CT colonography: investigation of the optimum reader paradigm by using computer-aided detection software

Area of Science:

• Medical Imaging
• Radiology
• Computer-Aided Detection

Background:

• Computed tomographic (CT) colonography is a vital tool for colorectal cancer screening.
• Computer-aided detection (CAD) systems aim to improve the accuracy and efficiency of interpreting medical images.
• Evaluating different CAD reading paradigms is crucial for optimizing diagnostic workflows.

Purpose of the Study:

• To compare the diagnostic performance and time efficiency of second read versus concurrent computer-aided detection (CAD) in CT colonography.
• To assess the impact of CAD on polyp detection rates and reading times.
• To establish a reference standard using consensus reading of colonoscopic findings.

Main Methods:

• Prospective comparison of second read CAD and concurrent read CAD paradigms.
• Ten radiologists interpreted 25 CT colonography datasets each, with randomized reading order.
• Data included 69 polyps (28 small, 41 large); analysis used logistic and linear regression, and ROC curves.

Main Results:

• Concurrent CAD reading was significantly faster than second read CAD (2.9 minutes less).
• Second read CAD showed a trend towards higher sensitivity for polyps >= 6 mm (odds ratio 1.5) compared to unassisted reads.
• Concurrent CAD sensitivity was similar to second read CAD for larger polyps but lower for smaller ones (polyps 1-5 mm).
• Mean areas under the ROC curve were 0.83 (unassisted), 0.86 (second read CAD), and 0.88 (concurrent read CAD).

Conclusions:

• Concurrent CAD offers time efficiency in CT colonography interpretation.
• Second read CAD provides the highest sensitivity for polyp detection, particularly for smaller lesions.
• The choice between concurrent and second read CAD depends on the priority between speed and maximal sensitivity.