Surgical Action Triplet Recognition 2021
Breaking News:
Challenge joint publication
- 30/11/2022: Challenge report manuscript accepted at Medical Image Analysis journal. See the arXiv report.
Challenge Winners
- 1st Prize: Team Trequartista, University of Chicago, USA
- 2nd Prize: Team SIAT CAMI, Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences, China
- 3rd Prize: Team HFUT-MedIA, Hefei University of Technology, China
The complete results of the challenge can be found here.
Dataset Release
- CholecT45 is a subset of CholecT50 dataset excluding the challenge test set. CholecT45 has been released under CC BY-NC-SA 4.0 license on April 12, 2022. Starting from this date, it can be used for submissions to conferences, journals and other venues. Download access is through the CAMMA website.
- A joint publication summarizing all the challenge methods and results is now available on the arXiv.
- Reference paper that introduces and presents the CholecT50 dataset:
- Nwoye, C.I., Yu T., Gonzalez C., Seeliger, B., Mascagni, P., Mutter, D., Marescaux, J., and Padoy, N. "Rendezvous: Attention Mechanisms for the Recognition of Surgical Action Triplets in Endoscopic Videos.” Medical Image Analysis, 78 (2022) 102433, arXiv:2109.03223, [Supplementary video].
- Official splits of the dataset for developing deep learning models is contained in:
- Nwoye, C.I., and Padoy, N. "Data Splits and Metrics for Method Benchmarking on Surgical Action Triplet Datasets”, 2022, arXiv:2204.05235.
- Did you miss the challenge? Did you enjoy the challenge? Are you interested in the next edition? Due to overwhelming success of the challenge, a number of people are requesting for a second run. Please, checkout the new edition: CholecTriplet2022 on surgical action triplet recognition and localization.
Overview:
With the development of context-aware decision support in the operating room, it has become imperative to analyze surgical workflow activities at a fine-grained level to foster safety and efficiency. Most of the existing works recognize surgical actions at a coarse-grained level (such as phases, stages, single verb, etc.) leaving out some detailed information needed to analyze surgical workflow at par with the current pace of deep learning and artificial intelligence on activity recognition. Hence, we aim to recognize surgical actions as a triplet of instrument, verb and target.
To this effect, we introduce a new and unique endoscopic dataset, CholecT50, in which every frame has been annotated with labels from the triplet classes. The dataset is very useful for the development and evaluation of algorithms targeting the recognition of instrument-tissue interaction in laparoscopic cholecystectomies.
This sub-challenge focuses on exploiting machine learning methods for the online automatic recognition of surgical actions as a series of triplets. Participants will develop and compete with algorithms to recognize action triplets directly from the provided surgical videos. This novel challenge investigates the state-of-the-art on surgical fine-grained activity recognition and will establish a new promising research direction in computer-assisted surgery.
This year's challenge will feature a colab code blog providing some sample codes for quick start, and a T50 slack community to share ideas and interacts with organizers and co-participants. The teams with the best results will be awarded prizes. We also plan a joint publication with all participants after the challenge. It indeed will be a rewarding experience.
A complete structured description of the accepted MICCAI challenge designs can be found here.
Industrial Sponsors:
Academic Sponsors:
