Edge detection in quantitative coronary arteriography (QCA): the impact of image zoom on edge positioning, accuracy, precision and discrimination

Abstract

In geometric QCA, the edge detection (ED) overestimates vessels in the small diameter range (SR, <1.5 mm) due to imaging limitations. Subsequent diameter correction enforces accurate but imprecise corrected diameters. To improve the starting conditions for further analysis the authors used 1-6-fold zoom in phantoms and measured each diameter 240 times by automated ED with optimized weighting of the first and second derivative and subsequent isocenter calibration. The positioning of the second derivative is zoom dependent. In derivative weighting ED, zoom shifts the edges over the entire range toward the vessel center, while in first derivative based ED this occurs only in SR. Zoom improves the accuracy but deteriorates precision of the native measurements. Based on discrimination as the optimality criterion, 3-4-fold zoom is optimal. By applying optimal zoom, precision after diameter correction was doubled.