Inter-observer Agreement in Landmark and Flexure Identification in Colon Capsule Endoscopy

Aim: Identification of landmarks are important in colon capsule endoscopy (CCE) as guidance for a subsequent
conventional colonoscopy (CC), but also as a quality indicator in the evaluation of bowel preparation per segment and
complete CCE. When a CC is carried out, a Scope Guide can assist the endoscopist in determining the localization. In
CCE this support is not available. The aim of this study was to investigate the inter-observer agreement in landmark and
flexure identification in CCE.
Method: The landmarks for CCE are first cecal image, hepatic flexure, splenic flexure and last rectal image. An interobserver study was carried out comparing the landmark in CCE investigations between four CCE readers. The group of
CCE readers was composed of three medical doctors with different levels of experience in CCE and the initial evaluation
given in by the external company who by routine performs CCE reading in our hospital. All CCE investigations analyzed
in this study were carried out as a part of the Danish screening program for colorectal cancer. A random sample of 20
CCE investigations was taken from the total sample of 856 videos.
Results: We found an overall inter-observer agreement on all landmarks of 51%. Inter-observer agreement on first cecal
image, hepatic flexure, splenic flexure and last rectal image were 72%, 29%, 22% and 83%, respectively. The overall
inter-observer agreement including only examinations with adequate bowel preparations (n=16) were 54% and for
individual landmarks 73%, 32%, 24% and 85%.
Conclusion: We found a poor overall inter-observer agreement (51%) on landmarks from CCE. The lowest inter-observer
agreements were found for the splenic flexure (22%) and hepatic flexure (28%). Measures are needed to improve the
landmark identification from CCE videos. Artificial intelligence could be a possible solution.