P0338 - The Evaluation of Artificial Intelligence on Adenoma Detection Rate in a Latin American Community Setting

Cynthia Contreras, MD¹, Cesar Ortiz, MD², Paola Peña, MD², Ana Finke, MD², Sarah Herrera, MD², Nicole Peña, MD², Leandro Tapia, MD³, Fernando Contreras, MD², Maria Alonso, MD^4
1TriHealth, Cincinnati, OH; ²Centro de Gastroenterologia Avanzada, Santo Domingo, Distrito Nacional, Dominican Republic; ³Hospital Germans Trias i Pujol, Barcelona, Catalonia, Spain; ⁴Centro de Gastroenterologia avanzada, Santo Domingo, Distrito Nacional, Dominican Republic

Introduction: The potential of Computer Assisted Diagnosis (CADe) to enhance the adenoma detection rate (ADR) during colonoscopy has generated growing interest within the gastroenterology field. In Latin America, where the incidence of colorectal cancer is on the rise, prevention and early detection are of utmost importance. Research has established that a 1% rise in adenoma detection rate (ADR) corresponds to a 3% decline in colorectal cancer (CRC) occurrence and a 5% reduction in mortality.

This study aims to evaluate the efficacy of Artificial Intelligence (AI) in improving adenoma and polyp detection rates during colonoscopy and the potential impact of CADe colonoscopy on the prevention and early detection of colorectal cancer.

Methods: Colonoscopies that used CADe and colonoscopies that didn’t use CADe were compared. ADR was calculated based on the total number of procedures that found at least one adenoma during a screening colonoscopy. Also, we calculated the overall ADR and overall PDR dividing each one by the total number of procedures. We also did calculate the adenomas found per colonoscopy. The polyps were divided based on CADe or Non-CADe and calculated mean size and age using a t-test, assuming significance when p< 0.05.

Results: The study included 624 patients, with CADe being utilized in 48.4% (n=302) of all procedures. ADR did not improve with CADe implementation (61% vs. 61%, p=0.98). CADe increased overall PDR compared to non-CADe procedures (62.25% vs. 59.9, p=0.04) but not overall ADR (56% vs. 57%, p=0.79). CADe did not lead to an increase in Adenomas per Colonoscopy when compared to non-CADe procedures (0.84 vs 0.97, p=0.14). CADe proved to be more useful in the visualization of smaller polyps than non-CADe procedures (3.5 vs. 4.4mm, p=0.0001) and in detecting smaller adenomas (3.62 vs. 4.5mm, p=0.004).

Discussion: The study failed to show a statistically significant difference in the adenomas or polyps detection rate between the computer-aided detection (CADe) group and the non-CADe group; this could be attributed to the fact that our study group already exhibited a much higher adenoma detection rate (61%) compared to the accepted standard. To evaluate whether AI can enhance adenoma detection, further studies should involve colonoscopists with lower adenoma detection rates. It is worth noting that the use of CADe did aid in the identification of smaller lesions.


Disclosures:


Cynthia Contreras, MD¹, Cesar Ortiz, MD², Paola Peña, MD², Ana Finke, MD², Sarah Herrera, MD², Nicole Peña, MD², Leandro Tapia, MD³, Fernando Contreras, MD², Maria Alonso, MD⁴. P0338 - The Evaluation of Artificial Intelligence on Adenoma Detection Rate in a Latin American Community Setting, ACG 2023 Annual Scientific Meeting Abstracts. Vancouver, BC, Canada: American College of Gastroenterology.