Endoscopy 2024; 56(S 02): S146
DOI: 10.1055/s-0044-1783007
Abstracts | ESGE Days 2024
Oral presentation
Artifical intelligence in colonosopy: Human versus machine! 27/04/2024, 13:30 – 14:30 Room 10

Autonomous Artificial Intelligence versus AI Assisted Human optical diagnosis of colorectal polyps: A randomized controlled trial

R. Djinbachian
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
C. Haumesser
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
M. Taghiakbari
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
H. Pohl
2   VA Medical Center, White River Junction, VT USA, United States of America
,
A. Barkun
3   McGill University, Montréal, Canada
,
S. Sidani
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
J. Liu Chen Kiow
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
B. Panzini
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
S. Bouchard
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
E. Deslandres
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
,
D. Von Renteln
1   CHUM – Centre hospitalier de l'Université de Montréal, Montréal, Canada
› Author Affiliations
 
 

    Aims Artificial intelligence-based optical diagnosis systems (CADx) have been developed to allow pathology prediction of colorectal polyps during colonoscopies. However, CADx systems have not yet been validated for autonomous performance. Therefore, we conducted a trial comparing Autonomous AI (AI-A) to AI assisted human (AI-H) optical diagnosis.

    Methods Artificial intelligence-based optical diagnosis systems (CADx) have been developed to allow pathology prediction of colorectal polyps during colonoscopies. However, CADx systems have not yet been validated for autonomous performance. Therefore, we conducted a trial comparing Autonomous AI (AI-A) to AI assisted human (AI-H) optical diagnosis.

    Results 467 patients were randomized (229 in the AI-A group, 238 in the AI-H). Overall accuracy for optical diagnosis was 77.2% (95%Confidence Interval [CI] 69.7-84.7) in the AI-A group and 72.1% (95% CI 65.5-78.6) in the AI-H group (p=0.86). Sensitivity, specificity, PPV and NPV for adenoma diagnosis were 84.8%, 64.4%, 85.6%, and 63.0% respectively in the AI-A group vs 83.6%, 63.8%, 78.6%, and 71.0% in the AI-H group. Diagnostic performance did not differ significantly between the two groups. AI-A had statistically significant higher agreement with pathology-based surveillance intervals compared to AI-H (91.5% [95% CI 86.9-96.1] vs 82.1% [95% CI 76.5-87.7]; p=0.016).

    Conclusions Autonomous AI-based optical diagnosis had non-inferior accuracy to endoscopist-based diagnosis but achieved higher agreement with pathology-based surveillance intervals. Resect-and-discard and diagnose-and-leave strategies can therefore be considered with current CADx versions without requiring human input.


    #

    Conflicts of interest

    Daniel von Renteln has received research funding from ERBE Elektromedizin GmbH, Ventage, Pendopharm, Fujifilm and Pentax, and has received consultant or speaker fees from Boston Scientific Inc., ERBE Elektromedizin GmbH, and Pendopharm. The remaining authors declare that they have no conflict of interest.

    Publication History

    Article published online:
    15 April 2024

    © 2024. European Society of Gastrointestinal Endoscopy. All rights reserved.

    Georg Thieme Verlag KG
    Rüdigerstraße 14, 70469 Stuttgart, Germany