KID Datasets

Description

This is an open access, non-profit database of high quality annotated wireless capsule endoscopy (WCE) images and videos. It aims to provide data and knowledge to aid and educate clinicians, and also for the development of automated medical decision support systems. We recognize the importance of annotated data for substantial progress in research on intelligent systems’ development and therefore encourage the provision of annotated files using RATSNAKE or similar tools. KID welcomes contributions from WCE researchers around the world as long as data provided are entirely anonymous with any potential identifiers redacted.

The high quality of the accumulated data and annotations is guaranteed by expert reviewers from the international scientific committee of KID.

If you have found this resource useful, please use the following reference to cite this project:

A. Koulaouzidis, D. K. Iakovidis, D. E. Yung, E. Rondonotti, U. Kopylov, J. N. Plevris, E. Toth, A. Eliakim, G. W. Johansson, W. Marlicz, G.W. Johansson, W. Marlicz, G. Mavrogenis, A. Nemeth, H. Thorlacius, G.E., Tontini, “KID Project: an internet-based digital video atlas of capsule endoscopy for research purposes,” Endoscopy International Open, vol. 5, no. 06, pp. E477–E483, 2017.

KID is based on annotated, anomymous image and video datasets contributed by a growing international community. Currently the following datasets are publicly available through the established Kaggle platform (https://www.kaggle.com) for research purposes.

KID Dataset 1

A total of 77 wireless capsule endoscopy (WCE) images obtained using MiroCam® (IntroMedic Co, Seoul, Korea) capsule endoscopes. These images illustrate various types of abnormalities, including angioectasias, apthae, chylous cysts, polypoid lesions, villous oedema, bleeding, lymphangiectasias, ulcers and stenoses.

[1] D.K. Iakovidis, and A. Koulaouzidis, “Automatic lesion detection in capsule endoscopy based on color saliency; closer to an essential adjunct for reviewing software,” Gastrointestinal Endoscopy, vol. 80, no. 5, pp. 877-883, 2014, doi:10.1016/j.gie.2014.06.026
[2] D.K. Iakovidis, A. Koulaouzidis, “Automatic Lesion Detection in Wireless Capsule Endoscopy – A Simple Solution for a Complex Problem,” in Proc. IEEE International Conference on Image Processing (ICIP), Paris, France, 2014, pp.2236-2240
[3] A. Koulaouzidis, D. K. Iakovidis, D. E. Yung, E. Rondonotti, U. Kopylov, J. N. Plevris, E. Toth, A. Eliakim, G. W. Johansson, W. Marlicz, G.W. Johansson, W. Marlicz, G. Mavrogenis, A. Nemeth, H. Thorlacius, G.E., Tontini, “KID Project: an internet-based digital video atlas of capsule endoscopy for research purposes,” Endoscopy International Open, vol. 5, no. 06, pp. E477–E483, 2017.

KID Dataset 2

A total of 2371 wireless capsule endoscopy (WCE) images obtained using MiroCam® (IntroMedic Co, Seoul, Korea) capsule endoscopes. These images illustrate assorted small bowel findings including polypoid, vascular and, inflammatory lesions. This dataset also includes normal images from the esophagus, stomach, small bowel and colon.

[1] D.K. Iakovidis, S.V. Georgakopoulos, M. Vasilakakis, A. Koulaouzidis, and V. Plagianakos, “Detecting and Locating Gastrointestinal Anomalies Using Deep Learning and Iterative Cluster Unification,” IEEE Transactions on Medical Imaging, 2018, doi:10.1109/TMI.2018.2837002
[2] A. Koulaouzidis, D. K. Iakovidis, D. E. Yung, E. Rondonotti, U. Kopylov, J. N. Plevris, E. Toth, A. Eliakim, G. W. Johansson, W. Marlicz, G.W. Johansson, W. Marlicz, G. Mavrogenis, A. Nemeth, H. Thorlacius, G.E., Tontini, “KID Project: an internet-based digital video atlas of capsule endoscopy for research purposes,” Endoscopy International Open, vol. 5, no. 06, pp. E477–E483, 2017.

KID Dataset – Video 1

Wireless capsule endoscopy video obtained using a MiroCam® (IntroMedic Co, Seoul, Korea) capsule endoscope.
This dataset includes only weak annotations.

[1] D.K. Iakovidis, S.V. Georgakopoulos, M. Vasilakakis, A. Koulaouzidis, and V. Plagianakos, “Detecting and Locating Gastrointestinal Anomalies Using Deep Learning and Iterative Cluster Unification,” IEEE Transactions on Medical Imaging, 2018, doi:10.1109/TMI.2018.2837002
[2] A. Koulaouzidis, D. K. Iakovidis, D. E. Yung, E. Rondonotti, U. Kopylov, J. N. Plevris, E. Toth, A. Eliakim, G. W. Johansson, W. Marlicz, G.W. Johansson, W. Marlicz, G. Mavrogenis, A. Nemeth, H. Thorlacius, G.E., Tontini, “KID Project: an internet-based digital video atlas of capsule endoscopy for research purposes,” Endoscopy International Open, vol. 5, no. 06, pp. E477–E483, 2017.

KID Dataset – Video 2

Wireless capsule endoscopy video obtained using a MiroCam® (IntroMedic Co, Seoul, Korea) capsule endoscope.
This dataset does not include annotations yet.

[1] D.K. Iakovidis, and A. Koulaouzidis, “Software for Enhanced Video Capsule Endoscopy: State of the Art and Challenges for Essential Progress,” Nature Reviews Gastroenterology & Hepatology, vol. 12, no 3, pp. 172-186, 2015, 10.1038/nrgastro.2015.13
[2] A. Koulaouzidis, D. K. Iakovidis, D. E. Yung, E. Rondonotti, U. Kopylov, J. N. Plevris, E. Toth, A. Eliakim, G. W. Johansson, W. Marlicz, G.W. Johansson, W. Marlicz, G. Mavrogenis, A. Nemeth, H. Thorlacius, G.E., Tontini, “KID Project: an internet-based digital video atlas of capsule endoscopy for research purposes,” Endoscopy International Open, vol. 5, no. 06, pp. E477–E483, 2017.

KID Dataset – Video 3

Wireless capsule endoscopy video obtained using a MiroCam® (IntroMedic Co, Seoul, Korea) capsule endoscope.
This dataset does not include annotations yet.

[1] D.K. Iakovidis, and A. Koulaouzidis, “Software for Enhanced Video Capsule Endoscopy: State of the Art and Challenges for Essential Progress,” Nature Reviews Gastroenterology & Hepatology, vol. 12, no 3, pp. 172-186, 2015, doi:10.1038/nrgastro.2015.13
[2] A. Koulaouzidis, D. K. Iakovidis, D. E. Yung, E. Rondonotti, U. Kopylov, J. N. Plevris, E. Toth, A. Eliakim, G. W. Johansson, W. Marlicz, G.W. Johansson, W. Marlicz, G. Mavrogenis, A. Nemeth, H. Thorlacius, G.E., Tontini, “KID Project: an internet-based digital video atlas of capsule endoscopy for research purposes,” Endoscopy International Open, vol. 5, no. 06, pp. E477–E483, 2017.

Scientific Committee

The International Scientific Committee of KID reviews the quality of submitted data from contributors worldwide. It has the authority to accept, request revisions or if necessary reject publication of data of insufficient quality. Its members are listed below in alphabetical order.

Tools

  • RATSNAKE Rapid image annotation tool
  • MAN Tool for merging image annotations
  • ART Image anonymization and renaming tool
  • VIDECOOL Tool for decomposition of a video into its frames
  • CAMSTUDIO Tool for capturing high quality videos from screen (use lossless video codec)
  • JVA Java video analysis framework (for software developers)

Register

In support of open science, KID has moved on Kaggle (https://www.kaggle.com). Kaggle is an established platform where both data and computer programs can be shared publicly. The previous KID database has been discontinued and all user data have been deleted; therefore, previously registered users must be re-registered in Kaggle by following the procedure described below. Registrations are accepted only for academic research.

The new Kaggle-based KID supports 3 user groups for enhanced user experience:
(a) Computer Scientists;
(b) Medical Doctors;
(c) Data Providers.

The exact registration process will be announced soon.

Registration requests should be sent by email to mdss-info [at] uth.gr. The email requesting registration should be sent only from your academic institutional account. Registration requests through gmail, hotmail or similar are not acceptable. Please include all the following information:
(i) Your full name;
(ii) The name and the address of your organization as well as a link to the public webpage of the research group to which you are a member.
(iii) Your username on Kaggle. Each user must be already registered in Kaggle.
(iv) The purpose of registering to KID. Users requesting access to current KID datasets, should specify exactly which datasets would like to use, e.g. “KID Dataset 2” and why.

Since KID includes contributions from various data providers, the registration requests for accessing each specific dataset are directly forwarded to the respective data providers.
Any information sent to the above email regarding KID datasets will be used only for the purposes of communication about the KID datasets.

All Data Providers are solely responsible for their datasets, i.e., they must have proper authorizations (it is a repsonsibility of each data provider to be knowledgable for and have obtained the necessary authorizations), who are eventually responsible for accepting or rejecting each application.

Login

In support of open science, KID has moved on Kaggle (https://www.kaggle.com/). Kaggle is an established platform where both data and computer programs can be shared publicly. If you have correctly followed the registration steps described above, you can access the respective KID datasets by clicking on the following links: