ASU-Mayo Clinic Colonoscopy Video (c) Database

ASU-Mayo Clinic Colonoscopy Video (c) Database is the first, largest, and a constantly growing set of short and long colonoscopy videos, collected and de-identified at the Department of Gastroenterology at Mayo Clinic in Arizona. Each frame in this database comes with a ground truth image or a binary mask that indicates the polyp region. If a frame contains no polyp, the corresponding ground truth image will be completely black. The ground truth images for the collected videos have been created by volunteer students at Arizona State University, and have been reviewed and corrected by a trained expert.

Training dataset

We have selected 20 short colonoscopy videos from ASU-Mayo Clinic Colonoscopy Video (c) Database, of which 10 videos have a unique polyp inside (positive shots) and the other 10 videos have no polyps (negative shots). The videos are selected so as to display maximum variation in colonoscopy procedures: Some videos are high resolutions but some are recorded in lower resolution, some videos display a careful colon examination while others show a hasty colon inspection, finally some videos have biopsy instruments in them and still others have a “play” logo at the top-right of the frames. Figure 1 shows examples of polyps in the training set along with the corresponding ground truth images.

Fig 1: Example of 10 different polyps from ASU-Mayo database with their corresponding ground truth images.

The training videos are named depending on whether they have a polyp inside or not. A video with a polyp inside has “_wp_” in its name, and a video with no polyps inside has “_np_” in its name. Each training video is placed in a folder with the same name as the video. The ground truth images are placed next to the video file in a subfolder named “GT”. For instance, the video named “ShortVD_wp_2.wmv”, has a polyp inside and can be found in the folder named “ShortVD_wp_2”. The corresponding ground truth can also be found at .\ ShortVD_wp_2\GT\. The following Table details the number of frames in each of the training videos:

Table 1: Description of the content of ASU-Mayo training database.

Testing dataset

This database contains 18  different videos without annotated ground truth, in order to assess the performance of the algorithms designed as part of 2015 ISBI Grand Challenge on Automatic Polyp Detection in Colonoscopy Videos. 

Terms of use

The ASU-Mayo Clinic Colonoscopy Video (c) Database is copyrighted. If you are interested in using this database, please contact Prof. Jianming Liang at Arizona State University.

Citation

Reference must be made to the following publications:

• N. Tajbakhsh, S. R. Gurudu and J. Liang, "Automated Polyp Detection in Colonoscopy Videos Using Shape and Context Information," in IEEE Transactions on Medical Imaging, vol. 35, no. 2, pp. 630-644, Feb. 2016.

We would also appreciate if you dropped Prof. Jianming Liang at Arizona State University an email to inform ASU of any publication using this dataset, so we can point to your publication on this website.