Polyp characteristics correctly annotated by computer-aided detection software but ignored by reporting radiologists...

Area of Science:

• Radiology
• Medical Imaging
• Gastroenterology

Background:

• Computer-aided detection (CAD) aims to improve radiologist performance in identifying polyps during CT colonography.
• Despite CAD, some polyps are still incorrectly dismissed by radiologists.

Purpose of the Study:

• To characterize polyps that were missed by radiologists despite appropriate CAD prompting during CT colonography.
• To identify factors contributing to the incorrect dismissal of polyps.

Main Methods:

• Retrospective analysis of 111 polyps (≥6 mm) from previous CT colonography studies.
• Two new observers evaluated polyps using predefined criteria, assessing size, morphology, visualization, and CAD interaction.
• Statistical analysis, including logistic regression, compared missed polyps with and without CAD gain.

Main Results:

• Before CAD, larger polyps (≥10 mm), easily visualized polyps, and uncoated polyps were less likely to be missed.
• After CAD, increasing polyp size and irregular morphology decreased the likelihood of CAD gain (correct reporting).

Conclusions:

• Larger, irregular polyps are a significant source of missed diagnoses in CT colonography.
• Radiologist dismissal of polyps persists even with effective CAD prompting, highlighting areas for improvement.