Centro Hospitalar de São João Porto, Porto, Portugal
Miguel Mascarenhas, MD1, Tiago Ribeiro, MD1, Joao Afonso, MD1, João Ferreira, PhD2, Pedro Marilio Cardoso, MD1, Hélder Cardoso, MD1, Patrícia Andrade, MD1, Marco Parente, PhD2, Renato Natal, PhD2, Guilherme Macedo, MD, PhD, FACG1 1Centro Hospitalar de São João, Porto, Porto, Portugal; 2University of Porto, Porto, Porto, Portugal
Introduction: Device-assisted enteroscopy, and particularly double-balloon enteroscopy (DBE), allow exploration of a deeper small bowel, and have the advantage of allowing tissue sampling and endoscopic therapy. Suspected mid-gastrointestinal bleeding (particularly after positive findings in capsule endoscopy) is the most frequent indication for DBE, and vascular lesions are the most commonly found lesion in these patients. Nevertheless, the lesion detection rate in this setting remains suboptimal (68%).
The application of artificial intelligence algorithms to different endoscopic techniques has provided promising results. Convolutional Neural Networks (CNN) are a multi-layer artificial intelligence architecture with high performance levels for image analysis. The application of these automated algorithms for detection of lesions in DBE has not been explored. We aimed to develop and test a CNN-based algorithm for automatic detection of vascular lesions in DBE exams.
Methods: A convolutional neural network was developed based on 72 DBE exams. A total of 6740 images were included, 1395 images containing vascular lesions. The remaining images showed normal mucosa or other findings. A training dataset comprising 80% of the total pool of images (n=5392) was used for development of the network. The performance of the CNN was evaluated using an independent validation dataset (20% of total image pool, n=1348). The output provided by the network was compared to a consensus classification provided by two gastroenterologists with experience in DBE (Figure 1a). The sensitivity, specificity, accuracy, positive and negative predictive values, and area under the curve (AUC) were calculated.
Results: After optimizing the architecture of the network, our model automatically detected small bowel protruding lesions with an accuracy of 95.3% (Figure 1b). 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 (Figure 1c). The CNN analyzed the validation dataset in 6 seconds, at a rate of approximately 237 frames per second.
Discussion: The authors developed, for the first time, an AI algorithm for automatic detection of vascular lesions during DBE exams. The development of these tools may enhance the diagnostic yield of deep enteroscopy techniques in patients with bleeding originating from the small bowel, which may have significant impact in the management of these patients.
Figure: Figure 1: 1a - Output obtained from the application of the convolutional neural network. A blue bar represents a correct prediction. 1b - Evolution of the accuracy of the convolutional neural network during training and validation phases, as the training and validation datasets were repeatedly inputted in the neural network. 1c - ROC analyses of the network’s performance. N: normal mucosa; PV – vascular lesions.
Disclosures:
Miguel Mascarenhas indicated no relevant financial relationships.
Tiago Ribeiro indicated no relevant financial relationships.
Joao Afonso indicated no relevant financial relationships.
João Ferreira indicated no relevant financial relationships.
Pedro Marilio Cardoso indicated no relevant financial relationships.
Hélder Cardoso indicated no relevant financial relationships.
Patrícia Andrade indicated no relevant financial relationships.
Marco Parente indicated no relevant financial relationships.
Renato Natal indicated no relevant financial relationships.
Guilherme Macedo indicated no relevant financial relationships.
Miguel Mascarenhas, MD1, Tiago Ribeiro, MD1, Joao Afonso, MD1, João Ferreira, PhD2, Pedro Marilio Cardoso, MD1, Hélder Cardoso, MD1, Patrícia Andrade, MD1, Marco Parente, PhD2, Renato Natal, PhD2, Guilherme Macedo, MD, PhD, FACG1. P0923 - Development of a Convolutional Neural Network for the Detection of Vascular Lesions in Double-Balloon Enteroscopy: A Proof-of-Concept Study, ACG 2021 Annual Scientific Meeting Abstracts. Las Vegas, Nevada: American College of Gastroenterology.
