Data Description¶
The data provided is a combination of endoscopic video and sensor data
from medical devices in an integrated operating room (Karl Storz OR1).
The data was obtained during laparoscopic surgeries at the University
Hospital of Heidelberg, a center of excellence for minimally invasive
surgery. All surgeries were annotated framewise for surgical phases by a
surgeon.
The surgeries recorded are colorectal surgeries, namely of type rectal
resection, sigmoid resection or proctocolectomy. In contrast to
surgeries such as cholecystectomy, these procedures are more complex.
This means there are more variations regarding the surgical strategy
(length or order of phases) and phases may change repeatedly. Also,
every procedure includes an extra-abdominal phase (bowel anastomosis,
the connection of two parts of bowel) that is prepared without use of
the laparoscope.
The dataset consists out of 30 different recorded surgeries. For each surgery, the video captured by the endoscope is provided. To ensure anonymity, frames corresponding to extra-abdominal views are censored by entire blue (RGB 0 0 255) frames. Furthermore, the data signals from medical devices are included in a csv file for every surgery. Each line in a csv file corresponds to a frame in the laparoscopic video.
The data will be released in three different sets: 2 training sets (the first set containing 24 videos and the second set containing probably 9 videos), which include framewise phase annotation and minute-wise instrument usage annotation and 2 testing set (first set consisting of 6 videos and the second probably 3 videos). If you participate in the challenge last year, please also download the dataset, as some changes have been made.
Reference Standard¶
The following phases are annotated in the recorded surgeries by surgeons:
| Phase ID | Phase |
|---|---|
| 0 | Preparation and orientation at abdomen |
| 1 | Dissection of lymphnodes and blood vessels |
| 2 | Retroperitoneal preparation to lower pancreatic border |
| 3 | Retroperitoneal preparation of duodenum and pancreatic head |
| 4 | Mobilizing the sigmoid and the descending colon |
| 5 | Mobilizing the spenic flexure |
| 6 | Mobilizing the tranverse colon |
| 7 | Mobilizing the ascending colon |
| 8 | Dissection and resection of rectum |
| 9 | Preparing the anastomosis extraabdominally |
| 10 | Preparing the anastomosis intraabdominally |
| 11 | Placing stoma |
| 12 | Finishing the operation |
| 13 | Exception (will be ignored during evaluation) |
Minute-wise instrument usage has also been annotated by surgical experts. This annotation can be used as additional information during training. No instrument usage data will be supplied for the test data. The following tools were annotated:
| Tool ID | Instrument |
| 0 | Grasper |
| 1 | Harmonic scalpel |
| 2 | J-Hook |
| 3 | Ligasure |
| 4 | Scissors |
| 5 | Stapler |
| 6 | Aspirator |
| 7 | Swapholder |
| 8 | Silicone drain |
| 9 | Clip applier |
| 10 | I-Hook |
| 11 | Needle Holder |
For evaluation of submitted results, standard metrics such as precision, recall, accuracy and the Jaccard index will be used. Participants are free to use third party public data to augmented the provided training data. Participants are encouraged, but not required, to also submit results without pretraining to allow future comparisons.