Mairin Joseph-Talreja, MD1, Sabrina Rosengarten, MPH1, Manuel Martinez, MD, FACG2; 1SUNY Downstate, Brooklyn, NY; 2VA NY Harbor Health Care System, Brooklyn, NY
Introduction: Capsule endoscopy, when utilized in the appropriate clinical setting, can help identify various pathologies in patients who may not tolerate a traditional endoscopic approach. However, transit times can vary, and a capsule study can easily reach upwards of 120 minutes of video content that requires review. Analytical software has been developed to extract repetitive frames, permitting shorter review times. The software algorithm analyzes several image frames that are deemed to be similar and only one representative image is selected and displayed on screen. Utilization of this method has been proven to significantly reduce reading times. This retrospective study sought to evaluate if the use of this type of algorithmic software is a reliable method for abbreviating capsule review times without sacrificing the accuracy of the examination. Methods: A random set of 51 temporally consecutive endoscopic capsule studies were selected. These included 40 small bowel, 7 esophageal and 4 colon capsule endoscopies. Each of these studies had been originally reviewed using the traditional non-algorithmic setting and read in their entirety. A repeat read for the purpose of this study was performed, using the algorithmic software setting. Each study was reviewed and interpreted by one expert capsule reader who was blinded to the original reading results and interpretations. The hypothesis was that there would be no additional pathology identified using the algorithmic software setting on the repeat read. Results: Out of the 50 included studies, four (8%) exams (3 small bowel and 1 esophageal) had new findings documented on review with the algorithmic settings activated (95% CI 2.2%-19.2%). There were also four studies (2 small bowel, 1 esophageal, and 1 colon) with findings on the original read that were missed on repeat. None of the new or missed findings warranted a change in clinical management. Discussion: Software assisted capsule endoscopy image review can dramatically reduce capsule review times, but its use has not yet been validated. This study unexpectedly identified new pathology from some capsule reads using the algorithmic setting in 8% of the studies reviewed. We consider that this may have been due to improved concentration facilitated by a shorter review time or simply by the Hawthorne effect. Larger studies are warranted to validate the utilization of the algorithm software setting to assist in capsule endoscopy review and interpretation.
Disclosures: Mairin Joseph-Talreja indicated no relevant financial relationships. Sabrina Rosengarten indicated no relevant financial relationships. Manuel Martinez: Medtronics – Consultant.