Post written by Ji Young Lee, MD, PhD, from the Health Screening and Promotion Center, Asan Medical Center, College of Medicine, University of Ulsan, Seoul, South Korea.

This study demonstrated objective assessment of bowel preparation adequacy using an artificial intelligence (AI)/convolutional neural network algorithm developed from colonoscopy videos.

An adequate bowel preparation impacts quality indicators such as adenoma detection rates and cecal intubation rate, and a reliable assessment of bowel preparation is important for colonoscopy-based research.

Although we have a few validated bowel preparation scales, these tools have limitations. There may be interobserver variability in the decision of degree of cleanliness. Because bowel cleanliness scoring is performed after the procedure, the scoring may be inaccurate and cannot improve quality of colonoscopy examination in real time.

We hypothesized that the high reproducibility of AI would allow an objective and consistent assessment of bowel preparation in hopes of eventually enabling real-time feedback of bowel cleanliness.

We validated the developed algorithm using 252 10-second video clips, an area under curve of .918, and accuracy of 85.3% for detection of inadequate bowel preparation. To test the algorithm performance, we compared the interobserver variability of 4 endoscopists to 4 endoscopists plus the algorithm using kappa statics.

In 10-second video clip testing, 4 experts’ degree of agreement was .858, and 4 experts plus the algorithm’s agreement was decreased to .733 for the assessment of bowel preparation.

However, in the withdrawal colonoscopy video testing, the degree of agreement was substantial for both 4 experts only and 4 experts plus the algorithm (Fleiss kappa = .694 vs .649 [Figure 3]).

Our study showed that the algorithm could give feedback on bowel cleanliness in real time during colonoscopy and provide the integrated Boston bowel preparation scale for full withdrawal colonoscopies without significant differences to the experts. 

Future studies should focus on the real-time application of our algorithm during colonoscopy to see if the algorithm’s feedback will be able to improve other colonoscopy quality indicators such as adenoma detection rate and cecal intubation rate. 

Interobserver agreement of bowel cleanliness between experts and the algorithm. AI-BBPS, Algorithm for assessment of Boston Bowel Preparation Scale; L, left-sided colon segment; R, right-sided colon segment; T, transverse colon.

Read the full article online.

The information presented in Endoscopedia reflects the opinions of the authors and does not represent the position of the American Society for Gastrointestinal Endoscopy (ASGE). ASGE expressly disclaims any warranties or guarantees, expressed or implied, and is not liable for damages of any kind in connection with the material, information, or procedures set forth.