In a collaboration between the University of Texas (software) and the University of Zürich (hardware) a compact, automatic system for biplane quantitative coronary arteriography was developed. The system is based on a 35 mm film projector, a slow-scan CCD-camera (image digitizing) and a computer workstation (Apollo DN 3000, image storage and processing). A new calibration procedure based on two fixed reference points in the center of the image intensifier was used (isocenter technique). Contour detection of coronary arteries was carried out in biplane projection using a geometric-densitometric edge-detection algorithm. The proximal and distal luminal areas, as well as the minimal luminal area of the stenotic vessel segment were determined. Accuracy and precision were determined from precision drilled holes in a plexiglas cube which were filled with 50%, 75% and 100% contrast medium. The diameter of the holes ranged from 0.5 to 5.0 mm. The mean difference and the standard deviation of the differences between the true and the measured diameters were 0.12 +/- 0.14 mm for plane A and 0.26 +/- 0.17 mm for plane B, respectively. After a second order correction the mean difference amounted to 0.02 +/- 0.09 mm for plane A and 0.02 +/- 0.12 mm for plane B, respectively. Intra- and interobserver variability were evaluated in 5 patients (age 60 +/- 10 years) with coronary artery disease using 16 normal and 5 stenotic vessel segments (cross-sectional area ranging from 0.8 to 8.7 mm2). Two independent observers analyzed the same vessel segment twice. Intraobserver variability expressed as the standard error of estimate in percent of the mean angiographic vessel area (SEE) amounted to 2.1% for observer 1 and 4.4% for observer 2, respectively. Interobserver variability expressed as SEE was 4.1% for measurement 1 and 3.6% for measurement 2, respectively.