DDW ePoster Library

ARTIFICIAL INTELLIGENCE FOR POLYP SIZE IN COLONOSCOPY: FELLOWS VERSUS FACULTY
DDW ePoster Library. Chela H. 05/22/22; 354912; Su1437
Dr. Harleen Chela
Dr. Harleen Chela
Contributions
Live Questions & Answers
Abstract
Submit your question(s) to the author at anytime.
Number: Su1437
ARTIFICIAL INTELLIGENCE FOR POLYP SIZE IN COLONOSCOPY: FELLOWS VERSUS FACULTY

Society: ASGE
Track: Colorectal Diseases

Author(s): Olalekan Akanbi1, Harleen K. Chela1, Matthew L. Bechtold1

Institution(s): 1. Medicine, University of Missouri - Columbia, Columbia, MO, United States.

Introduction: Artificial intelligence (AI) appears to be beneficial throughout the field of medicine. In gastroenterology, AI may significantly improve polyp identification, adenoma detection rates, and polyp detection rates. Due to a polyp's size significantly impacts surveillance timing and level of training or experience affects size estimation, AI may be used to accurately estimate size of polyps. Therefore, we performed a survey study on the use of AI for estimating polyp size between faculty and fellows.

Methods: A survey study was performed in November and December 2021 using a colon endoscopy phantom model. Artificial colon polyps were created (rubber-based or Play-Doh), measured with caliper, and placed in a colon phantom. Using a high definition sigmoidoscope, 11 videos were made in the colon phantom. In a single academic center, gastroenterology faculty and fellows were surveyed on the estimation of the size of the polyp in each of the videos. A newly AI system (Argus - EndoSoft - New York) was used as well and compared to the physicians for accuracy (by 2 methods) and impact on the timing of surveillance.

Results: Polyps were created, placed in the colon phantom, and videos were filmed (n=11). Faculty volunteered and performed the survey (n=4) with mean years of age 53 ± 11.2 and mean years of experience 19.3 ± 9.2. Fellows volunteered and performed the survey (n=7) with mean years of age 34.7 ± 2.3. Accuracy rates for all participants were 74% median (range 48-88%) as compared to 96% for Argus. Fellows appeared to have a higher accuracy rate than faculty (75% vs 71%). All the participants were within ± 1 mm range on the size estimation 48 times (40%) versus 9 times (82%) with Argus. Fellows appeared to be within ± 1 mm range on the size estimation more than faculty (44% vs 36%). Based on current guidelines, all participants' surveillance recommendations based on polyp size were significantly more incorrect as compared to Argus (34 vs 0) with 24 recommending too short (20%) and 10 recommending too long (6%) of interval for next colonoscopy. Faculty appeared to recommend too short of interval more than the fellows (10/44, 23% vs 14/77, 18%).

Discussion: AI is more accurate at estimating polyp size than experienced gastroenterologists and fellows in training. Fellows appear to be more accurate at estimating polyp size than the faculty. Given the variation in sizing between the entire group and since the size  1 cm impacts surveillance interval with both faculty and fellows being too short 20% of the time, AI may help all adhere to appropriate surveillance intervals.
Number: Su1437
ARTIFICIAL INTELLIGENCE FOR POLYP SIZE IN COLONOSCOPY: FELLOWS VERSUS FACULTY

Society: ASGE
Track: Colorectal Diseases

Author(s): Olalekan Akanbi1, Harleen K. Chela1, Matthew L. Bechtold1

Institution(s): 1. Medicine, University of Missouri - Columbia, Columbia, MO, United States.

Introduction: Artificial intelligence (AI) appears to be beneficial throughout the field of medicine. In gastroenterology, AI may significantly improve polyp identification, adenoma detection rates, and polyp detection rates. Due to a polyp's size significantly impacts surveillance timing and level of training or experience affects size estimation, AI may be used to accurately estimate size of polyps. Therefore, we performed a survey study on the use of AI for estimating polyp size between faculty and fellows.

Methods: A survey study was performed in November and December 2021 using a colon endoscopy phantom model. Artificial colon polyps were created (rubber-based or Play-Doh), measured with caliper, and placed in a colon phantom. Using a high definition sigmoidoscope, 11 videos were made in the colon phantom. In a single academic center, gastroenterology faculty and fellows were surveyed on the estimation of the size of the polyp in each of the videos. A newly AI system (Argus - EndoSoft - New York) was used as well and compared to the physicians for accuracy (by 2 methods) and impact on the timing of surveillance.

Results: Polyps were created, placed in the colon phantom, and videos were filmed (n=11). Faculty volunteered and performed the survey (n=4) with mean years of age 53 ± 11.2 and mean years of experience 19.3 ± 9.2. Fellows volunteered and performed the survey (n=7) with mean years of age 34.7 ± 2.3. Accuracy rates for all participants were 74% median (range 48-88%) as compared to 96% for Argus. Fellows appeared to have a higher accuracy rate than faculty (75% vs 71%). All the participants were within ± 1 mm range on the size estimation 48 times (40%) versus 9 times (82%) with Argus. Fellows appeared to be within ± 1 mm range on the size estimation more than faculty (44% vs 36%). Based on current guidelines, all participants' surveillance recommendations based on polyp size were significantly more incorrect as compared to Argus (34 vs 0) with 24 recommending too short (20%) and 10 recommending too long (6%) of interval for next colonoscopy. Faculty appeared to recommend too short of interval more than the fellows (10/44, 23% vs 14/77, 18%).

Discussion: AI is more accurate at estimating polyp size than experienced gastroenterologists and fellows in training. Fellows appear to be more accurate at estimating polyp size than the faculty. Given the variation in sizing between the entire group and since the size  1 cm impacts surveillance interval with both faculty and fellows being too short 20% of the time, AI may help all adhere to appropriate surveillance intervals.

By clicking “Accept Terms & all Cookies” or by continuing to browse, you agree to the storing of third-party cookies on your device to enhance your user experience and agree to the user terms and conditions of this learning management system (LMS).

Cookie Settings
Accept Terms & all Cookies