Comparison of automated edge detection and videodensitometric quantitative coronary arteriography

To increase the precision of the evaluation of coronary artery stenosis severity, we have implemented a digital radiographically based computer program which measures absolute and relative coronary artery dimensions using automated edge detection and relative coronary cross-sectional area using videodensitometry. We have compared automated edge detection and densitometric data for, (1) dimensional accuracy using in-vitro arterial phantoms, (2) independence from arteriographic projection in patients undergoing coronary angioplasty and, (3) correlation with coronary flow reserve and stress scintigraphic data in patients undergoing diagnostic catheterization. Both the diameter measurements assessed by automated edge detection and the videodensitometric cross-sectional area measurements were found to correlate closely (r > 0.98) with known dimensions of an arterial phantom containing contrast-filled cylindrical chambers ranging from 0.5–5.0 mm diameter. Although both methods were statistically equivalently accurate, automated edge detection tended to overestimate diameters less than 1.0 mm. Dependence of stenosis severity estimation on the arteriographic projection was evaluated in 13 patients undergoing coronary angioplasty. Comparison of arterial diameter and videodensitometric cross-sectional area estimations in the RAO- and LAO-views demonstrated greater independence from projection for the absolute diameter data. We have also compared quantitative arteriographic data with stress thallium scintigraphy, radionuclide ventriculography and digitally assessed coronary flow reserve information in 19 patients without prior myocardial infarction or coronary collaterals. Of all quantitative parameters of coronary stenosis severity, percent diameter stenosis best predicted functional impairment. Videodensitometric cross-sectional area predicted functional impairment least well. These data suggest that although in-vitro estimations of automated edge detection and videodensitometric methodologies are both highly accurate, absolute and relative diameter assessments are to be preferred over videodensitometric data, because of greater independence from arteriographic projection and closer correspondence to coronary flow reserve and exercise physiological data.