Computer-aided operator-interactive densitometry has been developed and applied to determine the percent stenosis for obstructive lesions in the coronary arterial tree. Phantom experiments performed to assess system linearity, accuracy of densitometric measurements, nonuniformity in image amplified response, and the interference of structure noise in measuring precision have been described. Three observers assessed the reproducibility of the method by repeating analyses on 20 stenoses in man recorded on 35-mm cine angiograms. Calculation of the intra- and interobserver errors found that their values significantly decrease with increasing percent stenosis. The reproducibility of the technique was also tested on stenoses that were exposed at multiple angles. The results agreed with those found for the intra- and interoperator errors. Limitations of the densitometric approach and desirability of further automation of the measurements are also discussed.