Statistical considerations on the precision of assessing blood vessel diameter in cine coronary angiography.

Problems related to the precision of computerized measurement and analysis of blood vessel diameter from digitized cineangiographic vessel images are described, and shortcomings of the conventional definition of percent stenosis are discussed. Formulae for the precision of percent stenosis, for the minimum significant measurable changes in blood vessel diameter, for the necessary scanning resolution, and for absolute vessel diameter measurement are provided. A Monte Carlo procedure-based technique for assessing the effect of radiographic noise on the location of vessel boundaries within the transverse densitometric image profile is presented and applied to contrast-filled cylindrical phantoms. The results demonstrate the feasibility for achieving subpixel resolution in computerized detection of vessel edges and diameter measurement.