Automatic vessel extraction from angiogram images

Vessel extraction from angiogram images is useful for blood vessel measurement and computer visualization of the coronary artery tree. A digitized angiogram can be treated as a height map and ridges in this map correspond to the centerlines of the blood vessels. The authors propose a ridge extraction algorithm based on the calculation of the local maximum ascent direction to extract blood vessel axes automatically from all angiogram image. The output of the ridge detection algorithm is a set of ridge pixels assembled in a natural hierarchy order, aligned with the centerlines of the blood vessel, plus a set of ridge pixels induced by image noise, which require a post processing stage to remove. To extract vessel axes from the angiogram images. First, the digitized angiogram is balanced by a median filter it is then smoothed by a projective non-linear diffusion method. Next, an ROI, which is crossed by the shadow of a vessel, is detected automatically by adaptive thresholding, both to cut the cost of the ridge extraction process and to reduce the number of the spurious ridges induced by image noise. The authors then apply the ridge detection algorithm to the ROI to find candidate vessel centerlines. Finally the detected ridges are piped into a curve relaxation process and the candidate vessel centerlines are assembled together; spurious lines induced by noise are removed.