Relative Insensitivity of densitometric stenosis measurement to lumen edge determination

Coronary artery stenosis measurement from coronary arteri is dependent on lumen edge determination. In theory, dependence may be minimized by densitometric integration of the entire lumen as compared with densitometric determination of edges alone coupled with calculation of lumen size by geometric formulas. To assess the impact of approximations of lumen border position when densitometric and geometric methodare employed, external, intermediate and internal borders were defined and used with each method to calculate lumennarrowing from images of contrast-filled, asymmetrically stenotic model vessels and from clinically generated coronary arterio-grams. Actual cross-sectional narrowing of models correlated almost perfectly with densitometric narrowing for each of the three border criteria (external: r = 0.99, n = 192; intermediate: r = 0.99, n = 192; internal: r = 0.99, n =192). Lumen narrowing calculated by assuming a circular lumen geometry with a diameter equal to the distance between these identical borders correlated less well with true am namwing (external: r = 0.79, n = 192; intermediate: r = 0.83, n = 192; Internal: r = 0.70, n = 192); the residual variance of the densitometric regression was significantly less than that of the geometric regression for each o the three border criteria (external: 0.13 versus 372, p < 0.001; intermediate: 7.39 versus 315, < 0.001; internal: 7.13 versus 531, p < 0.001). When severity of each of 21 stenotic coronary artery lesions was assessed from clinical arterlograms, the mean of the three densitometric estimations of stenosis obtained by using these border definitions correlated well with the mean of the three geometric estimations obtained by using the identical borders (r = 0.00, n = 21). However, the variance about the mean densitometric measurement of stenosis was significantly less than that about the mean geometric measurement of stenosis (4.24 versus 91.57, p < 0.005). Although no absolute measurement of actual coronary artery stenosis severity exists for comparison, the lower variability of the densitometric compared with the geometric approach supports the conclusion that densitometry is relatively less sensitive to imprecise border positioning than are geometric methods. Together with the model vessel results, these findings suggest that densitometry provides more precise measurement of stenosis than does geometric analysis when applied to clinical coronary arteriograms.