Comparison between manual and automatic assessment of carotid artery stenosis on CT Angiography

ABSTRACT Purpose: The purpose of this study was three-fold. First, to compare an automated method through software analysis with a well described manual method in the determination of carotid artery stenosis. Second, to evaluate if calcium volume scores of the carotid artery bifurcation compare to manually measured carotid artery stenosis. Finally, to evaluate the feasibility of both these methods in possible future CTA carotid stenosis measurements, taking correlations, time evaluation and overall observer preference into account. Methods: Two neuroradiologists separately evaluated a total of 182 carotid arteries twice, using both the automated and the manual method, blinded for previous collected, each others and other patient data. NASCET ratios were compared. A single run of carotid calcium volume score measurements was performed and compared to the manual measurements, after excluding near-occlusions according to a previous developed CTA standard. Intra- and interobserver correlations were evaluated for both the automated and manual method. Time consumption was evaluated, as well as overall observer preference. Results: Both intra- (r=0,961 & r=0,946) and interobserver (r=0,977) correlations were excellent for the automated measurements. Comparing these measurements to the manual method showed a high correlation (r=0,879). Calcium volume scores were not related to percent stenosis (r=0,141). Conclusion: Using an automated software package for carotid stenosis evaluation shows high levels of reproducibility, does not discriminate between different observers and correlates excellent to manual measurements. Using calcium volume scores as a quantification tool for carotid stenosis is unreliable.