Diagnostic Accuracy of Wireless Capsule Endoscopy in Polyp Recognition Using Deep Learning: A Meta-Analysis

Area of Science:

• Gastroenterology
• Medical Imaging
• Artificial Intelligence

Background:

• Small sample sizes and varied algorithms in current studies limit the assessment of deep learning's accuracy for polyp detection via WCE.
• Accurate polyp identification is crucial for diagnosing gastrointestinal conditions.

Purpose of the Study:

• To conduct a meta-analysis assessing the accuracy of deep learning algorithms in identifying polyps from wireless capsule endoscopy (WCE) images.
• To synthesize existing evidence on the diagnostic performance of AI-powered WCE for polyp detection.

Main Methods:

• A systematic search of PubMed, Embase, Web of Science, and Cochrane Library was performed for studies published up to December 8, 2021.
• Eight studies, analyzing 18,414 frames from 819 patients, were included in the meta-analysis using a random effects model.
• Statistical analysis included subgroup and regression analyses to explore heterogeneity.

Main Results:

• Deep learning models demonstrated high accuracy in polyp detection using WCE.
• Summary estimates included a sensitivity of 0.97 (95% CI, 0.95-0.98) and specificity of 0.97 (95% CI, 0.94-0.98).
• The area under the summary receiver operating characteristic (sROC) curve was 0.99, indicating excellent diagnostic performance.

Conclusions:

• Deep learning significantly enhances the accuracy of polyp identification in WCE.
• Further multicentre prospective randomized controlled studies are warranted to validate these findings.
• AI-assisted WCE shows promise for improved gastrointestinal diagnostics.