Small Bowel Capsule Endoscopy and artificial intelligence: First or second reader?

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Traditional Small Bowel Capsule Endoscopy (SBCE) reading faces challenges with accuracy and time.
• Early machine learning algorithms showed limited performance in SBCE analysis.
• Deep learning advancements offer potential for improved AI applications in SBCE.

Purpose of the Study:

• To evaluate the potential of Artificial Intelligence (AI) in enhancing Small Bowel Capsule Endoscopy (SBCE) feasibility and diagnostic accuracy.
• To explore the role of AI in the automatic detection and characterization of SBCE lesions like angioectasias, erosions, and ulcers.
• To investigate the optimal use of AI in SBCE, as either a first or second reader, considering accuracy and cost-effectiveness.

Main Methods:

• Development and application of advanced machine learning algorithms, particularly deep learning, for SBCE image analysis.
• Focus on automatic detection and characterization of gastrointestinal lesions.
• Comparative analysis of AI as a first reader versus a second reader in SBCE review.

Main Results:

• Deep learning systems show promise in improving the performance and accuracy of SBCE analysis.
• AI has the potential to significantly reduce SBCE reading time and enhance reader attention.
• Current evaluations primarily position AI as a first reader, but its role as a second reader requires further investigation.

Conclusions:

• Deep learning-based AI holds significant potential to revolutionize Small Bowel Capsule Endoscopy reading.
• AI can aid in the efficient and accurate identification of critical lesions, improving diagnostic outcomes.
• The debate on AI's role as a first or second reader necessitates further studies on accuracy and cost-effectiveness, especially for training and complex lesion characterization.