Carotid artery stenosis: Reproducibility of automated 3D CT angiography analysis method

The aim of this study was to assess the reproducibility and anatomical accuracy of automated 3D CT angiography analysis software in the evaluation of carotid artery stenosis with reference to rotational DSA (rDSA). Seventy-two vessels in 36 patients with symptomatic carotid stenosis were evaluated by 3D CT angiography and conventional DSA (cDSA). Thirty-one patients also underwent rotational 3D DSA (rDSA). Multislice CT was performed, with bolus tracking and slice thickness of 1.5 mm (l-mm collimation, table feed 5 mm/s) and reconstruction interval of 1.0 mm. Two observers independently performed the stenosis measurements on 3D CTA and on MPR rDSA according to the NASCET criteria. The first measurements on CTA utilized an analysis program with automatic stenosis recognition and quantitation. In the subsequent measurements, manual corrections were applied when necessary. Interfering factors for stenosis quantitation, such as calcifications, ulcerations, and adjacent vessels, were registered. Intra- and interobserver correlation for CTA were .89 and .90, respectively (P<.001). The interobserver correlation between two observers for MPR rDSA was .90 (P<.001). The intertechnique correlation between CTA and rDSA was .69 (P<.001), using automated measurements, but increased to .81 (P<.001) with the manually corrected measurements. Automated stenosis recognition achieved a markedly poorer correlation with MPR rDSA in carotids with interfering factors than those in cases where there were no such factors. Automated 3D CT angiography analysis methods are highly reproducible. Manually corrected measurements facilitated the avoidance of the interfering factors, such as ulcerations, calcifications, and adjacent vessels, and thus increased the anatomical accuracy of arterial delineation by automated CT angiography with reference to MPR rDSA.