Pith. sign in

REVIEW 4 major objections 3 minor 1 cited by

GynSurg: A Comprehensive Gynecology Laparoscopic Surgery Dataset

T0 review · 4 major / 3 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper introduces GynSurg, a public multi-task gynecologic laparoscopy dataset that pairs action and side-effect labels with pixel-level instrument and anatomy masks, and claims it is the largest and most diverse such dataset to date.

desk verdict GynSurg is a valuable multi-task dataset for gynecologic laparoscopy, but Table 2's clip counts do not follow from its stated extraction rule, so the reproducibility claims need a fix before the numbers can be trusted. read the letter →

arxiv 2506.11356 v1 pith:EUN2LMGC submitted 2025-06-12 cs.CV

Sahar Nasirihaghighi , Negin Ghamsarian , Leonie Peschek , Matteo Munari , Heinrich Husslein , Raphael Sznitman , Klaus Schoeffmann This is my paper · ORCID
classification cs.CV
keywords medicalvideoanalysissurgicalworkflowgynecologiclaparoscopicsurgeryactionrecognitioninstrumentsegmentationanatomicalstructuremulti-taskdatasetsideeffectdetection
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

GynSurg is a public dataset for gynecologic laparoscopic surgery that unites two annotation levels: temporal labels for four surgical actions (coagulation, needle passing, suction/irrigation, transection) and two side effects (bleeding, smoke) across 152 expert-annotated videos, and pixel-level masks for 21 instruments and four anatomical structures across 12,362 frames, plus 75 unannotated videos. The paper's central claim is that this makes GynSurg the largest and most diverse multi-task dataset for gynecologic laparoscopy to date, filling a gap left by smaller, single-task datasets. The authors demonstrate the claim by benchmarking action-recognition and segmentation models under one standardized protocol and reporting baselines. If the claim holds, researchers get a common resource for training and comparing systems for surgical workflow analysis, documentation, and intraoperative assistance.

What carries the argument

The load-bearing mechanism is the annotation protocol itself. GynSurg organizes videos into three-second clips with one-second overlap, which makes action and side-effect recognition comparable across models, and applies four-fold cross-validation with balanced sampling to handle severe class imbalance. For segmentation, the dataset groups instrument labels into 13 primary surgical instruments and 8 auxiliary tools, merges underrepresented classes for evaluation, and retains three anatomy classes (uterus, fallopian tube, ovary). This standardized setup lets one dataset support both temporal and pixel-level tasks under a single training and evaluation protocol.

What would settle it

Label the 75 unannotated videos with the same expert protocol and benchmark the published baselines on them; a substantial drop in per-class accuracy, or a strongly different action distribution, would show the annotated subset is not representative of the full procedure pool.

Watch

Extended reading notes

Core claim

The central discovery is the dataset itself, as the authors state it: no prior public resource for gynecologic laparoscopy combines action-level temporal annotations with pixel-level spatial masks at this scale. GynSurg includes 152 high-definition videos annotated for four actions plus a rest class, with separate binary labels for bleeding and smoke on a dedicated subset, and 12,362 frames at 750x480 with masks for 21 instrument classes and four anatomy labels, from laparoscopic hysterectomy recordings. The paper reports that existing datasets cover only one or two tasks and are smaller, and it demonstrates GynSurg's usefulness by training standard models: ResNet-LSTM gives the best average action accuracy, DeepLabV3 leads instrument segmentation, and a center-point prompted Segment Anything Model reaches Dice scores above 80 percent on several classes.

Load-bearing premise

The load-bearing premise is that the 152 videos chosen from more than 600 recorded procedures fairly represent routine gynecologic laparoscopy, since the paper gives no selection criteria; if easier or cleaner cases were favored, the dataset's diversity and baseline numbers would not generalize.

Editorial extensions

If this is right

  • Action recognition, side-effect detection, and segmentation can now be trained and evaluated on the same videos, so end-to-end surgical workflow models can be compared under one protocol.
  • The separate bleeding and smoke labels make it possible to train detectors for events that matter for intraoperative assistance and postoperative review.
  • The documented class imbalance can serve as a testbed for imbalance-robust training methods, since rare classes like transection are frequently misclassified.
  • The 75 unannotated videos provide a ready-made setting for semi-supervised and self-supervised learning, as the paper states.
  • Public release of the dataset and training splits makes the baselines reproducible and lets other researchers add new tasks.

Reading between the lines

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

  • Because the 152 videos were selected from more than 600 recorded procedures without reported criteria, the strongest claim is about the annotated subset; labeling the 75 unannotated videos would turn them into a direct out-of-distribution test of the published baselines.
  • The same videos could support additional annotations, such as surgical phase or skill labels, since actions and side effects are already localized in time; the paper does not propose this.
  • The near-chance results on rare classes suggest the resource could be used to separate data-scarcity effects from model-architecture effects, which the paper does not pursue.
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 / 3 minor

Summary. The paper introduces GynSurg, a multi-task dataset for gynecologic laparoscopic surgery, comprising 152 expert-annotated videos for four actions and two side effects, 12,362 frames with pixel-level masks for instruments and anatomical structures, and 75 unannotated videos. The authors report benchmarks for action recognition, side-effect detection, and semantic segmentation using several standard architectures and a four-fold cross-validation protocol. The central claim is that GynSurg is the largest and most diverse multi-task gynecologic laparoscopy dataset to date, and the dataset and code are released publicly.

Significance. If the dataset is released as described, it is a potentially valuable resource for the surgical video analysis community. The paper's strengths include the public release of the dataset, the provision of training IDs and mask-creation code, the use of a standardized four-fold protocol with fold-wise standard deviations, and the breadth of baselined architectures (VGG/ResNet/ResNet3D for actions, DeepLabV3/UNet/PP/CE-Net/CPFNet/RecalNet/AdaptNet/SAM for segmentation). These features support reproducibility and make the resource useful for comparing future methods. However, several load-bearing inconsistencies must be resolved before the claims of comprehensiveness and reproducibility can be fully accepted.

major comments (4)
  1. [Section 2.1, Table 2] The reported clip counts are not consistent with the stated extraction rule. For a segment of duration D, partitioning into 3-second clips with a 1-second overlap yields about (D - 3)/2 clips per segment, so the totals should be approximately (total_duration - 3 * num_segments)/2 + num_segments. For Rest (790 segments, 24,067 s) this gives about 11,638 clips, not 1,100; for NeedlePassing (510 segments, 7,036 s) about 3,269, not 1,206; for Non-smoke (948 segments, 29,507 s) about 14,279, not 4,200. Bleeding is roughly consistent, so this is not simply a different unit. Either the extraction was not exhaustive as written (e.g., subsampling or a per-class cap) or one of the reported columns is incorrect. Because Table 2 is the primary evidence for dataset scale and class balance, the exact extraction rule and corrected counts must be provided for the benchmarks to be reproducible.
  2. [Section 2.1] The paper states that the 152 action videos were 'selected from over 600 recorded procedures' but gives no selection criteria. Without specifying how cases were chosen (e.g., random sampling, balancing by procedure type, exclusion of incomplete or low-quality recordings), the 'diverse' and 'comprehensive' claims cannot be evaluated, and benchmark results may not generalize to the broader population of gynecologic laparoscopies. Please document the selection protocol or explicitly analyze the potential selection bias.
  3. [Section 2.2 and Instrument Segmentation Setting] The abstract and Table 1 advertise 21 surgical instruments and 4 anatomical structures, but the evaluated segmentation benchmark uses only 7 instrument classes (after excluding trocar, clip applier, and corkscrew and merging suture-carrier, knot-pusher, needle-holder, and needle), 4 auxiliary classes (after excluding clip, colpotomizer, and glove and merging cannula/in-cannula and thread/thread-fragment), and 3 anatomical classes (after excluding the organ category). The reported Dice scores therefore do not cover the claimed 21 classes. The paper should clarify which classes are actually available and whether the released masks include all originally annotated classes; as written, the headline '21 instruments' overstates the evaluated scope.
  4. [Section 4.1, Table 5] The text states that ResNet-LSTM achieves 92.65% accuracy for bleeding and 86.03% for smoke, but Table 5 reports 88.26% and 80.69%, respectively (with F1 values of 92.51% and 85.58%). The text also says ResNet3D achieves 85.80% accuracy for smoke, while the table reports 83.54%. Please correct the text or the table so that the reported benchmark results are internally consistent.
minor comments (3)
  1. [Section 3.1] The training protocol is first described as 40 epochs, but later the same section says 'Models are optimized using cross-entropy loss over 30 epochs.' Please clarify which number is correct and reconcile the training schedule description.
  2. [Table 4 and Figure 4] Table 4 is titled 'Action recognition performance' and reports per-class values, while Figure 4 shows F1-scores; please state explicitly which metric is used in Table 4 and whether the per-class values are accuracy, precision, recall, or F1.
  3. [Abstract and Section 2.3] The dataset URL in the abstract uses 'GynSurge' while the GitHub repository is named 'GynSurg'; please harmonize the naming to avoid confusion for users.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; the dataset's central claims rest on external annotations and public data, not on self-referential derivation.

full rationale

GynSurg is a dataset contribution, not a theoretical derivation. The central claim, that it is the largest and most diverse multi-task gynecologic laparoscopy dataset, is supported by independently collected video data, expert annotations, and public release, none of which are defined in terms of the paper's own benchmarks or conclusions. The action-recognition and segmentation experiments use standard architectures (VGG, ResNet, ResNet3D, DeepLabV3, UNet, SAM, and variants); while some of these baselines cite prior work by the same authors, the benchmark numbers are produced by training on the released dataset with conventional protocols and are externally reproducible from the public data and training IDs. Such self-citations are not load-bearing for the dataset's existence, scale, or annotation quality. The internal inconsistency in Table 2 between clip counts and total durations under the stated three-second, one-second-overlap extraction rule is a correctness/reproducibility issue, not a circularity issue, because the clip counts are not used to define the dataset's value and no prediction is forced by construction. No fitted parameter is renamed as a prediction, no uniqueness theorem is imported from the authors' prior work, and no known result is repackaged as an organizing principle. The paper is self-contained against external benchmarks and public availability, so no significant circularity is present.

Assumptions & free parameters 2 free parameters · 3 assumptions · 0 invented entities

No free parameters are fitted in the construction of the dataset itself; the two listed hyperparameters are training choices that affect benchmark tables. The load-bearing assumptions are annotation quality, video representativeness, and protocol validity.

free parameters (2)
  • lambda_ce (segmentation loss weighting) = 0.8
    Hand-set in Section 3.2, Eq. 1, to balance cross-entropy and region terms. It affects reported Dice scores but is not fitted to test data.
  • sigma_smooth (Laplacian smoothing constant) = 1
    Set in Section 3.2 to stabilize the segmentation loss. A conventional training choice, not fitted to the data.
assumptions (3)
  • domain assumption Manual annotations by clinical experts are accurate and consistent enough to serve as ground truth.
    Sections 2.1 and 2.2 state expert annotation with standardized guidelines, but no inter-observer agreement or validation set is reported.
  • domain assumption The selected 152 action videos and 15 segmentation videos are representative of gynecologic laparoscopy.
    Sections 2.1 and 2.2 name two source hospitals and say videos were selected from over 600 procedures, but give no selection criteria.
  • domain assumption The benchmark protocol and four-fold cross-validation yield meaningful estimates of model performance on this dataset.
    Section 3 specifies architectures and training details, but there is no same-protocol comparison with prior datasets and no external validation of the ground truth.

how reviews work

0 comments
Cite this review

Pith. "Pith review of GynSurg: A Comprehensive Gynecology Laparoscopic Surgery Dataset." pith.science (2026). https://pith.science/paper/EUN2LMGC

@misc{pith2026250611356,
  author       = {Pith},
  title        = {Pith review of: GynSurg: A Comprehensive Gynecology Laparoscopic Surgery Dataset},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EUN2LMGC}},
  note         = {Machine review of arXiv:2506.11356}
}
read the original abstract

Recent advances in deep learning have transformed computer-assisted intervention and surgical video analysis, driving improvements not only in surgical training, intraoperative decision support, and patient outcomes, but also in postoperative documentation and surgical discovery. Central to these developments is the availability of large, high-quality annotated datasets. In gynecologic laparoscopy, surgical scene understanding and action recognition are fundamental for building intelligent systems that assist surgeons during operations and provide deeper analysis after surgery. However, existing datasets are often limited by small scale, narrow task focus, or insufficiently detailed annotations, limiting their utility for comprehensive, end-to-end workflow analysis. To address these limitations, we introduce GynSurg, the largest and most diverse multi-task dataset for gynecologic laparoscopic surgery to date. GynSurg provides rich annotations across multiple tasks, supporting applications in action recognition, semantic segmentation, surgical documentation, and discovery of novel procedural insights. We demonstrate the dataset quality and versatility by benchmarking state-of-the-art models under a standardized training protocol. To accelerate progress in the field, we publicly release the GynSurg dataset and its annotations

Figures

Figures reproduced from arXiv: 2506.11356 by the authors.

Figure 1
Figure 1. Sample frames illustrating different surgical actions [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Statistics of instrument (left) and anatomical structures (right) instances in the laparoscopy dataset. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Visualization of pixel-level annotations for surgical instruments and anatomical structures in the GynSurg dataset. [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: F1-scores for each action across different models [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]

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. Stitch-Inferencer: Enhance Endoscopic Video Segmentation and Tracking via Panoramic Reconstruction

    cs.CV 2026-07 conditional novelty 6.0 of 10

    An inference-time framework builds an explicit panoramic memory from endoscopic video and uses it to improve off-the-shelf segmentation and tracking models without retraining.

Reference graph

Works this paper leans on

35 extracted references · 35 canonical work pages · cited by 1 Pith paper

  1. [1]

    A., Tan, M

    Aldahoul, N., Karim, H. A., Tan, M. J. T., and Fermin, J. L. Transfer learning and decision fusion for real time distortion classification in laparoscopic videos. IEEE Access 9 (2021), 115006–115018

    work page 2021

  2. [2]

    A., Jalal, N

    Alshirbaji, T. A., Jalal, N. A., Docherty, P. D., Neumuth, T., and Möller, K. A deep learning spatial-temporal framework for detecting surgical tools in laparoscopic videos. Biomedical Signal Processing and Control 68 (2021), 102801

    work page 2021

  3. [3]

    Encoder- decoder with atrous separable convolution for semantic image segmentation

    Chen, L.-C., Zhu, Y., Papandreou, G., Schroff, F., and Adam, H. Encoder- decoder with atrous separable convolution for semantic image segmentation. In Proceedings of the European conference on computer vision (ECCV) (2018), pp. 801– 818

    work page 2018

  4. [4]

    Surgical endoscopy 37, 7 (2023), 5164–5175

    Den Boer, R., Jaspers, T., De Jongh, C., Pluim, J., V an Der Sommen, F., Boers, T., van Hillegersberg, R., V an Eijnatten, M., and Ruurda, J.Deep learning- based recognition of key anatomical structures during robot-assisted minimally invasive esophagectomy. Surgical endoscopy 37, 7 (2023), 5164–5175

    work page 2023

  5. [5]

    Cpfnet: Context pyramid fusion network for medical image segmentation

    Feng, S., Zhao, H., Shi, F., Cheng, X., Wang, M., Ma, Y., Xiang, D., Zhu, W., and Chen, X. Cpfnet: Context pyramid fusion network for medical image segmentation. IEEE Transactions on Medical Imaging 39 , 10 (2020), 3008–3018

    work page 2020

  6. [6]

    T., Weitz, J., and Speidel, S

    Funke, I., Mees, S. T., Weitz, J., and Speidel, S. Video-based surgical skill assessment using 3d convolutional neural networks. International journal of computer assisted radiology and surgery 14 (2019), 1217–1225

    work page 2019

  7. [7]

    Lensid: a cnn-rnn-based framework to- wards lens irregularity detection in cataract surgery videos

    Ghamsarian, N., Taschwer, M., Putzgruber-Adamitsch, D., Sarny, S., El- Shabrawi, Y., and Schoeffmann, K. Lensid: a cnn-rnn-based framework to- wards lens irregularity detection in cataract surgery videos. In Medical Image Computing and Computer Assisted Intervention–MICCAI 2021: 24th International Conference, Strasbourg, France, September 27–October 1, 20...

    work page 2021

  8. [8]

    Recal-net: Joint region-channel-wise cali- brated network for semantic segmentation in cataract surgery videos

    Ghamsarian, N., Taschwer, M., Putzgruber-Adamitsch, D., Sarny, S., El- Shabrawi, Y., and Schöffmann, K. Recal-net: Joint region-channel-wise cali- brated network for semantic segmentation in cataract surgery videos. In Neural Information Processing (Cham, 2021), T. Mantoro, M. Lee, M. A. Ayu, K. W. Wong, and A. N. Hidayanto, Eds., Springer International P...

    work page 2021

Show all 35 references
  1. [9]

    Deep- pyramid: Enabling pyramid view and deformable pyramid reception for semantic segmentation in cataract surgery videos

    Ghamsarian, N., Taschwer, M., Sznitman, R., and Schoeffmann, K. Deep- pyramid: Enabling pyramid view and deformable pyramid reception for semantic segmentation in cataract surgery videos. In International Conference on Medical Image Computing and Computer-Assisted Intervention...

    2022

  2. [10]

    Ce-net: Context encoder network for 2d medical image segmentation

    Gu, Z., Cheng, J., Fu, H., Zhou, K., Hao, H., Zhao, Y., Zhang, T., Gao, S., and Liu, J. Ce-net: Context encoder network for 2d medical image segmentation. IEEE Transactions on Medical Imaging 38 , 10 (2019), 2281–2292

    2019

  3. [11]

    J., Shen, Y., W allis, P., Allen-Zhu, Z., Li, Y., W ang, S., W ang, L., Chen, W., et al

    Hu, E. J., Shen, Y., W allis, P., Allen-Zhu, Z., Li, Y., W ang, S., W ang, L., Chen, W., et al. Lora: Low-rank adaptation of large language models. ICLR 1, 2 (2022), 3

    2022

  4. [12]

    C., Lo, W.-Y., et al.Segment anything

    Kirillov, A., Mintun, E., Ravi, N., Mao, H., Rolland, C., Gustafson, L., Xiao, T., Whitehead, S., Berg, A. C., Lo, W.-Y., et al.Segment anything. In Proceedings of the IEEE/CVF international conference on computer vision (2023), pp. 4015–4026

    2023

  5. [13]

    Kitaguchi, D., Takeshita, N., Matsuzaki, H., Takano, H., Owada, Y., Enomoto, T., Oda, T., Miura, H., Y amanashi, T., W atanabe, M., et al.Real-time automatic surgical phase recognition in laparoscopic sigmoidectomy using the convolu- tional neural network-based deep learning a...

    2020

  6. [14]

    Laparoscopic surgery reduces the inci- dence of surgical site infections compared to the open approach for colorectal procedures: a meta-analysis

    Kulkarni, N., and Arulampalam, T. Laparoscopic surgery reduces the inci- dence of surgical site infections compared to the open approach for colorectal procedures: a meta-analysis. Techniques in coloproctology 24, 10 (2020), 1017–1024

    2020

  7. [15]

    Glenda: gynecologic laparoscopy endometriosis dataset

    Leibetseder, A., Kletz, S., Schoeffmann, K., Keckstein, S., and Keckstein, J. Glenda: gynecologic laparoscopy endometriosis dataset. In International Conference on Multimedia Modeling (2019), Springer, pp. 439–450

    2019

  8. [16]

    J., Kletz, S., Münzer, B., Schoeff- mann, K., and Keckstein, J

    Leibetseder, A., Petscharnig, S., Primus, M. J., Kletz, S., Münzer, B., Schoeff- mann, K., and Keckstein, J. Lapgyn4: a dataset for 4 automatic content analysis problems in the domain of laparoscopic gynecology. In Proceedings of the 9th ACM multimedia systems conference (2018...

    2018

  9. [17]

    Smoke detection in endoscopic surgery videos: a first step towards retrieval of semantic events

    Loukas, C., and Georgiou, E. Smoke detection in endoscopic surgery videos: a first step towards retrieval of semantic events. The International Journal of Medical Robotics and Computer Assisted Surgery 11 , 1 (2015), 80–94

    2015

  10. [18]

    Multimedia Tools and Applications 46 (2010), 521–544

    Lux, M., Marqes, O., Schöffmann, K., Böszörmenyi, L., and Lajtai, G.A novel tool for summarization of arthroscopic videos. Multimedia Tools and Applications 46 (2010), 521–544

    2010

  11. [19]

    Surgai: deep learning for computerized laparoscopic image understanding in gynaecology

    Madad Zadeh, S., Francois, T., Calvet, L., Chauvet, P., Canis, M., Bartoli, A., and Bourdel, N. Surgai: deep learning for computerized laparoscopic image understanding in gynaecology. Surgical endoscopy 34 (2020), 5377–5383

    2020

  12. [20]

    Relevance segmentation of laparoscopic videos

    Münzer, B., Schoeffmann, K., and Böszörmenyi, L. Relevance segmentation of laparoscopic videos. In 2013 IEEE international symposium on multimedia (2013), IEEE, pp. 84–91

    2013

  13. [21]

    Content-based processing and analysis of endoscopic images and videos: A survey

    Münzer, B., Schoeffmann, K., and Böszörmenyi, L. Content-based processing and analysis of endoscopic images and videos: A survey. Multimedia Tools and Applications 77 (2018), 1323–1362

    2018

  14. [22]

    Event recognition in laparoscopic gynecology videos with hybrid transformers

    Nasirihaghighi, S., Ghamsarian, N., Husslein, H., and Schoeffmann, K. Event recognition in laparoscopic gynecology videos with hybrid transformers. In International Conference on Multimedia Modeling (2024), Springer, pp. 82–95

    2024

  15. [23]

    Action recognition in video recordings from gynecologic laparoscopy

    Nasirihaghighi, S., Ghamsarian, N., Stefanics, D., Schoeffmann, K., and Hus- slein, H. Action recognition in video recordings from gynecologic laparoscopy. In 2023 IEEE 36th International symposium on computer-based medical systems (CBMS) (2023), IEEE, pp. 29–34

    2023

  16. [24]

    U-net: Convolutional networks for biomedical image segmentation

    Ronneberger, O., Fischer, P., and Brox, T. U-net: Convolutional networks for biomedical image segmentation. In Medical Image Computing and Computer- Assisted Intervention – MICCAI 2015 (2015), pp. 234–241

    2015

  17. [25]

    Keyframe extraction in endoscopic video

    Schoeffmann, K., Del Fabro, M., Szkaliczki, T., Böszörmenyi, L., and Keck- stein, J. Keyframe extraction in endoscopic video. Multimedia Tools and Appli- cations 74 (2015), 11187–11206

    2015

  18. [26]

    Video retrieval in laparoscopic video recordings with dynamic content descriptors

    Schoeffmann, K., Husslein, H., Kletz, S., Petscharnig, S., Muenzer, B., and Beecks, C. Video retrieval in laparoscopic video recordings with dynamic content descriptors. Multimedia Tools and Applications 77 (2018), 16813–16832

    2018

  19. [27]

    A., Ochoa-Ruiz, G., and Ali, S

    Teevno, M. A., Ochoa-Ruiz, G., and Ali, S. A semi-supervised teacher-student framework for surgical tool detection and localization. Computer Methods in Biomechanics and Biomedical Engineering: Imaging & Visualization 11 , 4 (2023), 1033–1041

    2023

  20. [28]

    Surgical endoscopy 35 (2021), 1651–1658

    Tokuyasu, T., Iwashita, Y., Matsunobu, Y., Kamiyama, T., Ishikake, M., Sak- aguchi, S., Ebe, K., Tada, K., Endo, Y., Etoh, T., et al.Development of an artificial intelligence system using deep learning to indicate anatomical landmarks during laparoscopic cholecystectomy. Surgi...

    2021

  21. [29]

    Learning spatiotemporal features with 3d convolutional networks

    Tran, D., Bourdev, L., Fergus, R., Torresani, L., and Paluri, M. Learning spatiotemporal features with 3d convolutional networks. In Proceedings of the IEEE international conference on computer vision (2015), pp. 4489–4497

    2015

  22. [30]

    Minimally invasive surgery: national trends in adoption and future directions for hospital strategy

    Tsui, C., Klein, R., and Garabrant, M. Minimally invasive surgery: national trends in adoption and future directions for hospital strategy. Surgical endoscopy 27 (2013), 2253–2257

    2013

  23. [31]

    Autolaparo: A new dataset of integrated multi-tasks for image-guided surgical automation in laparoscopic hysterectomy

    Wang, Z., Lu, B., Long, Y., Zhong, F., Cheung, T.-H., Dou, Q., and Liu, Y. Autolaparo: A new dataset of integrated multi-tasks for image-guided surgical automation in laparoscopic hysterectomy. In International Conference on Medical Image Computing and Computer-Assisted Interv...

    2022

  24. [32]

    Unified perceptual parsing for scene understanding

    Xiao, T., Liu, Y., Zhou, B., Jiang, Y., and Sun, J. Unified perceptual parsing for scene understanding. In Proceedings of the European conference on computer vision (ECCV) (2018), pp. 418–434

    2018

  25. [33]

    M., François, T., Comptour, A., Canis, M., Bourdel, N., and Bartoli, A

    Zadeh, S. M., François, T., Comptour, A., Canis, M., Bourdel, N., and Bartoli, A. Surgai3. 8k: a labeled dataset of gynecologic organs in laparoscopy with appli- cation to automatic augmented reality surgical guidance. Journal of Minimally Invasive Gynecology 30, 5 (2023), 397–405

    2023

  26. [34]

    Laparoscopic image-based critical action recognition and anticipation with explainable features

    Zhang, J., Zhou, S., Wang, Y., Shi, S., Wan, C., Zhao, H., Cai, X., and Ding, H. Laparoscopic image-based critical action recognition and anticipation with explainable features. IEEE Journal of Biomedical and Health Informatics 27 , 11 (2023), 5393–5404

    2023

  27. [35]

    Zhou, Z., Siddiqee, M. M. R., Tajbakhsh, N., and Liang, J.Unet++: Redesigning skip connections to exploit multiscale features in image segmentation. IEEE Transactions on Medical Imaging 39 , 6 (2020), 1856–1867

    2020

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.