Edge detection versus densitometry in the quantitative assessment of stenosis phantoms: an in vivo comparison in porcine coronary arteries.

The aim of this study was the in vivo validation and comparison of the geometric and densitometric technique of a computer-assisted automatic quantitative angiographic system (CAAS system). In six Landrace Yorkshire pigs (45 to 55 kg), precision-drilled phantoms with a circular lumen of 0.5, 0.7, 1.0, 1.4, and 1.9 mm were percutaneously introduced into the left anterior descending or left circumflex coronary artery. Twenty-eight coronary angiograms obtained with the phantom in a wedged intracoronary position could be quantitatively analyzed. Minimal lumen diameter, minimal cross-sectional area, percent diameter stenosis, and cross-sectional area stenosis were automatically measured with both the geometric and densitometric technique and were compared with the known phantom dimensions. When minimal lumen diameter was measured using the geometric approach, a nonsignificant underestimation of the phantom size was observed, with a mean difference of -0.06 +/- 0.14 mm. The larger mean difference observed with videodensitometry (-0.11 +/- 0.20 mm) was the result of the failure of the technique to differentiate the low lumen videodensities of two phantoms of smaller size (0.5 and 0.7 mm) from a dense background. Percent cross-sectional area stenosis measured with the two techniques showed a good correlation with the corresponding phantom measurements (mean difference between percent cross-sectional area stenosis calculated from the quantitative angiographic measurements and the corresponding phantom dimensions was equal to 2 +/- 6% for both techniques, correlation coefficient = 0.93 with both techniques, SEE = 5% with the geometric technique and 6% with the densitometric approach).(ABSTRACT TRUNCATED AT 250 WORDS)