1993 Volume 31 Issue 1 Pages 37-46
The reconstruction of three dimensional structure of cerebral arteries from two dimensional cineangiograms is useful to extract three dimensional information in the angiograms. In this paper, we consider the extraction of vessel center lines from the cineangiograms, which is a basic and important process for the three dimensional reconstruction. As the structure of cerebral arteries is extremely complicated and the vessel regions on the cineangiograms cross and overlap with each other, it is difficult to extract vessel center lines. A framewise extraction method is developed in this study, which repeats the vessel extraction upon each of a series of cineangiograms. On each frame, the system selects a proper image processing algorithm and parameters for the vessels under processing. The extraction method combines both the edge detection and the segmentation to improve the accuracy of vessel extraction from a single frame of cineangiograms. A vessel examining process is also proposed, which evaluates the possibility of extracted center lines being real cerebral arteries. The knowledge about cerebral arteries as well as that of cineangiograms is utilized in the processes for vessel extraction and vessel examination to improve the accuracy of the extraction results.