Computed tomographic colonography: assessment of radiologist performance with and without computer-aided detection

Area of Science:

• Medical Imaging
• Gastroenterology
• Radiology

Background:

• Computer-aided detection (CAD) for CT colonography shows promise for detecting significant adenomas.
• The interaction between CAD and human readers in CT colonography has not been fully evaluated.

Purpose of the Study:

• To assess the impact of CAD on reader performance in CT colonography.
• To evaluate changes in polyp detection sensitivity, specificity, and interpretation time with CAD assistance.

Main Methods:

• Ten readers interpreted CT colonography images twice, with and without CAD, after a 2-month interval.
• Per-patient and per-polyp detection rates were compared to known patient status.

Main Results:

• CAD significantly increased per-patient sensitivity (70% of readers) and polyp detection (average 9.1% increase).
• CAD improved detection of small and medium polyps; however, overall detection of large polyps remained suboptimal.
• Interpretation time decreased significantly with CAD (1.9-2.9 minutes).

Conclusions:

• CAD for CT colonography enhances polyp detection and efficiency.
• CAD is a valuable tool but does not eliminate the necessity for comprehensive reader training.