Many studies have reported the accuracy of quantitative coronary angio
graphy (QCA) based on experiments using moderated-size phantoms imaged
under unrealistic radiographic conditions. However, these observation
s may not be-generalizable to the setting of clinical angiography. To
determine QCA accuracy in a realistic radiographic setting and evaluat
e the impact of the x-ray system line spread function, plexiglass phan
toms were imaged inside and out of a human thorax. A realistic radiogr
aphic background was associated with a 38% increase in variability of
results (P < 0.05). Low concentrations of contrast and large image int
ensifier input screens were associated with significantly larger error
s and variability in results (p < 0.05). There was a systematic overes
timation of diameter in the smallest phantom. A mathematical model of
the x-ray line spread function was developed that explains the observe
d overestimation of the smallest phantom and provide a rational approa
ch for correction of the line spread function for QCA. Many factors en
countered in clinical coronary angiography such as nonuniform radiogra
phic background, low concentrations of contrast, and small vessel diam
eters have a significant adverse impact on the accuracy and/or variabi
lity of gradient-based edge detection QCA systems. (C) 1993 Wiley-Liss
, Inc.