In vitro studies have demonstrated that densitometric quantification of coronary artery stenoses is superior to geometric methods to assess non-circular lumens. However, in patients, several authors have reported significant discrepancies between area reduction percentages obtained densitometrically from two different imaging projections. Some of the factors causing the discrepancies can be reduced by simple precautions taken during image acquisition. Some others may be compensated for during analysis. Nevertheless, two factors remain problematic. The first is the inadequate spatial orientation of the vessel axes at the stenotic and reference cross sections with respect to the x-rays. The second is the difficulty in identifying the same vessel cross section in both planes at the time of analysis. We have designed a new densitometric technique that eliminates the error contributions of these two factors. The technique requires simultaneously acquired biplane coronary angiograms and biplane images of a translucent cube bearing steel markers acquired in exactly the same biplane geometry. Using the two projection matrices calculated from the images of the cube, the centrelines and the edges of the coronary arteries can be reconstructed in space from the biplane angiograms. The angles between the vessel axes and the x-ray beams can be determined and the densitometric cross sections can be corrected accordingly. Moreover, the 3D reconstruction allows the identification of the same cross section in the two planes for the determination of the area reduction percentages. Validation measurements were performed on a Perspex phantom and in patients, before and after angioplasty. In both types of measurement, the interplane discrepancies could be roughly halved. The densitometric technique presented can be incorporated into routine angiography and could become a strong alternative to the geometric approach that is presently dominating this field. Keywords: coronary angiography, coronary stenosis, densitometry, QCA