ARTIFICIAL INTELLIGENCE AND CAPSULE COLONOSCOPY: AUTOMATIC ASSESSMENT OF INTESTINAL PREPARATION SCORES/DEGREE

Introduction Medicine is evolving toward increasingly less invasive diagnostic modalities, with Gastroenterology being a prime example of this trend. The diagnostic assessment of the colon using capsule endoscopy, through specific protocols for colon capsule endoscopy or as part of panendoscopic evaluations, has gained growing importance as a first-line diagnostic method. Proper bowel preparation is essential for conclusive examinations, given the non-invasive nature of capsule endoscopy—which does not allow for interventional measures to improve mucosal visibility. Several scales have already been developed to classify bowel preparation for colon capsule endoscopy. However, their applicability is limited by inter-observer variability.

Objective The aim of this study was to develop a deep-learning algorithm for the automatic classification of colon bowel preparation.

Results The algorithm achieved highly satisfactory results, with sensitivity, specificity, and diagnostic accuracy of 91%, 97%, and 95%, respectively. The area under the curve ranged from 0.92 to 0.97.

Conclusion This study is pivotal for the future widespread use of capsule endoscopy in colon evaluation and for minimally invasive panendoscopic modalities.

No conflict of interest has been declared by the author(s).

Publication History

Article published online:
25 April 2025