Skip to main content
2025 ASCRS Annual Scientific Meeting Main banner
Final Program


Visit ASCRS U
Icon Legend
This session is not in your schedule.
This session is in your schedule. Click again to remove it.
Presentation Icons
CME
CME
Invitation/Ticketed Event
Invitation/Ticketed Event
Invited Lectureship
Invited Lectureship
Award Recipient
Award Recipient
Self-Assessment
Self-Assessment
Social
Social
Industry
Industry
Back

Session: ePoster: Other Neoplasia and Endoscopy 3 (Sun350)

EP564 - Artificial Intelligence Allied To Colonoscopy To Validate Image Quality Of The Intestinal Lumen

Sunday, May 11, 2025
3:55 PM – 4:00 PM PT
Location: Monitor 4, Sails Pavilion, Upper Level

    Author(s)

  • Lara Burlamaqui Veras (she/her/hers)

    166846
    Regular member
    Federal University of Ceará
    Fortaleza, Ceara, Brazil

    All of the relevant financial relationships listed below have been mitigated.
    This individual has no financial relationships with ineligible companies.

  • MN

    Miguel Ângelo Nobre e Souza, n/a

    Médico
    Universidade Federal do Ceará
    Fortaleza, Ceara, Brazil

    Disclosure information not submitted.

    • ePoster Presenter(s)

  • Carolina Vannucci Vasconcellos

    Médica
    Universidade Federal do Ceará
    Fortaleza, Ceara, Brazil

    All of the relevant financial relationships listed below have been mitigated.
    This individual has no financial relationships with ineligible companies.

    • Author(s)

  • AN

    Augusto Vasconcellos Nogueira Diogenes, n/a

    Acadêmico
    Universidade Federal do Ceará
    Quixadá, Ceara, Brazil

    Disclosure information not submitted.

  • Rd

    Rônei Reis de Castro e Silva, n/a

    Doutorando
    Universidade Federal do Ceará
    Fortaleza, Ceara, Brazil

    Disclosure information not submitted.

  • JL

    José Nilo de Lima Filho, n/a

    Acadêmico de Medicina
    Universidade Federal do Ceará
    Fortaleza, Ceara, Brazil

    Disclosure information not submitted.

All of the relevant financial relationships listed below have been mitigated.
Lara Burlamaqui Veras: This individual has no financial relationships with ineligible companies.
Carolina Vannucci Vasconcellos: This individual has no financial relationships with ineligible companies.
Image or Table (Do Not add text here):

Purpose/Background:

Introduction: Colorectal cancer is the third most prevalent type of cancer in the world. Its early detection, essential for effective treatment, depends on the quality of the images captured during colonoscopy. With images of low visual quality, there is a risk of failure to diagnose injuries. Artificial intelligence (AI) emerges as a promising tool to assist in the selection of ideal images for the diagnostic process.

Objective: This study aims to use a supervised machine learning model, specifically a convolutional neural network based on the AlexNet architecture, to categorize colonoscopy images according to their suitability for detecting polyps or lesions.

Methods/Interventions:

Method: The model was developed and trained using 1003 images previously classified by coloproctology professionals as "adequate" or "inadequate". The images were divided into a training set (801 images) and a validation set (202 images). Training took place in 10 cycles
(epochs), using libraries such as PyTorch, NumPy and Scikit-learn. The accuracy of the model was evaluated to determine its categorization capacity.

Results/Outcomes: After all epochs of training were complete, the model was tasked to classify the images of the validation set. This process returned an accuracy of 71.9%, which means that, in 71.9% of cases, the model’s classification of validation set images matched the classification assigned to them by human professionals.

Conclusion/Discussion: Although the model demonstrated potential, performance was limited, possibly due to the quality of the images, the choice of algorithm, and the size of the dataset. Improving these aspects could significantly increase the accuracy and clinical usefulness of the system.