The database gathers data of the colonoscope’s coordinates from electromagnetic coils inside the colonoscopes path through the colon. The data consists of coordinates from 1400 full clinical colonoscopies and 100 full colonoscopies from a simulated setting. These data are available for free, and can be used for mapping the mucosal inspection of the colon, generate heatmaps to ensure an equally distributed inspection, etc. Freely available for other researchers and scientists to use.
The advantages
The Copenhagen Colonoscopy Database goes beyond previous databases by not only relying on image frames but recordings of the full 3D Coordinates of the Colonoscope inside the colon with timestamps. This enables precise mapping of scope movement, tracking of exact position and orientation, flexibility and anatomy of the colon during real procedures.
Current datasets focus on polyp detection via Image recognition. With the new dataset researchers can map which areas of the colon that have been missed or inspected, and generate heat maps showing coverage, and reduce miss rates. Furthermore, novel AI and QA tools that go beyond mage recognition, by enabling the Colonoscopy Retraction Score (CoRS) and Colonoscopy Progression Score (CoPS), and by evaluating scope handling, withdrawal technique or inspection completeness, making it possible to rate patient discomfort etc.
The datasets from the Copenhagen Colonoscopy Database fills a gap by providing comprehensive spatial-temporal data on scope movement and mucosal coverage, enabling entirely new dimensions of quality assurance, AI training, procedural evaluation, and anatomical research that were not possible with prior datasets focused on images or clinical outcomes alone.
Authors
Kristoffer Mazanti Cold, Anishan Vamadevan, Amihai Heen , Andreas Slot Vilmann, Mustafa Bulut, Bojan Kovacevic, Morten Rasmussen, Lars Konge, Morten Bo Søndergaard Svendsen
See the article in Nature Journal: “Scientific Data” here.