PDF icon Download PDF
Open Access
CC BY 4.0 · Journal of Coloproctology 2024; 44(S 01): S1-S138
DOI: 10.1055/s-0045-1808860
Câncer do Cólon/Reto/Ânus
Colon/Rectum/Anus Cancer
ID – 141664
Open Topics (oral presentation)

ARTIFICIAL INTELLIGENCE AND HIGH-RESOLUTION ANOSCOPY: DEVELOPMENT OF AN INTEROPERABLE MODEL FOR THE DETECTION AND DIFFERENTIATION OF PRECUSORS OF SQUAMOUS CELL CARCINOMA OF THE ANAL CANAL – A TRANSATLANTIC MULTICENTER STUDY

Thiago da Silveira Manzione
1   Instituto de Infectologia Emilio Ribas, São Paulo, Brasil
,
Miguel Mascarenhas Saraiva
2   Faculdade de Medicina da Uniersidade do Porto, Porto, Portugal
,
Lucas Spindler
3   Hospital Paris Saint-Joseph, Paris, França
,
Luis Barroso
4   Atrium Health Wake Forest Baptist, North Carolina, United States
,
Miguel Martins
5   Hospital Universitário São João, Porto, Portugal
,
Vincent de Paredes
6   Hospital Paris Saint-Joseph, Paris, França
,
João Ferreira
7   Faculdade de Engenharia da Universidade do Porto, Portugal
,
Sidney Roberto Nadal
8   Instito de Infectologia Emilio Ribas, São Paulo, Brasil
› Author Affiliations
› Further Information
Also available ateRef
 
 PDF Download Permissions and Reprints

    Introduction High-resolution anoscopy (HRA) plays a central role in the screening of precursors to squamous cell carcinoma of the anal canal. Artificial intelligence (AI) algorithms can have a positive impact on the detection and differentiation between high-grade (HSIL) and low-grade (LSIL) squamous intraepithelial lesions in HRA images. The objective of this study was to develop a Convolutional Neural Network (CNN) for the automatic detection and differentiation of HSIL and LSIL using HRA images.

    Method A multicenter, prospective study (three distinct groups) was conducted to develop a CNN using 252 filmed HRA exams. Acetowhite lesions were marked and biopsied to define them as either LSIL or HSIL based on histopathology. This study enabled the recording of 65,380 images classified as HSIL or LSIL in an AI database to allow for more accurate diagnosis during HRA. Statistical analysis evaluated sensitivity, specificity, accuracy, positive predictive value (PPV), negative predictive value (NPV), and area under the curve (AUC).

    Results The evaluation of these data revealed sensitivity, specificity, PPV, and NPV values of 96.7%, 91.1%, 92%, and 96.3%, respectively. Accuracy was 94%, and AUC was 0.93. Our results demonstrated that the AI model had high diagnostic performance in HRA. Interoperability and the heterogeneity of the dataset are essential aspects for the implementation of AI models in HRA for clinical practice.

    Conclusion The AI algorithm developed in this study demonstrated its utility for detecting and differentiating precursors to squamous cell carcinoma of the anal canal in images obtained through HRA.


     

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

    Publication History

    Article published online:
    25 April 2025

    © 2025. The Author(s). This is an open access article published by Thieme under the terms of the Creative Commons Attribution 4.0 International License, permitting copying and reproduction so long as the original work is given appropriate credit (https://creativecommons.org/licenses/by/4.0/)

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

      Permissions and Reprints