Pith. sign in

REVIEW 4 major objections 7 minor 24 references

ColorGS: High-fidelity Surgical Scene Reconstruction with Colored Gaussian Splatting

T0 review · 4 major / 7 minor · reviewed 2026-08-05 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗

Pith's one-line read This paper claims that adding spatially varying color anchors and a time-independent global motion offset to 3D Gaussian Splatting lifts deformable surgical-scene reconstruction to 39.85 dB PSNR — 1.50 dB above the previous best.

desk verdict The color-anchor adaptation is a legitimate incremental result, but the EDM's constant offset is redundant with base Gaussian positions, so the paper's advertised deformation novelty does not survive close reading. read the letter →

arxiv 2508.18696 v1 pith:2CUUS7AV submitted 2025-08-26 cs.CV

Qun Ji , Peng Li , Mingqiang Wei This is my paper
classification cs.CV
keywords 3DGaussianSplattingsurgicalscenereconstructiondeformabletissueendoscopicvideospatiallyvaryingcolordeformationmodelingreal-timerendering
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

The paper tries to show that the two remaining quality gaps in splatting-based reconstruction of deformable surgical scenes — fixed per-Gaussian colors that miss subtle tissue differences, and localized basis-function motion models that miss global drift — can be closed without sacrificing speed. ColorGS gives each Gaussian a set of learnable color anchors on the image plane whose distance-weighted contributions add spatially varying, view-dependent color to the base spherical-harmonic term. It then augments the standard basis-function deformation with one time-independent offset per parameter, so global motion trends are carried by the offset while local dynamics stay flexible. If the reported numbers hold, the rendered tissue is measurably sharper (PSNR 39.85 vs 38.35, SSIM 97.25%) at real-time rates on a consumer GPU — the combination that matters for intraoperative guidance and AR/VR training.

What carries the argument

Two modules carry the argument. Colored Gaussian Primitives: each Gaussian gains k=4 dynamic anchors A_i with learnable colors c_i on the rendering plane; at an intersection point p, anchor i contributes exp(−0.1‖p−A_i‖²)·c_i, and the sum F_c(p) is added to the base spherical-harmonic color, so color becomes spatially varying and view-dependent. Enhanced Deformation Model (EDM): each deformation parameter (shown for x-position) is ψ_x(t)=Σ_{j=0}^{16} ω_j^x b̃(t;θ_j^x,σ_j^x)+δ^x, i.e., B=17 learnable Gaussian basis functions plus one time-independent offset δ. The basis functions model localized dynamics; δ carries consistent global motion (uniform drift) that localized bumps fit poorly. Remo

What would settle it

Train ColorGS and Deform3DGS on a surgical clip containing an actual discontinuity — for instance a cautery cut that separates a tissue surface into two pieces mid-clip — and compare held-out PSNR; if the 1.50 dB advantage vanishes or reverses on such motion, the smooth-basis-plus-constant-offset deformation model is the part of the claim that fails to generalize.

Watch

Extended reading notes

Core claim

The paper's central claim is that ColorGS achieves state-of-the-art reconstruction of deformable tissue from endoscopic video by fixing two representational limits of prior 3DGS methods. First, instead of a fixed color per Gaussian, each primitive carries four learnable color anchors on the rendering plane; an exponential decay of distance weights their colors and adds the result to the base direction-dependent color (spherical harmonics), so each Gaussian can render different colors at different pixels and viewpoints. Second, instead of modeling motion purely as a sum of localized Gaussian basis bumps, the deformation includes a learnable time-independent offset δ for each parameter (positi

Load-bearing premise

The whole claim depends on the assumption that every tissue point's motion is a smooth combination of 17 Gaussian bumps plus one constant offset; any real surgical motion with jumps, tearing, or contact-driven discontinuities would not be captured by that form.

Editorial extensions

If this is right

  • On the EndoNeRF benchmark, ColorGS reports PSNR 39.85 vs 38.35 for Deform3DGS (the second-best method), together with better SSIM and LPIPS; the gap is attributed to the two new modules.
  • The color module alone is credited with most of the gain: the ablation shows 3DGS at 38.55 dB, SuperGaussians at 39.31 dB, and ColorGS at 39.85 dB on EndoNeRF.
  • The deformation module prefers Gaussian basis functions plus a global offset over Fourier+Polynomial series or pure Gaussian functions without the offset, per the ablation on EndoNeRF.
  • Because the rendering path is unchanged 3DGS rasterization, the method keeps real-time rendering and trains in 3000 iterations on one RTX 3060, so the fidelity gain does not raise hardware demands.
  • The same gains reproduce on a second dataset (StereoMIS), where PSNR rises from 30.68 to 32.64, suggesting the two fixes transfer beyond the primary benchmark.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • Extension beyond the paper: the anchor count k and decay rate λe control the spatial frequency of color detail the model can express; k=4 and λe=0.1 are fixed globally, so tissue with much finer texture or severe specular highlights might need region-adaptive k/λ to keep the benefit.
  • Extension beyond the paper: because the global offset δ is time-independent, the model assumes each Gaussian's residual motion is constant across the whole clip; longer recordings with slowly drifting tissue would likely need the offset to become piecewise in time — a testable modification.
  • Extension beyond the paper: the two-part recipe (spatially varying color plus a constant-motion offset) is not surgery-specific; porting both modules to general monocular dynamic-scene splatting benchmarks would test how much of the gain is domain-driven.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 7 minor

Summary. ColorGS proposes two modifications to dynamic 3D Gaussian Splatting for endoscopic surgical scene reconstruction. First, Colored Gaussian Primitives augment each Gaussian's spherical-harmonic color with k=4 learnable color anchors on the rendering plane, using an exponential falloff to add spatially varying color. Second, an Enhanced Deformation Model (EDM) combines B=17 Gaussian basis functions with a time-independent offset δ to model Gaussian motion. The method is trained with RGB and depth losses on EndoNeRF and StereoMIS. The authors report state-of-the-art PSNR/SSIM/LPIPS (e.g., 39.85 dB PSNR on EndoNeRF, 1.50 dB above Deform3DGS) and claim real-time rendering efficiency.

Significance. If the empirical results are reproducible, the color-anchor component is a plausible and potentially useful extension of 3DGS for surgical scenes, where tissue appearance is homogeneous and subtle. The paper includes ablations on public benchmarks and reports a consistent margin over EndoGaussian and Deform3DGS. However, the deformation novelty is mathematically redundant as written, and the evaluation is too thin to support the state-of-the-art claim: only five clips are aggregated, with no per-scene breakdown, variance, or significance tests, and no runtime measurements back the real-time claim. The color/deformation ablations also do not isolate the two contributions. The central deformation claim therefore needs substantial rework before the paper can be accepted.

major comments (4)
  1. [Sec. 2.3, Eq. (8)] The time-independent offset δ_x is functionally redundant with the optimized base Gaussian position μ_x. For any solution with base position μ_x and deformation ψ_x(t)=Σ ω_j b(t)+δ_x, reparameterizing μ'_x=μ_x+δ_x and ψ'_x(t)=Σ ω_j b(t) gives exactly the same rendered positions. A constant offset has no t-dependence, so it cannot model uniform translation or any global motion trend; the stated motivation that Gaussian bases are local and hence global motion is hard to model is not addressed by Eq. (8). Consequently, the EDM-vs-GS gain in Table 2 cannot be attributed to a representational improvement. Please either introduce a genuinely time-dependent global term (e.g., δ_x·t) or reframe the contribution as an optimization/regularization prior, with experiments that support that reframing.
  2. [Sec. 3.2, Table 1] The evaluation aggregates results over only 2 EndoNeRF scenes and 3 StereoMIS clips, with no per-scene PSNR/SSIM/LPIPS, no standard deviations, and no significance tests. In such small, heterogeneous surgical datasets, aggregate means can be driven by a single clip. The claimed state-of-the-art performance is not assessable without per-scene tables, the number of runs, and ideally paired significance tests. Please provide these; if the number of scenes is fixed by the benchmark, state this limitation explicitly.
  3. [Abstract and Sec. 3.1] The abstract and introduction claim 'real-time rendering efficiency' and 'computational practicality,' but the paper reports no runtime measurements: no FPS, training time, GPU memory, or comparison of rendering cost against baselines. Since real-time performance is one of the two stated motivations for using 3DGS over NeRF, this is a load-bearing claim. Please report actual timings on the same hardware and, if possible, per-method FPS.
  4. [Sec. 3.3, Table 2] The ablation design does not isolate the two proposed components. In the first block, 'Ours' is the full ColorGS (color anchors + EDM), while the other rows differ in both the color representation and the deformation model (e.g., 2DGS and SuperGS do not use EDM). In the second block, 'FPS' and 'GS' are compared with 'Ours,' which again includes both contributions. To support the claims, provide a 2×2 ablation over the same base 3DGS pipeline: base+GS, base+EDM, base+color anchors+GS, base+color anchors+EDM, on both datasets. The current table cannot tell whether the gains come from the color anchors, the deformation change, or their interaction.
minor comments (7)
  1. [Sec. 2.3, Eq. (7)] The basis-function notation is inconsistent: Eq. (7) uses unsubscripted θ and σ, while Eq. (8) refers to θ_j and σ_j. Please fix the subscripts and clarify whether θ and σ are learnable per basis function.
  2. [Sec. 2.2] The 'dynamic anchors' are never assigned temporal dynamics; no time index appears in Eqs. (4)–(6). Please specify whether the anchor positions/colors evolve over time or are static per Gaussian, and clarify how the anchor coordinates are optimized.
  3. [Sec. 2.2] The notion of a ray 'intersecting' a 3D Gaussian at a point p=(u,v) on a rendering plane is imprecise: 3DGS is a volumetric splat, not an implicit surface. Please define p formally (e.g., the projected Gaussian center or per-pixel evaluation) and state whether the anchor coordinates are in image space or local Gaussian space.
  4. [Sec. 2.4, Eq. (9)] The loss uses an unqualified norm ∥·∥; please specify whether it is L1 or L2, and state how the mask M is obtained for both RGB and depth terms.
  5. [Sec. 3.3, Table 2] The table block header 'Gaussian tracking' is confusing; it should read 'Deformation modeling' or similar. Also, 'FPS' is an unusual abbreviation for Fourier and Polynomial Series; spell it out in the table caption or use 'FPS/Fourier-Polynomial'.
  6. [Sec. 3.3] The comparison with SuperGaussian (SuperGS) is not apples-to-apples because SuperGS is built on 2DGS while ColorGS uses 3DGS. Please include a 3DGS-based color variant as a control, or at minimum note this confound.
  7. [General] Minor typos: 'Exeperiment' in the Section 3 heading, and the abstract's '1.5 higher' should be '1.50 dB higher' for consistency. The qualitative figure lacks error maps or zoomed insets, which would help substantiate the claimed improvement in subtle color variation.

Circularity Check

1 steps flagged · score 6.0 of 10

EDM's 'time-independent global motion parameter' is a constant offset absorbable into the already-free Gaussian base positions, so the claimed global-motion modeling reduces by construction.

  1. self definitional [Sec. 2.3, Eq. (8)]
    "Taking the center position change of the Gaussian in the x-direction as an example, the position at any time t can be expressed as: ψx(t, Θx) = Σ_{j=0}^{B-1} ω_j^x b̃(t; θ_j^x, σ_j^x) + δ^x (8) where B = 17 denotes the total number of basis functions. EDM decouples the global motion trend, which is challenging to represent with Gaussian functions, allowing for more flexible modeling of diverse motion patterns and enhancing the overall consistency and smoothness of the deformation."

    δ^x is a constant independent of time t. In 3DGS, every Gaussian already has a free base position μ^x optimized by the same loss. For any solution (μ^x, ψ^x(t)), set μ'^x = μ^x + δ^x and ψ'^x(t) = ψ^x(t) − δ^x. Then ψ'^x(t) is still a linear combination of the same B Gaussian basis functions, and the rendered Gaussian positions are identical. Thus Eq. (8) adds no representational capacity for 'global motion trends': a constant offset has no time dependence and is absorbed into the base position. The claimed capacity to model consistent global motion is therefore a reparameterization of existing parameters, not a new mechanism; any PSNR difference in Table 2 could come from optimization dynamics rather than the new term.

full rationale

The paper otherwise reports held-out test metrics for a learned representation against external baselines; the color-anchor module is a genuinely new per-Gaussian color parameterization, and there is no fitted parameter disguised as a prediction. No self-citation is load-bearing. However, the central novelty of the Enhanced Deformation Model is Eq. (8)'s '+δ^x' term, which the paper explicitly claims decouples global motion trends. Because δ^x is time-independent and each Gaussian already optimizes a base position μ^x, the term is mathematically equivalent to a shift of μ^x; the deformation function class is unchanged. This is a reduction by construction of the proposed 'global motion' mechanism, making the EDM's stated contribution partially vacuous. The empirical gain over the 'GS' ablation therefore cannot be attributed to the added parameter's expressivity on the basis of the paper's own equations.

Assumptions & free parameters 3 free parameters · 4 assumptions · 2 invented entities

The central quality gain depends on hand-set hyperparameters (k, lambda_e, B) and on modeling assumptions (color anchors, global offset) that are not independently validated or ablated beyond one dataset.

free parameters (3)
  • color anchor count k = 4
    Chosen by hand in Sec. 2.2; controls number of spatial color anchors per Gaussian.
  • color decay rate lambda_e = 0.1
    Set in Sec. 2.2; controls spatial influence radius of color anchors. No sensitivity analysis is provided.
  • number of basis functions B = 17
    Set in Sec. 2.3 for the deformation model; no ablation is provided.
assumptions (4)
  • standard math 3DGS differentiable rasterization with alpha blending is a valid scene representation (Eq. 1-3)
    Pulled from [6], used as backbone.
  • ad hoc to paper Anchors on the rendering plane with exponential falloff yield view-dependent color that improves tissue appearance modeling
    Introduced in Sec. 2.2; motivated by tissue similarity but not proven.
  • ad hoc to paper Time-independent offset delta added to basis-function deformation captures global motion trends
    Proposed in Sec. 2.3; no theoretical justification beyond intuition.
  • domain assumption The ground-truth depth D and tissue mask M in the loss are accurate and available for both datasets
    Used in Eq. 9 without discussion of acquisition or error.
invented entities (2)
  • dynamic color anchors (Colored Gaussian Primitives)
    purpose: Provide spatially varying, view-dependent color to each Gaussian to capture subtle tissue color variations
    A new mechanism introduced in Sec. 2.2; no external falsifiable prediction, only used in this pipeline.
  • time-independent global deformation parameter delta
    purpose: Model global motion trends alongside local basis functions
    Introduced in Sec. 2.3 as an additive offset; not independently validated.

how reviews work

0 comments
Cite this review

Pith. "Pith review of ColorGS: High-fidelity Surgical Scene Reconstruction with Colored Gaussian Splatting." pith.science (2026). https://pith.science/paper/2CUUS7AV

@misc{pith2026250818696,
  author       = {Pith},
  title        = {Pith review of: ColorGS: High-fidelity Surgical Scene Reconstruction with Colored Gaussian Splatting},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/2CUUS7AV}},
  note         = {Machine review of arXiv:2508.18696}
}
read the original abstract

High-fidelity reconstruction of deformable tissues from endoscopic videos remains challenging due to the limitations of existing methods in capturing subtle color variations and modeling global deformations. While 3D Gaussian Splatting (3DGS) enables efficient dynamic reconstruction, its fixed per-Gaussian color assignment struggles with intricate textures, and linear deformation modeling fails to model consistent global deformation. To address these issues, we propose ColorGS, a novel framework that integrates spatially adaptive color encoding and enhanced deformation modeling for surgical scene reconstruction. First, we introduce Colored Gaussian Primitives, which employ dynamic anchors with learnable color parameters to adaptively encode spatially varying textures, significantly improving color expressiveness under complex lighting and tissue similarity. Second, we design an Enhanced Deformation Model (EDM) that combines time-aware Gaussian basis functions with learnable time-independent deformations, enabling precise capture of both localized tissue deformations and global motion consistency caused by surgical interactions. Extensive experiments on DaVinci robotic surgery videos and benchmark datasets (EndoNeRF, StereoMIS) demonstrate that ColorGS achieves state-of-the-art performance, attaining a PSNR of 39.85 (1.5 higher than prior 3DGS-based methods) and superior SSIM (97.25\%) while maintaining real-time rendering efficiency. Our work advances surgical scene reconstruction by balancing high fidelity with computational practicality, critical for intraoperative guidance and AR/VR applications.

Figures

Figures reproduced from arXiv: 2508.18696 by the authors.

Figure 1
Figure 1. Pipeline of ColorGS, composed of (a) ColorGS Initialization, (b) Enhanced Deformational Modeling, and (c) Optimization using color and depth loss functions. 2.1 Preliminaries of 3D Gaussian Splatting 3DGS [6] introduces anisotropic 3D Gaussian parameters and a tile-based raster￾izer to achieve high-quality and real-time 3D scene reconstruction. Each Gaus- [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Illustration of the rendered images of baselines and ours. 3.3 Quantitative Evaluation of Key Components The Effect of Colored Gaussians. We conduct experiments using differ￾ent 3D representations, including the original 3DGS and 2D Gaussian Splatting (2DGS) [3], to evaluate the color expression capabilities of our Colored Gaus￾sian. As shown in [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

24 extracted references · 15 canonical work pages

  1. [1]

    Langenbeck’s archives of surgery406, 911–915 (2021)

    Boedecker, C., Huettl, F., Saalfeld, P., Paschold, M., Kneist, W., Baumgart, J., Preim, B., Hansen, C., Lang, H., Huber, T.: Using virtual 3d-models in surgical planning: workflow of an immersive virtual reality application in liver surgery. Langenbeck’s archives of surgery406, 911–915 (2021)

    work page 2021

  2. [2]

    In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion

    Cao, A., Johnson, J.: Hexplane: A fast representation for dynamic scenes. In: Pro- ceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recogni- tion. pp. 130–141 (2023)

    2023

  3. [3]

    Geiger, A., Gao, S., Chen, A., Yu, Z., Huang, B.: 2d gaussian splatting for geo- metrically accurate radiance fields (2024)

    work page 2024

  4. [4]

    International journal of computer assisted radiology and surgery18(7), 1185–1192 (2023)

    Hayoz, M., Hahne, C., Gallardo, M., Candinas, D., Kurmann, T., Allan, M., Sznit- man, R.: Learning how to robustly estimate camera pose in endoscopic videos. International journal of computer assisted radiology and surgery18(7), 1185–1192 (2023)

    work page 2023

  5. [5]

    Information Sciences p

    Kasymov, A., Czekaj, B., Mazur, M., Tabor, J., Spurek, P.: Neggs: Negative gaus- sian splatting. Information Sciences p. 121912 (2025)

    work page 2025

  6. [6]

    ACM Trans

    Kerbl, B., Kopanas, G., Leimkühler, T., Drettakis, G.: 3d gaussian splatting for real-time radiance field rendering. ACM Trans. Graph.42(4), 139–1 (2023)

    2023

  7. [7]

    arXiv preprint arXiv:2406.02720 (2024)

    Li, H., Liu, J., Sznaier, M., Camps, O.: 3d-hgs: 3d half-gaussian splatting. arXiv preprint arXiv:2406.02720 (2024)

    arXiv 2024

  8. [8]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Li, Z., Chen, Z., Li, Z., Xu, Y.: Spacetime gaussian feature splatting for real- time dynamic view synthesis. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 8508–8520 (2024)

    2024

Show all 24 references
  1. [9]

    In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition

    Lin, Y., Dai, Z., Zhu, S., Yao, Y.: Gaussian-flow: 4d reconstruction with dynamic 3d gaussian particle. In: Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. pp. 21136–21145 (2024)

    2024

  2. [10]

    In: Medical Image Computing and Computer Assisted Intervention–MICCAI 2020: 23rd International Confer- ence, Lima, Peru, October 4–8, 2020, Proceedings, Part III 23

    Liu, X., Stiber, M., Huang, J., Ishii, M., Hager, G.D., Taylor, R.H., Unberath, M.: Reconstructing sinus anatomy from endoscopic video–towards a radiation-free approach for quantitative longitudinal assessment. In: Medical Image Computing and Computer Assisted Intervention–MIC...

    2020

  3. [11]

    arXiv preprint arXiv:2401.12561 (2024)

    Liu, Y., Li, C., Yang, C., Yuan, Y.: Endogaussian: Gaussian splatting for de- formable surgical scene reconstruction. arXiv preprint arXiv:2401.12561 (2024)

    2024 arXiv

  4. [12]

    Computers in biology and medicine140, 105109 (2022)

    Luo, H., Wang, C., Duan, X., Liu, H., Wang, P., Hu, Q., Jia, F.: Unsupervised learning of depth estimation from imperfect rectified stereo laparoscopic images. Computers in biology and medicine140, 105109 (2022)

    2022

  5. [13]

    Commu- nications of the ACM65(1), 99–106 (2021)

    Mildenhall, B., Srinivasan, P.P., Tancik, M., Barron, J.T., Ramamoorthi, R., Ng, R.: Nerf: Representing scenes as neural radiance fields for view synthesis. Commu- nications of the ACM65(1), 99–106 (2021)

    2021

  6. [14]

    Medical image analysis69, 101946 (2021)

    Pelanis, E., Teatini, A., Eigl, B., Regensburger, A., Alzaga, A., Kumar, R.P., Rudolph, T., Aghayan, D.L., Riediger, C., Kvarnström, N., et al.: Evaluation of a novel navigation platform for laparoscopic liver surgery with organ deformation compensation using injected fiducial...

    2021

  7. [15]

    Journal of Plastic, Reconstructive & Aesthetic Surgery 67(9), 1171–1185 (2014) 10 Qun Ji, Peng Li, Mingqiang Wei

    Rodby, K.A., Turin, S., Jacobs, R.J., Cruz, J.F., Hassid, V.J., Kolokythas, A., Antony, A.K.: Advances in oncologic head and neck reconstruction: systematic review and future considerations of virtual surgical planning and computer aided design/computer aided modeling. Journal...

    2014

  8. [16]

    Song,J.,Wang,J.,Zhao,L.,Huang,S.,Dissanayake,G.:Dynamicreconstructionof deformablesoft-tissuewithstereoscopeinminimalinvasivesurgery.IEEERobotics and Automation Letters3(1), 155–162 (2017)

    2017

  9. [17]

    Hepatobiliary & Pancreatic Diseases International17(2), 101–112 (2018)

    Tang, R., Ma, L.F., Rong, Z.X., Li, M.D., Zeng, J.P., Wang, X.D., Liao, H.E., Dong, J.H.: Augmented reality technology for preoperative planning and intra- operative navigation during hepatobiliary surgery: A review of current methods. Hepatobiliary & Pancreatic Diseases Inter...

    2018

  10. [18]

    In: International conference on med- ical image computing and computer-assisted intervention

    Wang, Y., Long, Y., Fan, S.H., Dou, Q.: Neural rendering for stereo 3d reconstruc- tion of deformable tissues in robotic surgery. In: International conference on med- ical image computing and computer-assisted intervention. pp. 431–441. Springer (2022)

    2022

  11. [19]

    In: CVPR

    Wu, G., Yi, T., Fang, J., Xie, L., Zhang, X., Wei, W., Liu, W., Tian, Q., Wang, X.: 4d gaussian splatting for real-time dynamic scene rendering. In: CVPR. pp. 20310–20320. IEEE (2024)

    2024

  12. [20]

    Wu, T.Y., Meng, Q., Yang, L., Kumari, S., Pirouz, M.: Amassing the security: An enhanced authentication and key agreement protocol for remote surgery in healthcare environment. Comput. Model. Eng. Sci134(1), 317–341 (2023)

    2023

  13. [21]

    arXiv preprint arXiv:2411.18966 (2024)

    Xu, R., Chen, W., Wang, J., Liu, Y., Wang, P., Gao, L., Xin, S., Komura, T., Li, X., Wang, W.: Supergaussians: Enhancing gaussian splatting using primitives with spatially varying colors. arXiv preprint arXiv:2411.18966 (2024)

    2024 arXiv

  14. [22]

    In: Interna- tional Conference on Medical Image Computing and Computer-Assisted Interven- tion

    Yang, S., Li, Q., Shen, D., Gong, B., Dou, Q., Jin, Y.: Deform3dgs: Flexible defor- mation for fast surgical scene reconstruction with gaussian splatting. In: Interna- tional Conference on Medical Image Computing and Computer-Assisted Interven- tion. pp. 132–142. Springer (2024)

    2024

  15. [23]

    In: International conference on medical image computing and computer-assisted intervention

    Zha, R., Cheng, X., Li, H., Harandi, M., Ge, Z.: Endosurf: Neural surface re- construction of deformable tissues with stereo endoscope videos. In: International conference on medical image computing and computer-assisted intervention. pp. 13–23. Springer (2023)

    2023

  16. [24]

    IEEE transactions on medical imaging39(2), 400–412 (2019)

    Zhou, H., Jagadeesan, J.: Real-time dense reconstruction of tissue surface from stereo optical video. IEEE transactions on medical imaging39(2), 400–412 (2019)

    2019

Pith tools

Reviewed August 5, 2026 · model on record in the stance chip above.