252641. doi:10.1371/journal.pone.0252641. 12

work page doi:10.1371/journal.pone.0252641 2021
[34]

Kawamoto, Y

K. Kawamoto, Y. Nagao, M. Naganuma, Y. Inatomi, Y. Hashimoto, T. Yonehara, M. Nakajima, Stent-retriever characteristics and strategies associated with recanalization in thrombectomy for acute ischemic stroke, Clinical Neurology and Neurosurgery 242 (2024) 108332. doi:10.1016/j.clineuro.2024.108332. 12

work page doi:10.1016/j.clineuro.2024.108332 2024
[35]

Luraghi, J

G. Luraghi, J. F. Rodriguez Matas, G. Dubini, F. Berti, S. Bridio, S. Duffy, A. Dwivedi, R. McCarthy, B. Fereidoonnezhad, P. McGarry, C. B. L. M. Majoie, F. Migliavacca, INSIST investigators, 15 Applicability assessment of a stent-retriever thrombectomy finite-element model, Interface Focus 11 (2021) 20190123.doi:10.1098/rsfs.2019.0123. 12

work page doi:10.1098/rsfs.2019.0123 2021
[36]

S. M. Mousavi J S, D. Faghihi, K. Sommer, M. M. S. Bhurwani, T. R. Patel, B. Santo, M. Waqas, C. Ionita, E. I. Levy, A. H. Siddiqui, V. M. Tutino, Realistic computer modelling of stent retriever thrombectomy: A hybrid finite-element analysis-smoothed particle hydrodynamics model, Journal of the Royal Society, Interface 18 (2021) 20210583.doi:10.1098/rsif....

work page doi:10.1098/rsif.2021.0583 2021
[37]

Guédon, V

A. Guédon, V. Lepetit, SuGaR: Surface-Aligned Gaussian Splatting for Efficient 3D Mesh Reconstruction and High-Quality Mesh Rendering (Dec. 2023).doi:10.48550/arXiv.2311.12775. 13

work page doi:10.48550/arxiv.2311.12775 2023
[38]

W. Cai, W. Ye, P. Ye, T. He, T. Chen, DynaSurfGS: Dynamic Surface Reconstruction with Planar-based Gaussian Splatting (Aug. 2024).doi:10.48550/arXiv.2408.13972. 13

work page doi:10.48550/arxiv.2408.13972 2024
[39]

I. Liu, H. Su, X. Wang, Dynamic Gaussians Mesh: Consistent Mesh Reconstruction from Dynamic Scenes (Mar. 2025).doi:10.48550/arXiv.2404.12379. 13

work page doi:10.48550/arxiv.2404.12379 2025
[40]

Zheng, L

C. Zheng, L. Xue, J. Zarate, J. Song, GauSTAR: Gaussian Surface Tracking and Reconstruction, in: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, 2025, pp. 16543–16553. 13

2025
[41]

S. Ma, Y. Luo, W. Yang, Y. Yang, MaGS: Reconstructing and Simulating Dynamic 3D Objects with Mesh-adsorbed Gaussian Splatting (Nov. 2024).doi:10.48550/arXiv.2406.01593. 13

work page doi:10.48550/arxiv.2406.01593 2024
[42]

Samurai: Adapting segment anything model for zero-shot visual tracking with motion-aware memory.arXiv preprint arXiv:2411.11922,

C.-Y. Yang, H.-W. Huang, W. Chai, Z. Jiang, J.-N. Hwang, SAMURAI: Adapting Segment Anything Model for Zero-Shot Visual Tracking with Motion-Aware Memory (Nov. 2024). doi:10.48550/arXiv.2411.11922. 13

work page doi:10.48550/arxiv.2411.11922 2024
[43]

H. K. Cheng, S. W. Oh, B. Price, J.-Y. Lee, A. Schwing, Putting the Object Back into Video Object Segmentation (Apr. 2024).doi:10.48550/arXiv.2310.12982. 13

work page doi:10.48550/arxiv.2310.12982 2024
[44]

H. K. Cheng, S. W. Oh, B. Price, A. Schwing, J.-Y. Lee, Tracking Anything with Decoupled Video Segmentation (Sep. 2023).doi:10.48550/arXiv.2309.03903. 13

work page doi:10.48550/arxiv.2309.03903 2023
[45]

SAM 3: Segment Anything with Concepts

N. Carion, L. Gustafson, Y.-T. Hu, S. Debnath, R. Hu, D. Suris, C. Ryali, K. V. Alwala, H. Khedr, A. Huang, J. Lei, T. Ma, B. Guo, A. Kalla, M. Marks, J. Greer, M. Wang, P. Sun, R. Rädle, T. Afouras, E. Mavroudi, K. Xu, T.-H. Wu, Y. Zhou, L. Momeni, R. Hazra, S. Ding, S. Vaze, F. Porcher, F. Li, S. Li, A. Kamath, H. K. Cheng, P. Dollár, N. Ravi, K. Saenko...

work page internal anchor Pith review Pith/arXiv arXiv doi:10.48550/arxiv.2511.16719 2025