Tree Structure Extraction in Vascular Images Using Edge Detecting Trace Algorithms

An edge detection-based vessel tracing algorithm is presented to extract vascular tree structures present in medical images. Vasculatures are important features in biomedical image processing applications, such as in analyzing the biological process of angiogenesis. The presented algorithm first identifies points where blood vessels are likely to be then initiates traces from each identified seed point. For the task of vasculature extraction, the tracing algorithm has several advantages over traditional edge detection methods: reduction of the amount of pixel processing, isolation of blood vessels, and ensured connectedness. The algorithm is executed iteratively with decreasing seed point detection and trace termination thresholds to overcome the problem of variant contrast levels. This provides an adaptive approach to threshold selection and results in improved feature extraction.