Effect of a Computer-Aided Device for Detecting Gastric Neoplasms: A Multicenter, Randomized Controlled Trial

Area of Science:

• Gastroenterology
• Medical Artificial Intelligence
• Clinical Trials

Background:

• Multicenter randomized controlled trials on artificial intelligence (AI) in upper endoscopy are limited.
• The efficacy of AI in enhancing gastric neoplasm detection requires further investigation.

Purpose of the Study:

• To evaluate the impact of AI-assisted upper endoscopy on gastric neoplasm detection rates.
• To assess secondary outcomes including early gastric cancer detection and procedural efficiency.

Main Methods:

• A multicenter randomized controlled trial involving 29,514 patients comparing AI-assisted versus non-assisted esophagogastroduodenoscopy.
• Primary outcome: detection rate of gastric neoplasms post-pathologic review. Secondary outcomes: detection rates before review, early gastric cancer ratio, metaplasia/atrophy detection, biopsy rates, blind spots, and procedure time.
• Intention-to-treat (ITT), per-protocol, and subgroup analyses were performed.

Main Results:

• AI did not significantly improve the overall detection rate of gastric neoplasms after pathologic review (1.42% vs 1.25%, P = .25).
• AI improved detection rates based on original pathology (4.06% vs 3.57%, P = .03) and significantly reduced blind spots (2.52 to 1.07, P < .001), while increasing procedure time.
• Subgroup analysis indicated potential benefits for less experienced endoscopists and during fatigue periods. AI demonstrated high diagnostic accuracy for gastric adenocarcinoma and intraepithelial neoplasia.

Conclusions:

• AI did not enhance the overall detection rate of gastric neoplasms in this large multicenter trial.
• AI shows promise in reducing blind spots and may assist less experienced endoscopists, but further real-world studies are needed to confirm its adaptability.
• The study highlights the need for continued research into AI's role in improving endoscopic diagnostics.