P3832 - Head-to-Head Comparison of Two Computer Aided Detection (CAD-e) Systems on Colonoscopy Performance Metrics

John Cooper, MD¹, Dalton A.. Norwood, MD², Charles D. Evers, MD, MBA¹, Ramzi Mulki, MD³, Sergio Sanchez-Luna, MD³, Fayez Sarkis, MD³, Amanda Cartee, MD³, Patricia Ajayi-Fox, MD³, Anam Hameed, MD³, Kirk B.. Russ, MD⁴, Douglas Morgan, MD³, Shajan Peter, MD^3
1University of Alabama at Birmingham Heersink School of Medicine, Birmingham, AL; ²UAB Minority Health & Health Equity Research Center, The University of Alabama at Birmingham Heersink School of Medicine, Birmingham, AL; ³Basil I. Hirschowitz Endoscopic Center of Excellence, The University of Alabama at Birmingham Heersink School of Medicine, Birmingham, AL; ⁴University of Alabama at Birmingham, Birmingham, AL

Introduction: In recent years, artificial intelligence (AI)-assisted screening colonoscopy has garnered significant interest as it has shown to increase adenoma detection rate (ADR) by reducing human error in polyp detection. A higher ADR is associated with a lower risk of postcolonoscopy colorectal cancer. We aimed to compare two computer aided detection (CADe) systems, SKOUT and GI Genius^TM, to unassisted colonoscopies.

Methods: We conducted a retrospective chart review to assess the performance metrics of first-time screening colonoscopies performed at a large academic medical center between January 2024 and April 2024. We evaluated three cohorts: GI Genius™, SKOUT, and traditional unassisted methods. Key performance metrics analyzed included number of polyps detected and removed, adenoma detection rate (ADR), advanced adenoma detection rate (AADR), sessile serrated polyp (SSP) detection rate and colorectal cancer detection rate (CRC-DR). Data analysis involved chi-square or Fisher’s exact tests for categorical data, two-sided t-tests for continuous parametric data, and Kruskal-Wallis tests for non-parametric data.

Results: 497 unassisted colonoscopies were compared with 174 GI Genius™ and 94 SKOUT-assisted procedures. The study observed statistically significant improvements in polyp removal rates with GI Genius™ compared to unassisted colonoscopies (64.4% vs. 49.7%, p< 0.001). GI Genius™ also showed a higher detection of SSP (< 1 cm) than unassisted methods (10.9% vs. 4.4%, p=0.002), indicating enhanced sensitivity in detecting diminutive polyps. Both CADe systems achieved similar ADR of 40.8% and 42.6%, respectively, which were not statistically different from each other (p=0.78), but both were higher than the unassisted group (34.4%). Furthermore, GI Genius™ demonstrated longer withdrawal times compared to both SKOUT (12.89 vs. 11.06 minutes, p=0.023) and the unassisted group (12.89 vs. 10.43 minutes, p< 0.001), potentially reflecting more thorough inspections.

Discussion: Both CADe systems demonstrated higher polyp removal rates and detection of diminutive polyps when compared to unassisted colonoscopy. Neither CADe system showed superiority in improving adenoma detection rates or other assessed metrics when compared to one another. This highlights that while both systems offer advantages over traditional methods, they may serve complementary roles in clinical settings rather than one being distinctly preferable over the other.


Disclosures:


John Cooper, MD¹, Dalton A.. Norwood, MD², Charles D. Evers, MD, MBA¹, Ramzi Mulki, MD³, Sergio Sanchez-Luna, MD³, Fayez Sarkis, MD³, Amanda Cartee, MD³, Patricia Ajayi-Fox, MD³, Anam Hameed, MD³, Kirk B.. Russ, MD⁴, Douglas Morgan, MD³, Shajan Peter, MD³. P3832 - Head-to-Head Comparison of Two Computer Aided Detection (CAD-e) Systems on Colonoscopy Performance Metrics, ACG 2024 Annual Scientific Meeting Abstracts. Philadelphia, PA: American College of Gastroenterology.