Endoscopy 2023; 55(S 02): S10
DOI: 10.1055/s-0043-1765012
Abstracts | ESGE Days 2023
Oral presentation
Artificial Intelligence for polyp detection 20/04/2023, 08:30 – 09:30 Liffey Meeting Room 3

Clinical validation of a computer-aided detection model for colorectal polyp detection (CAD-ARTIPOD) trial using a second observer and real-time unblinding

P. Sinonquel
1   UZ Leuven Gasthuisberg Campus, Leuven, Belgium
2   Katholieke Universiteit Leuven, Leuven, Belgium
,
T. Eelbode
2   Katholieke Universiteit Leuven, Leuven, Belgium
,
O. Pech
3   Krankenhaus Barmherzige Brüder Regensburg, Regensburg, Germany
,
D. De Wulf
4   AZ Delta campus Rumbeke, Roeselare, Belgium
,
P. Dewint
5   Maria Middelares, Gent, Belgium
,
H. Neumann
6   Dr. med. Helmut Neumann, Bad Salzuflen, Germany
,
G. Antonelli
7   Gastroenterology and Digestive Endoscopy Unit, Castelli Hospital, Roma, Italy
,
D. J. Tate
8   Ghent University Hospital, Ghent, Belgium
,
A. Lemmers
9   Hospital Erasme, Bruxelles, Belgium
,
N. D. Pilonis
10   Centrum Onkologii – Instytut im. Marii Skłodowskiej – Curie, Warszawa, Poland
,
M. F. Kaminski
10   Centrum Onkologii – Instytut im. Marii Skłodowskiej – Curie, Warszawa, Poland
,
I. Demedts
1   UZ Leuven Gasthuisberg Campus, Leuven, Belgium
2   Katholieke Universiteit Leuven, Leuven, Belgium
,
C. Hassan
11   Humanitas Research Hospital, Cascina Perseghetto, Italy
,
P. Roelandt
1   UZ Leuven Gasthuisberg Campus, Leuven, Belgium
2   Katholieke Universiteit Leuven, Leuven, Belgium
,
F. Maes
2   Katholieke Universiteit Leuven, Leuven, Belgium
,
R. Bisschops
1   UZ Leuven Gasthuisberg Campus, Leuven, Belgium
2   Katholieke Universiteit Leuven, Leuven, Belgium
› Author Affiliations
› Further Information
Also available at
 

Aims This trial aimed to validate a CADe system for colorectal polyp detection versus trained endoscopists in an endoscopy blinded set-up with real-time unblinding.

Methods The CAD-ARTIPOD trial is a prospective multicenter trial conducted in 9 European centers. CADe specifics were published earlier [1]. Sample size analysis for superiority required 2048 polyps assuming a difference in sensitivity of 2% with endoscopic diagnosis as gold standard. All endoscopists (ADR 20-50%) were blinded for the CADe output. A second observer classified in real time the CADe output in (1) true positive (detected/not detected by the endoscopist) (2) false negative or (3) clinically relevant or irrelevant false positive detection. If a polyp was only detected by CADe, the endoscopist was unblinded for confirmation of the finding. Histological assessment of endoscopically confirmed polyps was performed by two independent histopathologists.

Results A number of 2080 polyps were collected, 1549 were histologically confirmed polyps.

For the primary endpoint superiority was not reached between CADe and endoscopists (sensitivity (Sn) of 0.94 vs 0.96) (non-inferiority margin of 5%, p = 0.0216). False positive rate was 1.8/min. With histology as standard, CADe outperformed the endoscopists (Sn 0.96 versus 0.95 (p=0.03)) and showed an extra detection rate of 4.8% (Fig. [1]).

Zoom Image
Fig. 1  

Conclusions This manufacturer independent CADe system used in an innovative and competitive study design demonstrated non-inferiority in comparison to endoscopists’ polyp detection. However, with histology as gold standard CADe outperformed endoscopists. Our study also shows that “watching over a shoulder” in itself leads to higher PDR and ADR



Publication History

Article published online:
14 April 2023

© 2023. European Society of Gastrointestinal Endoscopy. All rights reserved.

Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany

 
  • 1 Sinonquel P, Eelbode T, Hassan C. et al. Real-time unblinding for validation of a new CADe tool for colorectal polyp detection. Gut 2021; 70 (04) 641-643
    CrossrefPubMedSearch in Google Scholar