An Improved Seed Point Detection Algorithm for Centerline Tracing in Coronary Angiograms

We presents a new method to detect initial seed points for automatic tracing of the vessel center lines in coronary angiograms. Vessel tracing algorithms are known to be fast and efficient among several feature extraction methods. However, most of them suffer from incomplete results due to inappropriate trade-off between the completeness of seed point detection and computational efficiency. Imposing strict validation rules decreases the number of background traces, but results in more false negatives and more computation time. We show that using the geometrical properties of gradient vectors calculated at vessel boundary points as a validation criterion, improves the performance of the seed point detection algorithm. The results illustrate that the proposed method improves upon the prior method in both performance and computation time.continue-->