Evaluation of Streamline Clustering Techniques for Blood Flow Data

Understanding the hemodynamics of blood flow in vascular pathologies such as aneurysms is essential for both their diagnosis and treatment. Computational fluid dynamics (CFD) simulations of blood flow based on patient-individual data are performed to better understand aneurysm initiation and progression and for predicting treatment success. A CFD simulation results in a complex, multiparameter dataset comprising scalar as well as vectorial data attributes. For its comprehensive investigation, the contained flow information is often visualized by a highly dense and cluttered set of integral curves colored according to one of the attributes. We aim at a fully automatic approach for reducing visual clutter and exposing characteristic flow structures by grouping similar curves and computing group representatives. In this work, we lay the foundations by evaluating different clustering techniques for grouping curves. We evaluate Spectral Clustering and four versions of Agglomerative Hierarchical Clustering. Both are particularly suited since they can be based on inter-curve distances rendering the construction of feature vectors unnecessary. Our work focuses on steadystate simulations of blood flow in intracranial aneurysms and the visualization by means of streamlines. Our results indicate that Spectral Clustering as well as Agglomerative Hierarchical Clustering with average link or Ward’s method as proximity measure generate meaningful groups of similar streamlines.