PROJECT

Development of an artificial intelligence system for the discovery and diagnosis of colon polyps

 

MAIN DESCRIPTION

○ To develop an artificial intelligence system for colonoscopy in the form of open software which assists the discovery of colon polyps and prediction of tissue types using the most widely used colonoscopy systems (non-magnified imaging and white light imaging systems)

 

○ Colon polyp detection system

- Develop a system that quickly recognizes polyps that enter the visual field of fast-moving endoscopy during real-time colonoscopy with artificial intelligence and transmits them to the screen

- The level and degree of artificial intelligence's perception of flat polyps of less than 5mm show the excellence of the system, and the performance of accurately distinguishing polyps that enter the screen even in poor bowel cleanliness with feces could be checked at the same time

○ Colon polyp diagnosis system

- When developing the system, the tissue type of polyps is divided into high confidence (90% or more possibility of predicting correct answers) and low confidence

- In the case of high confidence among the judgments of all polyps, high artificial intelligence success is required to reach 95% or more in accuracy and sensitivity. In the case of low confidence, the diagnostic results are not evaluated

- As the performance of artificial intelligence gradually advances, it aims to develop artificial intelligence for reliable optical diagnosis even with the current general colonoscopy by gradually increasing the diagnosis rate of high confidence