ARTIFICIAL INTELLIGENCE AND PILLCAM CROHN’S CAPSULE: AUTOMATIC CLASSIFICATION OF COLON PREPARATION USING A CONVOLUTIONAL NEURAL NETWORK – A MULTICENTRIC PILOT STUDY

Introduction PillCam Crohn's Capsule (PCC) was introduced to provide a panenteric assessment of the gastrointestinal (GI) tract. It enables the performance of a minimally-invasive inspection of the entire GI tract. An adequate bowel preparation is crucial for a conclusive PCC exam. Different scales have been developed for the evaluation of colon preparation in colon capsule videos; however, their application is time-consuming, and they have a high interobserver variability. To date, no model based on artificial intelligence (AI) has been developed for the automatic evaluation of colon preparation in CE.

Objective To develop a model based on a convolutional neural network (CNN) for the automatic classification of colon preparation in PCC exams.

Materials and Methods We developed, trained, and validated a CNN based on PCC images. Each frame was labelled after a consensus of 3 experts according to the quality of bowel preparation – excellent: ≥ 90% of visible mucosa; satisfactory: 50% to 90% of visible mucosa; and unsatisfactory: < 50% of visible mucosa. A training dataset was used for the development of the model, and the performance of the network was evaluated using an independent dataset. The output of the CNN was compared to the classification provided by the experts, and its performance was measured in terms of the area under the curve (AUC), accuracy, sensitivity, specificity, and positive and negative predictive values (PPV and NPV respectively).

Results A total of 59 PCC exams from 2 centres (ManopH Gastroenterology Clinic and Centro Hospitalar Universitário de São João) were used, with a total of 5,774 frames used to develop the CNN. The model had an overall accuracy of 95.8%, a sensitivity of 92.7%, and a specificity of 97.3% for the differentiation of classes of bowel preparation.

Conclusion We developed a CNN-based model that was able to automatically classify the quality of colon preparation in PCC exams based on a simple quantitative scale. The implementation of these systems for the automatic assessment of bowel preparation may improve the reliability and reproducibility of bowel preparation scales.

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

Publication History

Article published online:
16 March 2023