P0357 - Artificial Intelligence in Colonoscopy in the Community Setting

Sunday, October 22, 2023

3:30 PM – 7:00 PM PT

Location: Exhibit Hall

Has Audio

Presenting Author(s)

Kunaal Patel, MD, MS, MBA

Trident Medical Center-HCA
North Charleston, SC

Kunaal Patel, MD, MS, MBA¹, Brett VanLeer-Greenberg, MD²
¹Trident Medical Center-HCA, North Charleston, SC; ²Trident Medical Center-HCA, Summerville, SC

Introduction: Artificial Intelligence (AI), specifically machine learning, has had a growing role in the field of endoscopy. AI systems in colonoscopy augment real time polyp detection, with studies showing up to a 14.4% increase in absolute Adenoma Detection Rate (ADR). An increase in ADR results in a lower risk of interval colon cancer. Some estimates show that a 1% increase in ADR lowers the risk of interval colon cancer by up to 6%. AI can serve as an effective second endoscopist, positively impacting the ADR, with the largest improvements in low detector endoscopists. We aim to assess the impact of the use of AI in colonoscopy on the ADR in a community-based outpatient setting.

Methods: We conducted a single-center retrospective chart review study of all patients undergoing a screening colonoscopy in an outpatient community setting. Colonoscopy reports from two endoscopists and correlating pathology reports from five months before and after the implementation of the AI system were reviewed. We included 1,278 patients ages 45 to 80 years undergoing an elective colonoscopy for average risk screening purposes. Patients were excluded from the study if they had a prior history of colon resection, colon polyps, history of gastrointestinal diseases such as Ulcerative Colitis (UC) and Crohn’s Disease (CD), or prior positive Cologuard testing. We compared the ADRs of two endoscopists, pre- and post-implementation of an AI system. Patient recall times were also recorded as observational data.

Results: A total of 1,278 patients met the inclusion criteria. Comparison of ADRs from endoscopist 1 shows a pre-AI ADR of 37% compared to a post-AI ADR of 35.12%. Comparison of ADRs from endoscopist 2 shows a pre-AI ADR of 60.59% compared to a post-AI ADR of 56.82%. Crude analysis showed no significant difference in pre-AI and post AI ADRs (48.81% vs 45.97%, p of 0.1006). Additionally, recall time to colonoscopy were recorded during chart review and are available as observational data.

Discussion: Based on our findings, there appears to be no statistically significant difference in ADR with the addition of an AI systems during a screening colonoscopy in an outpatient community setting. Previous studies have shown the impact of AI is most pronounced for polyps ≤5mm, with the biggest difference in detection in low ADR detectors. This technology may not yet have significant benefit in high ADR detectors. As advancements in AI are made, additional studies will be needed to reassess its impact on the screening colonoscopy.

Recall Time (Endoscopist 1)			Recall Time (Endoscopist 2)
	Frequency	Percent		Frequency	Percent
Pre-System			Pre-System
None	7	1.99	None	2	0.59
Less than 1 year	6	1.70	Less than 1 year	29	8.50
2 years	0	0	2 years	2	0.59
3 years	72	20.45	3 years	78	22.87
5 years	104	29.55	5 years	149	43.70
10 years	163	46.31	10 years	81	23.75

Post-System			Post-System
None	8	2.86	None	5	1.62
Less than 1 year	4	1.43	Less than 1 year	16	5.18
2 years	0	0	2 years	2	0.65
3 years	51	18.21	3 years	67	21.68
5 years	99	35.36	5 years	141	45.63
10 years	118	42.14	10 years	78	25.24

Table: Table 1: Observational Data Outlining Recall Time to Endoscopy by Endoscopist

Disclosures:

Kunaal Patel indicated no relevant financial relationships.

Brett VanLeer-Greenberg indicated no relevant financial relationships.

Kunaal Patel, MD, MS, MBA¹, Brett VanLeer-Greenberg, MD². P0357 - Artificial Intelligence in Colonoscopy in the Community Setting, ACG 2023 Annual Scientific Meeting Abstracts. Vancouver, BC, Canada: American College of Gastroenterology.