The authors used radiopaque vascular casts to test a new computer-based analysis method that employs a densitometric profile to determine arterial diameters. Results from this method were compared with those obtained with caliper measurements of overhead projection images of the radiograph and direct caliper measurements of the vascular cast. In addition, the corresponding in vivo angiograms were compared with the measurements obtained from the vascular casts. Vessel diameters obtained from cast radiographs by caliper measurement, operator-determined and automated densitometric curve fitting showed excellent correlations with acceptable intraobserver and interobserver variation. All methods, however, underestimated arterial diameters by about 7% compared with diameters obtained from casts by direct caliper measurements. In vivo angiograms revealed smaller arterial diameters than those obtained from specimens. However, we observed higher correlations for measurements at stenotic sites than in non-stenotic regions. Our results suggest definite limitations for the often used direct comparison between in vivo angiographic and postmortem findings.