P0922 - Deep Learning and Capsule Endoscopy: Automatic Classification of Small Bowel Cleansing Using a Convolutional Neural Network

Sunday, October 24, 2021

3:30 PM – 7:00 PM PT

Location: Shoreline Exhibit Hall

Has Audio

Presenting Author(s)

Miguel Mascarenhas, MD

Centro Hospitalar de São João
Porto, Porto, Portugal

Miguel Mascarenhas, MD¹, Joao Afonso, MD¹, Tiago Ribeiro, MD¹, João Ferreira, PhD², Hélder Cardoso, MD¹, Patrícia Andrade, MD¹, Renato Natal, PhD², Guilherme Macedo, MD, PhD, FACG¹
¹Centro Hospitalar de São João, Porto, Porto, Portugal; ²University of Porto, Porto, Porto, Portugal

Introduction: Capsule endoscopy (CE) allows minimally invasive inspection of the small bowel and is the first line for investigation of small bowel diseases. An adequate bowel preparation is crucial for a conclusive CE exam. The quality of bowel preparation is mandatory in CE reports and a wide range of qualitative and semi-quantitative scales are available. The reliability of these scales is a limitation as significant intra and interobserver variability has been reported. The application of artificial intelligence (AI) algorithms to endoscopy, including for assessment of bowel preparation, 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 automatic evaluation of bowel preparation in CE. We aimed to develop a CNN-based model for automatic classification of bowel preparation in CE exams.

Methods: We developed, trained, and validated a CNN based on CE images. Each frame was labelled after a consensus of three experts in CE according to the quality of bowel preparation: excellent – ≥ 90% of visible mucosa; satisfactory – 50 to 90% of visible mucosa; unsatisfactory – < 50% of visible mucosa. A training dataset was used for development of the model (80% of the full image dataset). The performance of the network was evaluated using an independent dataset (20% of the full image dataset). The CNN’s output was compared to the classification provided by the experts (Figure 1a). The performance of the CNN was measured by the area under the curve (AUC), accuracy, sensitivity, specificity, positive and negative predictive values (PPV and NPV, respectively).

Results: A total of 5070 CE images were included: 1570 labelled as excellent (E) bowel preparation; 2150 with satisfactory (S) bowel preparation and 1350 with unsatisfactory (U) bowel preparation. The model had an overall accuracy of 94.3% (Figure 1c), 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 (Figure 1c).

Discussion: We developed a CNN-based model for automatic classification of bowel preparation in CE exams base on a simple quantitative scale. The development of systems for automatic assessment of bowel preparation in CE may improve the reliability and reproducibility of bowel preparation scales.

Figure: Figure 1: 1a - Output obtained from the application of the CNN. A blue bar represents a correct prediction. 1b - Evolution of the accuracy oin predicting bowel preparation classification of the CNN 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. CNN: convolutional neural network; E: excellent bowel preparation (≥90% visualized mucosa); S: satisfactory bowel preparation (50 – 90% visualized mucosa); U: unsatisfactory bowel preparation (<50% visualized mucosa).

Table 1 – Confusion matrix of the validation dataset of automatic prediction of bowel preparation by the CNN versus expert classification.

Abbreviations: CNN – Convolutional neural network; Excellent bowel preparation (≥90% visualized mucosa); Satisfactory bowel preparation (50 – 90% visualized mucosa); Unsatisfactory bowel preparation (< 50% visualized mucosa).

		Expert classification
		Excellent	Satisfactory	Unsatisfactory
CNN classification	Excellent	311	9	2
	Satisfactory	3	397	61
	Unsatisfactory	0	24	207

Disclosures:

Miguel Mascarenhas indicated no relevant financial relationships.

Joao Afonso indicated no relevant financial relationships.

Tiago Ribeiro indicated no relevant financial relationships.

João Ferreira indicated no relevant financial relationships.

Hélder Cardoso indicated no relevant financial relationships.

Patrícia Andrade indicated no relevant financial relationships.

Renato Natal indicated no relevant financial relationships.

Guilherme Macedo indicated no relevant financial relationships.

Miguel Mascarenhas, MD¹, Joao Afonso, MD¹, Tiago Ribeiro, MD¹, João Ferreira, PhD², Hélder Cardoso, MD¹, Patrícia Andrade, MD¹, Renato Natal, PhD², Guilherme Macedo, MD, PhD, FACG¹. P0922 - Deep Learning and Capsule Endoscopy: Automatic Classification of Small Bowel Cleansing Using a Convolutional Neural Network, ACG 2021 Annual Scientific Meeting Abstracts. Las Vegas, Nevada: American College of Gastroenterology.