Controversies in Computer-Assisted Detection in Colonoscopy.

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence in Medicine

Background:

• Computer-aided detection (CADe) in endoscopy shows promise in randomized controlled trials (RCTs) for improving adenoma detection rate (ADR).
• Despite RCT benefits, the adoption of CADe in routine colonoscopy is debated, with conflicting international guideline recommendations.

Purpose of the Study:

• To evaluate the efficacy and effectiveness of CADe in colonoscopy.
• To analyze the impact of CADe on adenoma detection, cost-effectiveness, and potential harms.

Main Methods:

• Review of evidence from randomized controlled trials (RCTs) and real-world implementation studies.
• Analysis of cost-effectiveness studies and consideration of potential harms like deskilling.

Main Results:

• CADe significantly increases ADR in RCTs, mainly detecting diminutive adenomas, with limited impact on advanced adenomas.
• Real-world studies show diminished CADe benefits compared to RCTs, potentially due to factors like Hawthorne effect and polyp management.
• Cost-effectiveness is generally favorable but depends on assumptions regarding ADR improvement and CADe costs.

Conclusions:

• CADe's real-world effectiveness is inconsistent, with gains primarily in diminutive adenomas and uncertain impact on significant clinical outcomes.
• Potential harms include increased polypectomy of non-neoplastic lesions and endoscopist deskilling.
• Further research is needed to bridge the gap between trial efficacy and real-world effectiveness, optimize surveillance, and address human-AI interaction before widespread CADe implementation.