Densitometric arterial lesion size using morphologically filtered images

Densitometric measurement of coronary artery narrowing requires a correction for the nonarterial contributions to the density ("background"). Two correction methods were studied. The first samples background in two strips of pixel parallel to the artery edges, and excludes nearby arteries. The second method measures density on a morphologically filtered image with no additional background correction. Measurements of % narrowing are consistent, reproducible and well correlated between the two methods. The first method corrects completely for background, the second less so. Using a set of routine arteriograms, densitometric % narrowing correlates with the visual estimate. Both methods underestimate lesion severity. Morphological filtering was less variable when comparing values measured on different views of same lesion.<>