CORONARY ARTERY DISEASE Computerized image analysis for quantitative measurement of vessel diameter from cineangiograms

Subjective estimates of the angiographic severity of coronary artery stenoses show variability and inaccuracy. We therefore tested the accuracy of a newly developed computerized image analysis system for quantitating vessel diameter from cineangiograms. Fourteen cylindrical phantoms of known diameter were filled with contrast medium and filmed over a wide range of clinically relevant radiographic conditions in order to develop regression equations that related computer-derived to anatomic diameters. Computer measurements of vessel diameter were unaffected by vessel size, magnification, focal spot size, thickness of scattering medium, kilovolt peak, or location within the radiographic field, but a correction factor was necessary for a small but significant (p < .01) linear dependence on contrast medium concentration. The accuracy of computerized vessel diameter measurements ranged between + 59 and 137 ,u for all conditions except for rapid vessel motion and contrast medium concentrations of 50% or less meglumine diatrizoate (Renografin 76), both of which resulted in reduced accuracy as well as in the inability to locate lumen edges of vessels less than 1 mm in diameter. Circulation 68, No. 2, 453-461, 1983. SUBJECTIVE VISUAL ESTIMATES of percent stenosis of a coronary artery from cineangiograms have been shown to be characterized by a large interobserver variabilityl-3 and a descrepancy between angiographic and postmortem estimates of lesion severity has been noted in a number of studies."7 As a result, attempts have recently been made to quantitate luminal dimensions more precisely. For visual measurements calipers, optical scale devices,>'0 and computer-assisted edge tracking11 have recently been used to attain this goal. Semiautomated'2 and automated"3 edge-tracking techniques applied to digitized cineangiographic images have likewise been recently used in attempts to From the Charles A. Dana Research Institute and the Harvard-Thorndike Laboratory of Beth Israel Hospital, Department of Medicine, Cardiovascular Division, the Radiology Research Laboratory of the Department of Radiology, Beth Israel Hospital, and the Department of Radiology, Brigham and Women's Hospital, Harvard Medical School, Boston. This work was supported in part by a Young Investigatorship Research Award HL27328-02 from the National Heart, Lung and Blood Institute. Address for correspondence: J. Richard Spears, M.D., Cardiovascular Division, Beth Israel Hospital, 330 Brookline Ave., Boston MA 02215. Received March 2, 1982; revision accepted April 7, 1983. Vol. 68, No. 2, August 1983 further objectify measurements of luminal dimensions. Although the reproducibility of various objective methods of diameter measurement has been studied, the accuracy of these measurements has received little attention. The fact that radiographic vessel images have edge gradients rather than sharply demarcated edges is not widely appreciated by angiographers. 14 Accurate diameter measurement is dependent on precise localization of the anatomic vessel edge within the edge gradient, and the magnitude of error in diameter measurements associated with arbitrary assignment of the vessel edge within the edge gradient may be considerable. A new method for computerized edge detection of coronary arteries from digitized 35 mm cineangiographic images with the use of specific algorithms to locate spatially disparate points within the edge gradient has been developed. The hypothesis that each of these points can be mathematically related to the anatomic edge was tested by filming contrast mediumfilled cylindrical phantoms over a wide range of cineradiographic conditions and comparing computer-derived measurements of diameter from digitized frames with the known diameters. 453 by gest on Jauary 5, 2018 http://ciajournals.org/ D ow nladed from