Related Experiment Video
Updated: Jul 5, 2025
06:55
Diagnosis of Neoplasia in Barrett’s Esophagus using Vital-dye Enhanced Fluorescence Imaging
Published on: May 11, 2014
12.1K
Artificial Intelligence and Panendoscopy-Automatic Detection of Clinically Relevant Lesions in Multibrand
Francisco Mendes1,2, Miguel Mascarenhas1,2,3, Tiago Ribeiro1,2,3
1Alameda Professor Hernâni Monteiro, Department of Gastroenterology, São João University Hospital, 4200-427 Porto, Portugal.
Cancers
|January 11, 2024
Summary
A new artificial intelligence model using convolutional neural networks (CNNs) can detect lesions during device-assisted enteroscopy (DAE) with high accuracy. This AI tool shows promise for improving diagnostic yield in gastrointestinal evaluations.
Area of Science:
- Gastroenterology
- Medical Imaging
- Artificial Intelligence
Background:
- Device-assisted enteroscopy (DAE) enables comprehensive gastrointestinal tract evaluation but often has suboptimal diagnostic yield.
- Convolutional neural networks (CNNs) are advanced AI models for image analysis, yet their application in DAE remains understudied.
Purpose of the Study:
- To develop and evaluate a multidevice CNN for detecting clinically relevant lesions during DAE panendoscopy.
- To assess the performance of the CNN in improving the diagnostic capabilities of DAE.
Main Methods:
- Retrospective analysis of 338 DAE exams across single-balloon, double-balloon, and motorized spiral enteroscopy devices.
- Development of a CNN model trained on 36,599 images and tested on 4,066 images.
- Comparison of CNN output against expert consensus classification, evaluating sensitivity, specificity, PPV, NPV, accuracy, and AUC-PR.
Main Results:
- The developed CNN achieved high performance metrics: 88.9% sensitivity, 98.9% specificity, 95.8% PPV, 97.1% NPV, and 96.8% accuracy.
- The model demonstrated a strong area under the precision-recall curve (AUC-PR) of 0.97.
- This represents the first multidevice CNN developed for lesion detection in DAE panendoscopy.
Conclusions:
- A novel multidevice CNN effectively detects clinically relevant lesions during DAE panendoscopy.
- The development of accurate deep learning models is crucial for enhancing the diagnostic yield of DAE.
- This AI approach holds significant potential for improving patient diagnosis and management in gastroenterology.
