Accurate vessel segmentation with optimal combination of features

We describe a novel appearance model with optimal combined features to produce the accurate vessel segmentation. It starts with investigating a set of multi-scale vessel features, followed by a weighed approach to optimally combine different features. Then the optimally combined features advantage the appearance model to reveal more detailed information of vessel. The novelty of the work lies in the integration of optimal combined multi-scale features in the appearance model. The main advantage of our framework is that it detects vessel boundary in problematic regions that contain small vessels and noise. It is particularly suitable for accurate segmentation of thin and low contrast vessels. Two state-of-the-art vessel segmentation methods were used to compare with our method. Quantitative results on synthetic data indicate that our method is more accurate than these methods. Furthermore, our method performs good in clinical experiments, it is capable of detecting more detailed information of vessel. Compare with two state-of-the-art methods, our method is more accurate and robust, and more suited for automatic vessel extraction.