Endoscopy 2022; 54(S 01): S183-S184
DOI: 10.1055/s-0042-1745061
Abstracts | ESGE Days 2022
ESGE Days 2022 Digital poster exhibition

PERFORMANCE OF AN ARTIFICIAL INTELLIGENCE ALGORITHM FOR THE DETECTION OF GASTROINTESTINAL ANGIOECTASIA IN DEVICE-ASSISTED ENTEROSCOPY: A PILOT STUDY

T. Ribeiro
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
,
M. Mascarenhas Saraiva
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
,
J. Afonso
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
,
P. Cardoso
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
,
J. Ferreira
2   Faculdade de Engenharia da Universidade do Porto, Department of Mechanical Engineering, Porto, Portugal
,
P. Andrade
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
,
H. Cardoso
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
,
G. Macedo
1   Centro Hospitalar Universitário de São João, Department of Gastroenterology, Porto, Portugal
› Author Affiliations
› Further Information
Also available at
 

Aims Device-assisted enteroscopy (DAE) allows deeper exploration of the small bowel and has the advantage of allowing tissue sampling and endoscopic therapy. Suspected mid-gastrointestinal bleeding (particularly after positive capsule endoscopy) is the most frequent indication for DAE, and angiectasia is the most common lesion. Nevertheless, the detection rate in this setting remains suboptimal (68%).

The application of artificial intelligence (AI) to different endoscopic modalities has produced exciting results. Nevertheless, their application to DAE has not been explored. We aimed to develop and test a convolutional neural network (CNN) algorithm for automatic detection of angioectasia in DAE exams.

Methods A CNN was developed based on 72 DAE exams. A total of 6740 images were included, 1395 images angioectasia, and the remaining showing normal mucosa. A training dataset and a validation dataset, comprising 80% and 20% of the total pool of images, respectively, were constructed. The output provided by the network was compared to a consensus classification by two DAE experts ([Fig.1]). The performance of the CNN was evaluated.

Zoom Image
Fig. 1

Results Our model automatically detected angioectasia with an accuracy of 95.3%. Our CNN had a sensitivity, specificity, positive and negative predictive values of 88.5%, 97.1%, 88.1%, and 97.0%, respectively. The AUC was 0.98. The CNN analyzed the validation dataset at a rate of 237 frames per second.

Conclusions The authors developed a pioneer AI algorithm for automatic detection of GI angiectasia in DAE exams. The potential increase in diagnostic yield provided by these algorithms may lead to more efficient treatment of these patients.



Publication History

Article published online:
14 April 2022

© 2022. European Society of Gastrointestinal Endoscopy. All rights reserved.

Georg Thieme Verlag KG
Rüdigerstraße 14, 70469 Stuttgart, Germany