Evaluating the Effectiveness of AI in Polyp Detection and Characterization
In recent years, the integration of artificial intelligence (AI) into medical diagnostics has transformed various healthcare practices. One of the notable advancements is the application of AI systems in enhancing the detection and characterization of polyps during colonoscopies. This article delves into the clinical effectiveness and diagnostic accuracy of specific AI systems equipped with computer-aided polyp detection (CADe) and computer-aided polyp characterization (CADx) capabilities, based on a comprehensive analysis of studies conducted up until June 2025.
Methods and Analysis
For this evaluation, studies that employed CADe/CADx systems alongside endoscopist assessment during real-time colonoscopies were meticulously reviewed. Key outcomes included the adenoma detection rate (ADR) and diagnostic accuracy metrics such as sensitivity and specificity. Other aspects like additional polyp detection outcomes, procedure duration, and adverse events (AEs) were also scrutinized.
Meta-analyses were stratified for CADe based on individual AI systems. However, due to concerns about heterogeneity, meta-analyses for CADx were not conducted. The studies also explored subgroup analyses based on colonoscopy indication and endoscopist experience to provide a nuanced understanding of the AI systems’ effectiveness.
Results
A total of 52 studies were included, comprising 43 focusing on CADe and 9 on CADx systems. The findings reveal that each CADe system increased the ADR compared to traditional colonoscopy methods without CADe. However, two systems did not achieve statistical significance in their results. There is also evidence suggesting that some systems may improve the detection of advanced adenomas, sessile serrated lesions (SSLs), and other polyp types, although these findings are less certain.
Regarding CADx, the diagnostic accuracy outcomes were inconsistent, often due to limitations like unavailable comparisons with endoscopic optical diagnoses and exclusion of low-confidence diagnoses. Notably, SSLs were frequently categorized as non-neoplastic diagnoses.
The study found that the use of CADe systems may slightly extend the procedure time, but no significant concerns about adverse events were identified. However, there was a lack of data on the long-term impact of CADe/CADx on outcomes such as interval colorectal cancer (CRC) or mortality. Furthermore, no definitive conclusions could be drawn about their impact on different colonoscopy indications or endoscopist experience subgroups.
Conclusion
The research indicates that CADe systems can significantly enhance ADR with improved detection of various polyp types, without raising substantial concerns about procedure time or adverse events. However, the study underscores the necessity for further research into CADx diagnostic accuracy, the long-term impacts of CADe/CADx on CRC incidence, and their effectiveness across different subgroups based on colonoscopy indications and endoscopist experience.
For a deeper dive into this research, please visit the source Here.
“`
