Pith. sign in

REVIEW 3 major objections 5 minor 1 cited by

C3VDv2 -- Colonoscopy 3D video dataset with enhanced realism

T0 review · 3 major / 5 minor · reviewed 2026-08-06 · deepseek-v4-flash

Pith's one-line read C3VDv2 expands colonoscopy benchmarking from clean clips to 192 videos with per-frame 3D ground truth and realistic clinical artifacts.

desk verdict A genuinely useful, well-documented dataset extension whose only real weakness is that the registration validation metric is partly circular; still deserves peer review. read the letter →

arxiv 2506.24074 v2 pith:D7Z2VY33 submitted 2025-06-30 eess.IV cs.CV

classification eess.IVcs.CV
keywords colonoscopyvideodataset3Dreconstructiongroundtruthpixel-leveldepthandnormalsopticalflowsiliconecolonphantom2D/3DregistrationendoscopicSLAMvalidationsimulatedclinicalartifacts
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

C3VDv2 is a public video dataset built to give 3D-reconstruction researchers something existing colonoscopy datasets lack: per-pixel ground truth in videos that look and move like real clinical colonoscopy. The paper's claim is that 192 sequences, including 169 with pixel-level depth, normals, optical flow, occlusion, diffuse shading, 6-DoF poses, and 3D coverage models, capture artifacts such as fecal debris, blood, mucus, and lens occlusion, along with fast motion, loops, and paired clean-versus-debris trajectories. If the claim holds, colonoscopy SLAM, depth estimation, and reconstruction algorithms can be measured quantitatively under clinically relevant degradation rather than only on clean or synthetic scenes. The dataset also includes eight full-colon screening videos with recorded poses and fifteen deformation videos.

What carries the argument

The load-bearing mechanism is a 2D/3D registration pipeline that fits a rigid transform $T$ aligning the known 3D colon model to video frames. The optimization minimizes $(1 - S_e)$, where $S_e$ is the normalized overlap between binary edges from RGB keyframes and edges rendered from depth maps of the 3D model, and the search is carried out with CMA-ES over rotations of $\pm0.1$ radians and translations of $\pm7.5$ millimeters. A spherical omnidirectional camera model maps each pixel to a ray, reproducing the colonoscope's wide field of view, and hand-eye calibration converts 500 Hz robotic-arm poses into camera poses. Because the phantom stays fixed and the robotic arm repeats its trajectory with sub-millimeter precision, the same rigid transform is reused for the paired debris-filled video, so pixel-level ground truth can be rendered even when debris obscures the scene.

What would settle it

Take one of the phantom segments after imaging, scan it with CT or a structured-light scanner, and compare the scanned surface to the dataset's 3D model under the published per-video transform; if the residual misalignment on even the highest-quality videos exceeds a few millimeters, the registration is biased and the downstream pixel-level ground truth is in question.

Watch

Extended reading notes

Core claim

The central discovery is that high-fidelity silicone phantoms, robotic-arm trajectory logs, and edge-based 2D/3D registration can together produce pixel-level registered ground truth for colonoscopy videos that include realistic artifacts, at seven times the scale of the prior dataset. The paper demonstrates this on 169 registered videos totaling 67,886 frames, with per-frame depth, surface normals, optical flow, occlusion maps, diffuse maps, and camera poses, plus 3D coverage meshes. It further records eight simulated screening colonoscopies performed by an experienced gastroenterologist and fifteen deformation videos. The contribution is the dataset itself as a benchmark: 3D reconstruction and SLAM methods can now be quantitatively evaluated under fecal debris, blood, mucus, water and debris on the lens, fast and abrupt motion, and other conditions previously absent from paired ground-truth colonoscopy datasets.

Load-bearing premise

The whole dataset assumes the edge-based 2D/3D registration lands on the true phantom-to-camera transform for every video; since the true transform for real phantom videos is unknown and the validation uses the same edge-overlap signal that the optimization maximizes, a biased or locally stuck alignment would silently corrupt all pixel-level ground truth for that video.

Editorial extensions

If this is right

  • Depth, normal, optical-flow, and occlusion maps give pixel-level supervision for monocular depth estimation, normal prediction, and flow networks in colonoscopy-like images.
  • The provided 6-DoF poses and coverage meshes turn every registered video into a quantitative test bed for endoscopic SLAM, NeRF, and 3D Gaussian Splatting reconstruction.
  • Paired clean and debris-filled videos recorded on the same trajectory make it possible to measure exactly how much debris degrades reconstruction and tracking performance.
  • Fast motion, loops, en-face-to-down-the-barrel transitions, and lens occlusion provide stress cases that clean colon datasets cannot provide.
  • Eight full-colon withdrawal videos with poses support validation of tracking on realistic screening trajectories, and fifteen deformation videos support qualitative robustness checks.

Reading between the lines

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

  • Because the debris-filled videos inherit the clean colon's geometry as ground truth, an algorithm that reconstructs the debris surface itself will appear wrong; the intended reading is that the benchmark measures recovery of the underlying clean surface through visual clutter.
  • The loop and mirrored trajectories with known start-end correspondence make the dataset a ready-made drift benchmark: a SLAM system's endpoint pose error can be read directly from the provided poses.
  • The paired clean-versus-debris videos allow a controlled experiment of training on clean frames and testing on debris frames, isolating the effect of artifacts on depth and pose accuracy.
  • An independent geometric verification of the registration, such as CT-scanning one phantom after imaging and comparing the scanned surface to the published 3D model under the released transforms, would strengthen the dataset's accuracy claims, since the current quantitative validation uses the same edge-overlap cue that the registration optimizes.
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

3 major / 5 minor

Summary. The paper presents C3VDv2, a colonoscopy 3D video dataset containing 192 videos (169,371 frames) acquired from silicone colon phantoms with a clinical colonoscope mounted on a robotic arm. For 169 videos, the authors provide per-frame ground truth depth, surface normals, optical flow, occlusion, diffuse maps, 6-DoF pose, and 3D coverage meshes, generated by registering a 3D phantom model to video frames via an edge-based 2D/3D alignment (CMA-ES optimization of an edge similarity score). The dataset includes clean and debris-filled paired videos, simulated fecal debris, mucus, blood, foam, water/debris on the lens, fast camera motion, and en-face views; 15 deformation videos and 8 simulated screening videos are provided without pixel-level GT. Technical validation consists of an SSIM-based synchronization check for the paired clean/debris videos and a combination of qualitative scoring and a quantitative edge-overlap score (Dice and Chamfer distances) for the 2D/3D registration.

Significance. If the registration accuracy holds, C3VDv2 is a substantial contribution: it is one of the largest public colonoscopy datasets with pixel-level 3D ground truth under realistic artifacts, and its paired clean/debris setup enables controlled evaluation of robustness. The manuscript is strong in its detailed fabrication and acquisition protocols, open-source code, and honest enumeration of limitations (e.g., debris GT does not model debris, deformation videos lack pixel-wise GT). The cross-modal SSIM validation for synchronization is a good use of independent data. However, the load-bearing assumption that the edge-based registration recovers the true rigid transform for real phantom videos is not independently verified. The quantitative validation metric is computed from the same edge-overlap cue that is optimized, so it mainly demonstrates convergence of the optimizer, not correctness against physical truth. The qualitative scoring provides a modality-independent check, but it cannot rule out moderate global errors. These issues are addressable and do not invalidate the dataset, but they need to be addressed before the central claim is fully supported.

major comments (3)
  1. [Technical Validation, '2D/3D Registration Scoring'] The quantitative alignment score Qalign (Eq. 5) is computed from DexiNed edges on the RGB frames and Canny edges on the rendered depth frames. This is the same cue family that the registration objective in Eqs. (3)-(4) maximizes, so a high Qalign mainly indicates that the CMA-ES optimizer converged to a local maximum of the edge-overlap objective. The paper should provide an independent geometric verification on real phantom videos. A practical approach would be to perform a synthetic re-validation on the C3VDv2 phantoms: render depth and RGB images from known transforms, run the proposed registration pipeline, and report translational and rotational errors. Alternatively, adding external fiducials to a subset of phantoms and measuring their projected positions would provide a direct check. The current reliance on C3VD19's synthetic validation is insufficient because the v2 phantoms have new manufacturing defects and the acquisition conditions differ.
  2. [Data Records / Usage Notes] The paper does not report how many of the 169 registered videos received qualitative score 1, 2, or 3, nor the distribution of Qalign values. Since the usage notes explicitly recommend using only score 1 and 2 videos for tasks requiring high pixel-level accuracy, readers need the per-score counts to determine the effective size of the high-quality subset. Please include a summary table or histogram with the number of videos per qualitative score and the Qalign statistics for each group.
  3. [Methods, '2D/3D Registration Pipeline'] The registration pipeline uses an initial transform T_init that is 'manually estimated using the model-video overlay tool', followed by CMA-ES refinement. If the initialization is poor or if the search bounds (±0.1 rad, ±7.5 mm) are too tight, the optimizer can converge to a wrong local optimum. The paper should report the sensitivity of Qalign to the initialization and to the choice of the five keyframes, or at least state how often the search bounds had to be adjusted. This is important because a small number of keyframes (5) and manual substitution of low-edge frames could bias the estimated transform.
minor comments (5)
  1. [Abstract] The abstract states that 'Ground truth depth ... are provided for 169 colonoscopy videos', which is potentially misleading for the v3 debris videos because the depth, normal, and occlusion maps are those of the underlying clean 3D model and do not reflect the debris. Please qualify the sentence (e.g., 'with ground truth maps derived from the clean 3D model').
  2. [Table 1] In the row for C3VDv2 (Pixel-wise GT videos), the 'Debris' column is checked, but the pixel-wise GT does not model the debris surface. A footnote should clarify that the GT maps correspond to the clean colon geometry.
  3. [Figure 7] Figure 7b would be more informative if the box plot included the number of videos in each qualitative-score category, and if the axes were fully described in the caption (e.g., what the boxes and whiskers represent).
  4. [Technical Validation] The statement that 'the true transformation for real videos is unknown' appears only in the Technical Validation section. It should also be prominently stated in the Usage Notes or Data Records so that all users are aware of this limitation before downloading the dataset.
  5. [Methods, Eq. (1)] The polynomial in Eq. (1) omits α1, and the text later says α1=0. This is standard for the Scaramuzza model, but an explicit note in the text would avoid confusion for readers not familiar with the model.

Circularity Check

1 steps flagged · score 4.0 of 10

The quantitative alignment score is partly circular because it reuses the edge-overlap objective that defines the registration transform, while the dataset pipeline itself is otherwise self-contained.

  1. self definitional [Methods, 'CMAES Optimization and Edge Loss', Eqs. (3)-(4); Technical Validation, Eq. (5)]
    "Tfinal = argmin_T (1.0−S_e), (4) ... For each clean colon video, edges were extracted from both rendered depth frames and raw RGB frames using the DexiNed model. These edges were binarized, and the Dice coefficient and the Chamfer distance between the edge maps were computed and averaged across all video frames. Quantitative alignment scores Qalign were computed by combining standardized z-scores of the Dice coefficient Dz (edge similarity) and Chamfer distance Cz (edge displacement). The final metric is defined as: Qalign = 1/2 (Dz − Cz). (5)"

    The rigid transform T is estimated by minimizing (1−S_e), i.e., by maximizing overlap between RGB edge maps and rendered-depth edge maps. The quantitative validation metric Qalign is then computed from Dice and Chamfer distances between the same kinds of edge maps extracted from the same RGB and rendered-depth sources. A high Qalign therefore largely restates that CMA-ES found a local maximum of its own edge-overlap objective rather than independently confirming geometric correctness. This matters because the paper states 'the true transformation for real videos is unknown' and no independent geometric reference, such as a CT scan of the phantom, is used for real sequences.

full rationale

The dataset construction itself is not circular: the 3D colon models are sculpted independently, phantoms are cast from known molds, camera trajectories are logged by a robotic arm or EM tracker, and pixel-level maps are rendered from the model along the registered camera path. These are separate physical and geometric inputs. The single load-bearing step is the rigid transform T that connects the trajectory to the 3D model, and the paper's only quantitative check on T for real videos is Qalign, which reuses the same edge-overlap cue that Eq. (4) maximizes. Consequently, a high numerical alignment score partly reflects convergence to the optimizer's own objective rather than independent geometric accuracy. The qualitative overlay review provides some independent, modality-different evidence, and the cited C3VD19 synthetic validation is a genuine prior benchmark with known transforms, but neither rules out systematic bias on real phantoms with manufacturing defects, lighting variation, and debris. Because the paper explicitly acknowledges that the true transform for real videos is unknown, this is a material limitation of the validation, though it does not make the rendering pipeline itself definitionally circular. The score is therefore moderate: one central validation step is partly self-referential, while the dataset's other construction steps and qualitative checks retain independent content.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The central scientific contributions are the phantom designs, data collection, and registration; they rest on standard camera modeling and rigid-body assumptions. No new physical entities are introduced. The free parameters listed are engineering choices in the registration pipeline, not scientific constants fitted to a theory.

free parameters (4)
  • Number of keyframes per sequence = 5
    Empirically chosen in C3VD19 to balance accuracy and efficiency; affects registration quality.
  • CMA-ES search bounds and initial step size = ±0.1 rad, ±7.5 mm, sigma=0.1
    Hand-set; sometimes reduced for finer alignment.
  • EM tracker drift correction linear model = drift and offset estimated by linear regression
    Fitted to manually identified temporal correspondences between video and EM tracker timestamps.
  • Box filter window for pose log = 54 ms
    Hand-set noise reduction window before downsampling to 60 Hz.
assumptions (5)
  • domain assumption The phantom model remains stationary relative to the robotic arm during video capture.
    Stated in 2D/3D Registration Pipeline; single rigid transform represents model pose.
  • domain assumption Edge-based 2D/3D registration converges to the true model transform for phantom videos.
    No independent ground truth transform for real phantoms; relied on CMA-ES optimization of edge similarity (Eq. 3-4).
  • domain assumption The model transform estimated from the clean colon video (v2) is valid for the debris-filled video (v3).
    Relies on stationary phantom orientation and sub-millimeter robotic repeatability; validated only by SSIM on 10 frames per pair.
  • domain assumption The Scaramuzza omnidirectional camera model, with calibrated intrinsics, accurately maps pixels to rays.
    Used to render pixel-level ground truth from the 3D model.
  • domain assumption Silicone phantom surfaces match the 3D model closely enough for sub-millimeter ground truth.
    Acknowledged manufacturing defects cause Score 2-3 videos.

how reviews work

0 comments
Cite this review

Pith. "Pith review of C3VDv2 -- Colonoscopy 3D video dataset with enhanced realism." pith.science (2026). https://pith.science/paper/D7Z2VY33

@misc{pith2026250624074,
  author       = {Pith},
  title        = {Pith review of: C3VDv2 -- Colonoscopy 3D video dataset with enhanced realism},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/D7Z2VY33}},
  note         = {Machine review of arXiv:2506.24074}
}
read the original abstract

Spatial computer vision techniques have the potential to improve the diagnostic performance of colonoscopy. However, the lack of 3D colonoscopy datasets for training and validation hinders their development. This paper introduces C3VDv2, the second version (v2) of the high-definition Colonoscopy 3D Video Dataset, featuring enhanced realism designed to facilitate the quantitative evaluation of 3D colon reconstruction algorithms. 192 video sequences totaling 169,371 frames were captured by imaging 60 unique, high-fidelity silicone colon phantom segments. Ground truth depth, surface normals, optical flow, occlusion, diffuse maps, six-degree-of-freedom pose, coverage map, and 3D models are provided for 169 colonoscopy videos. Eight simulated screening colonoscopy videos acquired by a gastroenterologist are provided with ground truth poses. Lastly, the dataset includes 15 videos with colon deformations for qualitative assessment. C3VDv2 emulates diverse and challenging scenarios for 3D reconstruction algorithms, including fecal debris, mucous pools, blood, debris obscuring the colonoscope lens, en-face views, and fast camera motion. The enhanced realism of C3VDv2 will allow for more robust and representative development and evaluation of 3D reconstruction algorithms. Project Page - https://durrlab.github.io/C3VDv2/

Figures

Figures reproduced from arXiv: 2506.24074 by the authors.

Figure 1
Figure 1. The C3VDv2 dataset comprises 169 colonoscopy videos with paired ground truth data, including depth, normals, occlusion, optical flow, camera pose, diffuse images, and 3D model with coverage. New enhancements in the C3VDv2 dataset videos offer improved realism, adding colonic artifacts such as fecal debris, mucous pools, blood, and foam. Challenging cases like lens cleaning, water on the lens, debris on the lens, col… view at source ↗
Figure 2
Figure 2. Video snapshots of challenging scenarios: The C3VDv2 dataset features a wide range of challenging scenarios such as fast non-smooth motion, scope dipping in debris, lens cleaning, water on the lens, water jet, exploratory motion, loops, and en face to down the barrel transitions. Evaluating 3D reconstruction algorithms on C3VDv2 will provide a more representative assessment of their performance in real-world clinica… view at source ↗
Figure 3
Figure 3. Overview of the C3VDv2 data collection process. (a) High-resolution 3D models of colons are digitally sculpted. (b) Three-part molds are 3D printed. The high-resolution core insert (black) represents the colon lumen, while the outer shells (white) define the external contours of the phantom. (c) The phantom casting process involves (1) painting silicone layer-by-layer onto the insert mold with the vasculature hand-p… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Cross-section of Colon models: High-resolution 3D models were digitally sculpted referencing anatomical images. For positional reference, magnified views of lesions and Ileocecal valves are shown. The colors represent the distinct colon segments used for making phantom…
Figure 5
Figure 5. Figure 5: Sample frames from ‘clean’ colon sequences (Video v1 & Video v2): Four phantoms with varied tissue colors and textures were cast for each colon model. Paired ground truth maps are available for all RGB frames. 18/19 [PITH_FULL_IMAGE:figures/full_fig_p018_5.png]
Figure 6
Figure 6. Figure 6: Sample frames from debris-filled colon sequences (Video v3): C3VDv2 bridges the gap between phantom datasets and real-world colonoscopy by including colonic artifacts like fecal debris, mucous pools, blood, foam, en-face textureless wall views, and water on lens, which…
Figure 7
Figure 7. Figure 7: Technical validation results: (a) To verify the synchronization between the clean (v2) and debris-filled (v3) videos, the structural similarity index (SSIM) was computed across a window of frames with offsets. Maximum SSIM at offset 0 indicates that clean & debris-fill…

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

  1. ExtraGS: Enhancing Endoscopic View Extrapolation via Diffusion-Guided 3D Gaussian Splatting

    cs.CV 2026-07 reject novelty 5.5 of 10

    Diffusion-refined pseudo views from uncertainty-sampled virtual cameras reduce 3DGS extrapolation artifacts in endoscopic novel view synthesis.

Reference graph

Works this paper leans on

60 extracted references · 56 canonical work pages · cited by 1 Pith paper

  1. [1]

    Cancer statistics, 2025.CA: A Cancer Journal for Clinicians, 75(1):10, 2025

    Rebecca L Siegel, Tyler B Kratzer, Angela N Giaquinto, Hyuna Sung, and Ahmedin Jemal. Cancer statistics, 2025.CA: A Cancer Journal for Clinicians, 75(1):10, 2025

  2. [2]

    Quality indicators for colonoscopy.Gastrointestinal endoscopy, 81(1):31–53, 2015

    Douglas K Rex, Philip S Schoenfeld, Jonathan Cohen, Irving M Pike, Douglas G Adler, M Brian Fennerty, John G Lieb, Walter G Park, Maged K Rizk, Mandeep S Sawhney, et al. Quality indicators for colonoscopy.Gastrointestinal endoscopy, 81(1):31–53, 2015

  3. [3]

    Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths.New England Journal of Medicine, 366(8):687–696, 2012

    Ann G Zauber, Sidney J Winawer, Michael J O’Brien, Iris Lansdorp-V ogelaar, Marjolein van Ballegooijen, Benjamin F Hankey, Weiji Shi, John H Bond, Melvin Schapiro, Joel F Panish, et al. Colonoscopic polypectomy and long-term prevention of colorectal-cancer deaths.New England Journal of Medicine, 366(8):687–696, 2012

  4. [4]

    Magnitude, risk factors, and factors associated with adenoma miss rate of tandem colonoscopy: a systematic review and meta-analysis.Gastroenterology, 156(6):1661–1674, 2019

    Shengbing Zhao, Shuling Wang, Peng Pan, Tian Xia, Xin Chang, Xia Yang, Liliangzi Guo, Qianqian Meng, Fan Yang, Wei Qian, et al. Magnitude, risk factors, and factors associated with adenoma miss rate of tandem colonoscopy: a systematic review and meta-analysis.Gastroenterology, 156(6):1661–1674, 2019. 9/19

  5. [5]

    Artificial intelligence identifies and quantifies colonoscopy blind spots.Endoscopy, 53(12):1284–1286, 2021

    Sarah K McGill, Julian Rosenman, Rui Wang, Ruibin Ma, Jan-Michael Frahm, and Stephen Pizer. Artificial intelligence identifies and quantifies colonoscopy blind spots.Endoscopy, 53(12):1284–1286, 2021

  6. [6]

    Rnnslam: Reconstructing the 3d colon to visualize missing regions during a colonoscopy.Medical image analysis, 72:102100, 2021

    Ruibin Ma, Rui Wang, Yubo Zhang, Stephen Pizer, Sarah K McGill, Julian Rosenman, and Jan-Michael Frahm. Rnnslam: Reconstructing the 3d colon to visualize missing regions during a colonoscopy.Medical image analysis, 72:102100, 2021

  7. [7]

    SLAM Endoscopy enhanced by adversarial depth prediction

    Richard J Chen, Taylor L Bobrow, Thomas Athey, Faisal Mahmood, and Nicholas J Durr. Slam endoscopy enhanced by adversarial depth prediction.arXiv preprint arXiv:1907.00283, 2019

  8. [8]

    Real-time 3d reconstruction of colonoscopic surfaces for determining missing regions

    Ruibin Ma, Rui Wang, Stephen Pizer, Julian Rosenman, Sarah K McGill, and Jan-Michael Frahm. Real-time 3d reconstruction of colonoscopic surfaces for determining missing regions. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 573–582. Springer, 2019

Show all 60 references
  1. [9]

    A deep learning framework for quality assessment and restoration in video endoscopy.Medical image analysis, 68:101900, 2021

    Sharib Ali, Felix Zhou, Adam Bailey, Barbara Braden, James E East, Xin Lu, and Jens Rittscher. A deep learning framework for quality assessment and restoration in video endoscopy.Medical image analysis, 68:101900, 2021

  2. [10]

    Anita Rau, PJ Eddie Edwards, Omer F Ahmad, Paul Riordan, Mirek Janatka, Laurence B Lovat, and Danail Stoyanov. Implicit domain adaptation with conditional generative adversarial networks for depth prediction in endoscopy.International journal of computer assisted radiology and...

  3. [11]

    Bimodal camera pose prediction for endoscopy.IEEE Transactions on Medical Robotics and Bionics, 2023

    Anita Rau, Binod Bhattarai, Lourdes Agapito, and Danail Stoyanov. Bimodal camera pose prediction for endoscopy.IEEE Transactions on Medical Robotics and Bionics, 2023

  4. [12]

    Simcol3d—3d reconstruction during colonoscopy challenge.Medical Image Analysis, 96:103195, 2024

    Anita Rau, Sophia Bano, Yueming Jin, Pablo Azagra, Javier Morlana, Rawen Kader, Edward Sanderson, Bogdan J Matuszewski, Jae Young Lee, Dong-Jae Lee, et al. Simcol3d—3d reconstruction during colonoscopy challenge.Medical Image Analysis, 96:103195, 2024

  5. [13]

    Faisal Mahmood and Nicholas J. Durr. Deep learning and conditional random fields-based depth estimation and topograph- ical reconstruction from conventional endoscopy.Medical Image Analysis, 48:230–243, 2018

  6. [14]

    Bobrow, Gulfize Coskun, Kagan Incetan, Yasin Almalioglu, Faisal Mahmood, Eva Curto, Luis Perdigoto, Marina Oliveira, Hasan Sahin, Helder Araujo, Henrique Alexandrino, Nicholas J

    Kutsev Bengisu Ozyoruk, Guliz Irem Gokceler, Taylor L. Bobrow, Gulfize Coskun, Kagan Incetan, Yasin Almalioglu, Faisal Mahmood, Eva Curto, Luis Perdigoto, Marina Oliveira, Hasan Sahin, Helder Araujo, Henrique Alexandrino, Nicholas J. Durr, Hunter B. Gilbert, and Mehmet Turan. ...

  7. [15]

    Endomapper dataset of complete calibrated endoscopy procedures.Scientific Data, 10(1):671, 2023

    Pablo Azagra, Carlos Sostres, Ángel Ferrández, Luis Riazuelo, Clara Tomasini, O León Barbed, Javier Morlana, David Recasens, Víctor M Batlle, Juan J Gómez-Rodríguez, et al. Endomapper dataset of complete calibrated endoscopy procedures.Scientific Data, 10(1):671, 2023

  8. [16]

    Deep learning-based depth estimation from a synthetic endoscopy image training set

    Faisal Mahmood and Nicholas J Durr. Deep learning-based depth estimation from a synthetic endoscopy image training set. InMedical Imaging 2018: Image Processing, volume 10574, pages 521–526. SPIE, 2018

  9. [17]

    Deep learning with cinematic rendering: fine-tuning deep neural networks using photorealistic medical images.Physics in Medicine & Biology, 63(18):185012, 2018

    Faisal Mahmood, Richard Chen, Sandra Sudarsky, Daphne Yu, and Nicholas J Durr. Deep learning with cinematic rendering: fine-tuning deep neural networks using photorealistic medical images.Physics in Medicine & Biology, 63(18):185012, 2018

  10. [18]

    Synthesized colonoscopy dataset from high-fidelity virtual colon with abnormal simulation.Computers in Biology and Medicine, 186:109672, 2025

    Dongdong He, Ziteng Liu, Xunhai Yin, Hao Liu, Wenpeng Gao, and Yili Fu. Synthesized colonoscopy dataset from high-fidelity virtual colon with abnormal simulation.Computers in Biology and Medicine, 186:109672, 2025

  11. [19]

    Colonoscopy 3d video dataset with paired depth from 2d-3d registration.Medical image analysis, 90:102956, 2023

    Taylor L Bobrow, Mayank Golhar, Rohan Vijayan, Venkata S Akshintala, Juan R Garcia, and Nicholas J Durr. Colonoscopy 3d video dataset with paired depth from 2d-3d registration.Medical image analysis, 90:102956, 2023

  12. [20]

    Timo Teufel, Hongchao Shu, Roger D Soberanis-Mukul, Jan Emily Mangulabnan, Manish Sahu, S Swaroop Vedula, Masaru Ishii, Gregory Hager, Russell H Taylor, and Mathias Unberath. Oneslam to map them all: a generalized approach to slam for monocular endoscopic imaging based on trac...

  13. [21]

    Cudasift-slam: multiple-map visual slam for full procedure mapping in real human endoscopy.arXiv preprint arXiv:2405.16932, 2024

    Richard Elvira, Juan D Tardós, and José MM Montiel. Cudasift-slam: multiple-map visual slam for full procedure mapping in real human endoscopy.arXiv preprint arXiv:2405.16932, 2024

  14. [22]

    A surface-normal based neural framework for colonoscopy reconstruction

    Shuxian Wang, Yubo Zhang, Sarah K McGill, Julian G Rosenman, Jan-Michael Frahm, Soumyadip Sengupta, and Stephen M Pizer. A surface-normal based neural framework for colonoscopy reconstruction. InInternational Conference on Information Processing in Medical Imaging, pages 797–8...

  15. [23]

    Lightneus: Neural surface reconstruction in endoscopy using illumination decline

    Víctor M Batlle, José MM Montiel, Pascal Fua, and Juan D Tardós. Lightneus: Neural surface reconstruction in endoscopy using illumination decline. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 502–512. Springer, 2023. 10/19

  16. [24]

    Enerf-slam: a dense endoscopic slam with neural implicit representation

    Jiwei Shan, Yirui Li, Ting Xie, and Hesheng Wang. Enerf-slam: a dense endoscopic slam with neural implicit representation. IEEE Transactions on Medical Robotics and Bionics, 2024

  17. [25]

    Gaussian pancakes: geometrically-regularized 3d gaussian splatting for realistic endoscopic reconstruction

    Sierra Bonilla, Shuai Zhang, Dimitrios Psychogyios, Danail Stoyanov, Francisco Vasconcelos, and Sophia Bano. Gaussian pancakes: geometrically-regularized 3d gaussian splatting for realistic endoscopic reconstruction. InInternational Conference on Medical Image Computing and Co...

  18. [26]

    Endogslam: Real-time dense reconstruction and tracking in endoscopic surgeries using gaussian splatting

    Kailing Wang, Chen Yang, Yuehao Wang, Sikuang Li, Yan Wang, Qi Dou, Xiaokang Yang, and Wei Shen. Endogslam: Real-time dense reconstruction and tracking in endoscopic surgeries using gaussian splatting. InInternational Conference on Medical Image Computing and Computer-Assisted...

  19. [27]

    Multi-task learning with cross-task consistency for improved depth estimation in colonoscopy.Medical Image Analysis, 99:103379, 2025

    Pedro Esteban Chavarrias Solano, Andrew Bulpitt, Venkataraman Subramanian, and Sharib Ali. Multi-task learning with cross-task consistency for improved depth estimation in colonoscopy.Medical Image Analysis, 99:103379, 2025

  20. [28]

    Leveraging near-field lighting for monocular depth estimation from endoscopy videos

    Akshay Paruchuri, Samuel Ehrenstein, Shuxian Wang, Inbar Fried, Stephen M Pizer, Marc Niethammer, and Roni Sengupta. Leveraging near-field lighting for monocular depth estimation from endoscopy videos. InEuropean Conference on Computer Vision, pages 473–491. Springer, 2024

  21. [29]

    Endo-depth-and-motion: Reconstruction and tracking in endoscopic videos using depth networks and photometric constraints.IEEE Robotics and Automation Letters, 6(4):7225–7232, 2021

    David Recasens, José Lamarca, José M Fácil, JMM Montiel, and Javier Civera. Endo-depth-and-motion: Reconstruction and tracking in endoscopic videos using depth networks and photometric constraints.IEEE Robotics and Automation Letters, 6(4):7225–7232, 2021

  22. [30]

    Photometric single-view dense 3d reconstruction in endoscopy

    Víctor M Batlle, José MM Montiel, and Juan D Tardós. Photometric single-view dense 3d reconstruction in endoscopy. In 2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 4904–4910. IEEE, 2022

  23. [31]

    Tracking monocular camera pose and deformation for slam inside the human body

    Juan J Gómez Rodríguez, José MM Montiel, and Juan D Tardós. Tracking monocular camera pose and deformation for slam inside the human body. In2022 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 5278–5285. IEEE, 2022

  24. [32]

    Automated detection of anatomical landmarks during colonoscopy using a deep learning model.Journal of the Canadian Association of Gastroenterology, 6(4):145–151, 2023

    Mahsa Taghiakbari, Sina Hamidi Ghalehjegh, Emmanuel Jehanno, Tess Berthier, Lisa di Jorio, Saber Ghadakzadeh, Alan Barkun, Mark Takla, Mickael Bouin, Eric Deslandres, et al. Automated detection of anatomical landmarks during colonoscopy using a deep learning model.Journal of t...

  25. [33]

    Multimodal densenet.arXiv preprint arXiv:1811.07407, 2018

    Faisal Mahmood, Ziyun Yang, Thomas Ashley, and Nicholas J Durr. Multimodal densenet.arXiv preprint arXiv:1811.07407, 2018

  26. [34]

    Automated sizing of colorectal polyps using computer vision.Gut, 71(1):7–9, 2022

    Mohamed Abdelrahim, Hiroyasu Saiga, Naoto Maeda, Ejaz Hossain, Hitoshi Ikeda, and Pradeep Bhandari. Automated sizing of colorectal polyps using computer vision.Gut, 71(1):7–9, 2022

  27. [35]

    Batlle, Juan D

    Raúl Iranzo, Víctor M. Batlle, Juan D. Tardós, and José M.M. Montiel. EndoMetric: Near-light monocular metric scale estimation in endoscopy. InInt. conf. on Medical Image Computing and Computer Assisted Intervention (MICCAI), 2025

  28. [36]

    Computer vision and augmented reality in gastrointestinal endoscopy.Gastroenterology report, 3(3):179–184, 2015

    Nadim Mahmud, Jonah Cohen, Kleovoulos Tsourides, and Tyler M Berzin. Computer vision and augmented reality in gastrointestinal endoscopy.Gastroenterology report, 3(3):179–184, 2015

  29. [37]

    An objective comparison of detection and segmentation algorithms for artefacts in clinical endoscopy.Scientific reports, 10(1):2748, 2020

    Sharib Ali, Felix Zhou, Barbara Braden, Adam Bailey, Suhui Yang, Guanju Cheng, Pengyi Zhang, Xiaoqiong Li, Maxime Kayser, Roger D Soberanis-Mukul, et al. An objective comparison of detection and segmentation algorithms for artefacts in clinical endoscopy.Scientific reports, 10...

  30. [38]

    Mayank Golhar, Luojie Huang, and Nicholas J. Durr. HalF-SAM: SAM-based Haustral Fold Detection In Colonoscopy with Debris Suppression and Temporal Consistency . Inproceedings of Medical Image Computing and Computer Assisted Intervention – MICCAI 2025, volume LNCS 15968. Spring...

  31. [39]

    A self-supervised approach for detecting the edges of haustral folds in colonoscopy video

    Wenyue Jin, Rema Daher, Danail Stoyanov, and Francisco Vasconcelos. A self-supervised approach for detecting the edges of haustral folds in colonoscopy video. InMICCAI Workshop on Data Engineering in Medical Imaging, pages 56–66. Springer, 2023

  32. [40]

    Gan inversion for data augmenta- tion to improve colonoscopy lesion classification.IEEE Journal of Biomedical and Health Informatics, 2024

    Mayank V Golhar, Taylor L Bobrow, Saowanee Ngamruengphong, and Nicholas J Durr. Gan inversion for data augmenta- tion to improve colonoscopy lesion classification.IEEE Journal of Biomedical and Health Informatics, 2024

  33. [41]

    Realistic endoscopic illumination modeling for nerf-based data generation

    Dimitrios Psychogyios, Francisco Vasconcelos, and Danail Stoyanov. Realistic endoscopic illumination modeling for nerf-based data generation. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention, pages 535–544. Springer, 2023

  34. [42]

    Improving colonoscopy lesion classification using semi-supervised deep learning.IEEE Access, 9:631–640, 2020

    Mayank Golhar, Taylor L Bobrow, Mirmilad Pourmousavi Khoshknab, Simran Jit, Saowanee Ngamruengphong, and Nicholas J Durr. Improving colonoscopy lesion classification using semi-supervised deep learning.IEEE Access, 9:631–640, 2020. 11/19

  35. [43]

    Development and validation of the open-source automatic bowel preparation scale

    Kristoffer Mazanti Cold, Amihai Heen, Anishan Vamadevan, Andreas Slot Vilmann, Lars Konge, Morten Rasmussen, and Morten Bo Søndergaard Svendsen. Development and validation of the open-source automatic bowel preparation scale. Gastrointestinal Endoscopy, 2024

  36. [44]

    Diverse image generation with diffusion models and cross class label learning for polyp classification.arXiv preprint arXiv:2502.05444, 2025

    Vanshali Sharma, Debesh Jha, MK Bhuyan, Pradip K Das, and Ulas Bagci. Diverse image generation with diffusion models and cross class label learning for polyp classification.arXiv preprint arXiv:2502.05444, 2025

  37. [45]

    The paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to december 1, 2002.Gastrointest Endosc, 58:S3–S43, 2003

    R Lambert. The paris endoscopic classification of superficial neoplastic lesions: esophagus, stomach, and colon: November 30 to december 1, 2002.Gastrointest Endosc, 58:S3–S43, 2003

  38. [46]

    Fabrication and characterization of silicone-based tissue phantoms with tunable optical properties in the visible and near infrared domain

    Frederick Ayers, Alex Grant, Danny Kuo, David J Cuccia, and Anthony J Durkin. Fabrication and characterization of silicone-based tissue phantoms with tunable optical properties in the visible and near infrared domain. InDesign and Performance Validation of Phantoms Used in Con...

  39. [47]

    Wide-field optical property mapping and structured light imaging of the esophagus with spatial frequency domain imaging.Journal of biophotonics, 12(9):e201900005, 2019

    Jordan A Sweer, Mason T Chen, Kevan J Salimian, Richard J Battafarano, and Nicholas J Durr. Wide-field optical property mapping and structured light imaging of the esophagus with spatial frequency domain imaging.Journal of biophotonics, 12(9):e201900005, 2019

  40. [48]

    Ganpop: generative adversarial network prediction of optical properties from single snapshot wide-field images.IEEE transactions on medical imaging, 39(6):1988–1999, 2019

    Mason T Chen, Faisal Mahmood, Jordan A Sweer, and Nicholas J Durr. Ganpop: generative adversarial network prediction of optical properties from single snapshot wide-field images.IEEE transactions on medical imaging, 39(6):1988–1999, 2019

  41. [49]

    How to make cost-effective polyp simulators for high-fidelity simulation-based training in postpolypectomy bleeding management and emr.VideoGIE, 8(12):520–526, 2023

    Ally Y Zhao, Nikko Gimpaya, Mary Raina Angeli Fujiyoshi, Yusuke Fujiyoshi, Rishad Khan, James Lisondra, Catharine M Walsh, and Samir C Grover. How to make cost-effective polyp simulators for high-fidelity simulation-based training in postpolypectomy bleeding management and emr...

  42. [50]

    American college of gastroenterology guidelines for colorectal cancer screening 2008.Official journal of the American College of Gastroenterology| ACG, 104(3):739–750, 2009

    Douglas K Rex, David A Johnson, Joseph C Anderson, Phillip S Schoenfeld, Carol A Burke, and John M Inadomi. American college of gastroenterology guidelines for colorectal cancer screening 2008.Official journal of the American College of Gastroenterology| ACG, 104(3):739–750, 2009

  43. [51]

    An image inpainting technique based on the fast marching method.Journal of graphics tools, 9(1):23–34, 2004

    Alexandru Telea. An image inpainting technique based on the fast marching method.Journal of graphics tools, 9(1):23–34, 2004. 52.Karel J Zuiderveld et al. Contrast limited adaptive histogram equalization.Graphics gems, 4(1):474–485, 1994

  44. [53]

    Soria, E

    X. Soria, E. Riba, and A. Sappa. Dense extreme inception network: Towards a robust cnn model for edge detection. In 2020 IEEE Winter Conference on Applications of Computer Vision (WACV), pages 1912–1921, Los Alamitos, CA, USA, mar 2020. IEEE Computer Society

  45. [54]

    A computational approach to edge detection.IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-8(6):679–698, 1986

    John Canny. A computational approach to edge detection.IEEE Transactions on Pattern Analysis and Machine Intelligence, PAMI-8(6):679–698, 1986

  46. [55]

    Robot sensor calibration: solving ax= xb on the euclidean group.IEEE Transactions on Robotics and Automation, 10(5):717–721, 1994

    Frank C Park and Bryan J Martin. Robot sensor calibration: solving ax= xb on the euclidean group.IEEE Transactions on Robotics and Automation, 10(5):717–721, 1994

  47. [56]

    A toolbox for easily calibrating omnidirectional cameras

    Davide Scaramuzza, Agostino Martinelli, and Roland Siegwart. A toolbox for easily calibrating omnidirectional cameras. In2006 IEEE/RSJ International Conference on Intelligent Robots and Systems, pages 5695–5701. IEEE, 2006

  48. [57]

    Müller, and Petros Koumoutsakos

    Nikolaus Hansen, Sibylle D. Müller, and Petros Koumoutsakos. Reducing the time complexity of the derandomized evolution strategy with covariance matrix adaptation (cma-es).Evolutionary Computation, 11(1):1–18, 2003

  49. [58]

    Endoomni: Zero-shot cross-dataset depth estimation in endoscopy by robust self-learning from noisy labels.arXiv preprint arXiv:2409.05442, 2024

    Qingyao Tian, Zhen Chen, Huai Liao, Xinyan Huang, Lujie Li, Sebastien Ourselin, and Hongbin Liu. Endoomni: Zero-shot cross-dataset depth estimation in endoscopy by robust self-learning from noisy labels.arXiv preprint arXiv:2409.05442, 2024

  50. [59]

    Blood vessel delineation in endoscopic images with deep learning based scene classification

    Mayank Golhar, Yuji Iwahori, Manas Kamal Bhuyan, Kenji Funahashi, and Kunio Kasugai. Blood vessel delineation in endoscopic images with deep learning based scene classification. InPattern Recognition Applications and Methods: 6th International Conference, ICPRAM 2017, Porto, P...

  51. [60]

    A robust method for blood vessel extraction in endoscopic images with svm-based scene classification

    Mayank Golhar, Yuji Iwahori, Manas Kamal Bhuyan, Kenji Funahashi, and Kunio Kasugai. A robust method for blood vessel extraction in endoscopic images with svm-based scene classification. InICPRAM, pages 148–156, 2017

  52. [61]

    Comparing visual odometry systems in actively deforming simulated colon environments

    Mitchell J Fulton, J Micah Prendergast, Emily R DiTommaso, and Mark E Rentschler. Comparing visual odometry systems in actively deforming simulated colon environments. In2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), pages 4988–4995. IEEE, 202...

Pith tools

Reviewed August 6, 2026 · model on record in the stance chip above.