A multisystem-compatible deep learning-based algorithm for detection and characterization of angiectasias in small-bowel capsule endoscopy. A proof-of-concept study. Issue 12 (December 2021)

• Record Type: Journal Article
• Authors: Houdeville, Charles; Souchaud, Marc; Leenhardt, Romain; Beaumont, Hanneke; Benamouzig, Robert; McAlindon, Mark; Grimbert, Sylvie; Lamarque, Dominique; Makins, Richard; Saurin, Jean-Christophe; Histace, Aymeric; Dray, Xavier
• Is Part Of: Digestive and liver disease. Volume 53:Issue 12(2021)
• Page Start: 1627
• Page End: 1631
• Publication Date: 2021-12
• Subjects: Artificial intelligence -- Capsule endoscopy -- Deep learning -- Small bowel; AI Artificial Intelligence -- CD Crohn's disease -- CE capsule endoscopy -- DL deep learning -- GI gastrointestinal -- GIA gastrointestinal angiectasia (or angiodysplasia) -- ms millisecond -- NPV Negative Predictive Value -- OGIB obscure gastrointestinal bleeding -- PPV Positive Predictive Value -- ROI Region of Interest -- SB small bowel -- SBCE small bowel capsule endoscopy -- Se sensitivity -- Sp specificity -- TP True Positive; Digestive organs -- Diseases -- Periodicals; Liver -- Diseases -- Periodicals; 616.33005
• DOI: 10.1016/j.dld.2021.08.026
• Languages: English
• ISSNs: 1590-8658

Abstract

Abstract: Background and aims: Current artificial intelligence (AI)-based solutions for capsule endoscopy (CE) interpretation are proprietary. We aimed to evaluate an AI solution trained on a specific CE system (Pillcam®, Medtronic) for the detection of angiectasias on images captured by a different proprietary system (MiroCam®, Intromedic). Material and methods: An advanced AI solution (Axaro®, Augmented Endoscopy), previously trained on Pillcam® small bowell images, was evaluated on independent datasets with more than 1200 Pillcam® and MiroCam® still frames (equally distributed, with or without angiectasias). Images were reviewed by experts before and after AI interpretation. Results: Sensitivity for the diagnosis of angiectasia was 97.4% with Pillcam® images and 96.1% with Mirocam® images, with specificity of 98.8% and 97.8%, respectively. Performances regarding the delineation of regions of interest and the characterization of angiectasias were similar in both groups (all above 95%). Processing time was significantly shorter with Mirocam® (20.7 ms) than with Pillcam® images (24.6 ms, p <0.0001), possibly related to technical differences between systems. Conclusion: This proof-of-concept study on still images paves the way for the development of resource-sparing, "universal" CE databases and AI solutions for CE interpretation.