ARTIFICIAL INTELLIGENCE AND CAPSULE ENDOSCOPY: AUTOMATIC CLASSIFICATION OF SMALL BOWEL PREPARATION USING A CONVOLUTIONAL NEURAL NETWORK

Aims Capsule endoscopy (CE) allows non-invasive inspection of the small bowel. An adequate bowel preparation is crucial for a conclusive exam. The application of artificial intelligence (AI) algorithms to endoscopy has produced promising results. Convolutional neural networks (CNNs) are a highly efficient architecture designed for image analysis. To date, no AI-based model has been developed for evaluation of bowel preparation in CE. We aimed to develop a deep learning model for automatic classification of bowel preparation in CE.

Methods We developed a CNN-based on CE images. Each frame was labelled according to the quality of bowel preparation (excellent (E):≥90% of visible mucosa; satisfactory (S): 50-90% of visible mucosa; unsatisfactory (U):<50% of visible mucosa. A training dataset and a validation dataset, comprising 80% and 20% of the total pool of images, respectively, were constructed. The CNN’s output was compared to the classification provided by the experts. The performance of the CNN was evaluated.

Results A total of 5070 images were included: 1570 labelled as E, 2150 as S and 1350 as U. The model had an overall accuracy of 94.3%, a sensitivity of 93.6%, a specificity of 93.1%, a PPV of 92.6% and NPV of 95.7% for differentiation of classes of bowel preparation ([Table 1]). The AUC for E, S and U classes was 1.00, 0.96, 0.97, respectively.

Table 1

	Expert classification
	Excellent	Satisfactory	Unsatisfactory
Convolutional neural network classification	Excellent	311	9	2
	Satisfactory	3	397	61
	Unsatisfactory	0	24	207

Conclusions We developed a CNN-based model for automatic classification of bowel preparation in CE. The development of these automated systems may improve the reliability and reproducibility of bowel preparation scales in CE.

Publication History

Article published online:
14 April 2022