Open Access
CC BY-NC-ND 4.0 · Journal of Coloproctology 2022; 42(S 01): S1-S219
DOI: 10.1055/s-0043-1764832
Enteroscopia, Colonoscopia e Pólipos
ID – 114659
Pôster Eletrônico

DEEP LEARNING AND CAPSULE ENDOSCOPY: AUTOMATIC IDENTIFICATION AND DIFFERENTIATION OF SMALL BOWEL LESIONS WITH DISTINCT HEMORRHAGIC POTENTIAL USING A CONVOLUTIONAL NEURAL NETWORK

Authors

  • Miguel Mascarenhas

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal

  • João Afonso

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal

  • Tiago Ribeiro

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal

  • Francisco Costa Mendes

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal

  • João Ferreira

    2   Faculdade de Engenharia da Universidade do Porto, Porto, Portugal

  • Patricia Andrade

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal

  • Helder Cardoso

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal

  • Guilherme Macedo

    1   Centro Hospitalar Universitário de São João, Faculdade de Medicina da Universidade do Porto, Porto, Portugal
Further Information
Also available at
 
 

Objective Capsule endoscopy (CE) is pivotal for the evaluation of small bowel disease. Obscure gastrointestinal bleeding most often originates from the small bowel, and CE frequently identifies a wide range of lesions with different bleeding potentials in these patients. However, reading CE examinations is a time-consuming task. Convolutional neural networks (CNNs) are highly efficient artificial intelligence tools for image analysis. The present study aims to develop a CNN-based model for the identification and differentiation of multiple small bowel lesions with distinct hemorrhagic potential using CE images.

Design We have developed, trained, and validated a denary CNN based on CE images. Each frame was labelled according to the type of lesion (lymphangiectasia, xanthomas, ulcers, erosions, vascular lesions, protruding lesions, and blood). The hemorrhagic potential was assessed by the Saurin classification. The entire dataset was divided into training and validation subsets. The performance of the CNN was measured by the area under the receiver operating characteristic curve, sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV).

Results A total of 53,555 CE images were included. The model had an overall accuracy of 99%, a sensitivity of 88%, a specificity of 99%, a PPV of 87%, and an NPV of 99% for the detection of multiple small bowel abnormalities and the respective classification of the bleeding potential.

Conclusion We developed and tested a CNN-based model for the automatic detection of multiple types of small bowel lesions and the classification of the respective bleeding potential. This system may improve the diagnostic yield of CE for these lesions and overall CE efficiency.


No conflict of interest has been declared by the author(s).

Publication History

Article published online:
16 March 2023

© 2023. Sociedade Brasileira de Coloproctologia. This is an open access article published by Thieme under the terms of the Creative Commons Attribution-NonDerivative-NonCommercial License, permitting copying and reproduction so long as the original work is given appropriate credit. Contents may not be used for commecial purposes, or adapted, remixed, transformed or built upon. (https://creativecommons.org/licenses/by-nc-nd/4.0/)

Thieme Revinter Publicações Ltda.
Rua do Matoso 170, Rio de Janeiro, RJ, CEP 20270-135, Brazil