Use of artificial intelligence in submucosal vessel detection during third-space endoscopy.

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Therapeutic endoscopy, particularly third-space procedures, carries a risk of intraprocedural bleeding.
• The utility of artificial intelligence (AI) in therapeutic endoscopy, unlike diagnostic endoscopy, requires further clarification.
• Blood vessel identification is critical for preventing complications during these interventions.

Purpose of the Study:

• To develop and evaluate an AI algorithm for detecting submucosal blood vessels during third-space endoscopic procedures.
• To assess the impact of AI support on endoscopists' vessel detection rate (VDR) and vessel detection time (VDT).

Main Methods:

• A dataset of 101 standardized video clips with 200 predefined submucosal blood vessels was used.
• Nineteen endoscopists, categorized by experience in endoscopic submucosal dissection, performed vessel detection with and without AI assistance.
• Vessel detection rate (VDR) and time (VDT) were measured for each endoscopist.

Main Results:

• AI support increased the overall VDR from 56.4% to 72.4%.
• AI assistance reduced the VDT from 6.7 seconds to 5.2 seconds.
• False-positive detections occurred in 4.5% of frames and were marked for a significantly shorter duration than true positives.

Conclusions:

• AI algorithms can significantly enhance endoscopists' performance in identifying blood vessels during third-space endoscopy.
• The developed AI tool shows potential for improving both the safety and efficiency of therapeutic endoscopic interventions.
• Further clinical validation is warranted to confirm AI's role in real-world endoscopic procedures.