Validation of Artificial Intelligence Computer-Aided Detection of Colonic Neoplasm in Colonoscopy

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Colonoscopic quality is crucial for detecting colon polyps, thereby reducing colorectal cancer risk.
• Artificial intelligence (AI) is increasingly utilized across medical disciplines.
• The ALPHAON® computer-aided detection (CADe) algorithm was previously developed for polyp identification.

Purpose of the Study:

• To validate the ALPHAON® AI CADe algorithm for colon polyp detection.
• To compare the diagnostic performance of ALPHAON® against expert endoscopists.
• To assess AI's impact on endoscopists' polyp detection capabilities.

Main Methods:

• Retrospective analysis of 500 still colon images (100 polyp, 400 healthy).
• Validation of the ALPHAON® CADe algorithm's diagnostic performance.
• Comparison with two expert endoscopists and six trainees, with and without AI assistance after a washout period.

Main Results:

• The ALPHAON® CADe algorithm achieved high detection accuracy (0.97), sensitivity (0.91), specificity (0.99), and AUC (0.967).
• ALPHAON® outperformed expert endoscopists in accuracy, sensitivity, and specificity.
• AI assistance significantly improved polyp detection capabilities for both expert and trainee endoscopists.

Conclusions:

• The ALPHAON® CADe system demonstrates high performance in colon polyp detection during colonoscopy.
• ALPHAON®'s superior sensitivity suggests its potential to outperform expert endoscopists.
• The algorithm effectively enhances polyp detection for both novice and experienced endoscopists, indicating its value as an endoscopic assistant.