부산대학교 도서관

Tracking of coronary vessel centerlines in angiograms is essential for accurate, automated quantitative vascular analysis of the coronary vessel tree. Such analyses are important for diagnosis of atherosclerosis, for surgical or treatment planning, for monitoring disease progress or remission, and for comparing efficacies of treatment. The authors have developed a tracking method that automatically identifies the centerline of each vessel after user indication of 2-3 points. The tracking is based on a modified sector search approach. The perimeters of the sectors centered on previously determined tracking points are searched for the pixel with maximum contrast. In angiographic images, the distinction between the vessel and the background can be difficult in regions with poor signal-to-noise ratios (SNR), especially for thin vessels. Standard edge enhancement algorithms can lead to artifacts such as edge overshoot and noise magnification. To improve the performance of the tracking technique, a non-linear adaptive filtering technique has been developed with which background structures are effectively removed while preserving the vessel information. Additional background removal is achieved using pixel connectivity and thresholding. The effect of the filtering on the tracking algorithm was investigated using several clinical angiographic images, The tracking technique successfully determined centerlines of 100% of the vessels after filtering, whereas without filtering only 12% of the vessels were successfully tracked.