This year, alongside with the annual workshop, M2CAI is also holding two challenge events. The challenges focus on two tasks that need to be solved to develop a surgical workflow monitoring system.

In each challenge, awards sponsored by Labex CAMI and IHU Strasbourg will be given to the submissions obtaining the best results.


Surgical workflow challenge (Challenge Results)

Detecting the surgical workflow of a minimally invasive surgery has many applications, from surgeon skill evaluation and report generation to automatic triggering of actions in the OR. There exist many approaches that can detect the current workflow phase based on instruments in use, but this data usually requires unreliable, additional sensors, while laparoscopic or endoscopic video is always available in high quality. For this reason, various approaches that rely on visual features have been proposed in the recent years.

For this challenge, we have defined eight surgical phases for cholecystectomy procedures. This challenge consists of identifying the phase at time t using solely visual information from the laparoscopic videos. In particular, the challenge focuses on surgical phase recognition in online mode, where the algorithm estimates the surgical phase at time without seeing any future information (i.e., images that come after time t).

The m2cai16-workflow dataset consists of 41 cholecystectomy videos with ground truth annotations of the phases. The dataset is split is into two parts: training subset (27 videos) and testing subset (14 videos).

To obtain access to our training and testing data for the surgical workflow challenge, please kindly contact us by filling this form. You will be added to our mailing list and the link to download the dataset will be sent to you once it is available.

To submit your results, please go to the submission page.


Surgical tool detection challenge (Challenge Results)

The objective of this challenge is to identify all surgical tools that are present in an image. The problem of detecting surgical tools during a surgery is a prevalent topic in the community. It can be used to various applications, such as surgical video indexing and also report generation. In addition, there is inherently a strong correlation between this challenge and the surgical workflow recognition challenge.

We have defined seven surgical tools (seen below) that are typically used in cholecystectomy procedures. This challenge consists of detecting these surgical tools from the images. Note that, the localization of the tools is not the concern of this challenge. Thus, from an image, the output is a binary vector (of seven elements), indicating the presence of all seven tools.

The m2cai16-tool dataset consist of 15 cholecystectomy videos with ground truth binary annotations of the present tools. The dataset is split into two parts: training subset (10 videos) and testing subset (5 videos).

tool_grasper_labeled tool_hook_labeled tool_clipper_labeled

tool_bipolar_labeled tool_irrigator_labeled tool_scissors_labeled

tool_specimenBag_labeled

To have access to our training and testing data for the surgical tool detection challenge, please kindly contact us by filling this form. You will be added to our mailing list and the link to download the dataset will be sent to you once it is available.

To submit your results, please go to the submission page.