Search

Presence of intestinal content in capsule endoscopy (CE) procedures reduces quality of visualization with consequent negative impact in its diagnostic yield. The aim of our study was to validate a tool for automatic detection of intestinal content in CE procedures. Unidentified mpg type videos were extracted from 30 consecutive CE procedures. One image every minute was selected from each video for evaluation. Images were analysed randomly with an automatic tool capable of distinguishing between clean images and images with intestinal content both by analysing whole images and by dividing the images into segments. The same images were evaluated twice by two experienced CE readers who were blinded to automatic tool results. The first evaluation was performed in segments in a similar way as the automatic tool analysis and the second was performed in complete images applying a grading scale. The created automatic tool for the evaluation of cleanliness of the small bowel in CE shows high precision with an acceptable interobserver agreement. This tool has demonstrated that it is reproducible, objective and consistent. Future application of its benefits can be used in a clinical trial that compares different bowel preparations.