Real-time illustrative visualization of cardiovascular hemodynamics

Healthcare institutions generate vast amounts of clinical imaging data. Because of the advances in acquisition techniques, contemporary imaging data can be multimodal, multi-dimensional and multi-valued by nature. In particular, modern magnetic resonance imaging (MRI) techniques enable acquisition of multiple image series that supply anatomical and functional information. In this thesis, we concentrate on visual analysis of MRI-acquired blood-flow information in the heart and the thoracic arteries. In addition to anatomical information, MRI enables non-invasive acquisition of time-resolved blood-flow velocity data that capture the intricate cardiovascular hemodynamics. These quantitative velocity data describe the blood flow by means of volumetric velocity fields during a heart beat. This is often referred to as four-dimensional blood-flow data, based on the three spatial dimensions plus the time. For this relatively new MRI acquisition technique, physicians are rather unsure what to expect from the data. Nevertheless, there are clear indications that the data contain valuable information. Quantitative and qualitative analyses of these data should provide insight into the blood-flow dynamics, improving the understanding of the cardiovascular system and its pathologies. This improved understanding conceivably leads to better diagnosis and prognosis of cardiovascular diseases, and may facilitate risk assessment, as well as evaluation of treatment and follow-up studies. With qualitative analyses, physicians aim for newfound insight into the intricate blood-flow dynamics, and therefore there are no a priori questions to be answered, or tasks to be performed. The visual analysis should enable exploration of the complex high-dimensional data. However, exploration through the typical series-by-series and slice-by-slice inspection requires a full mental reconstruction of the unsteady bloodflow velocity data, as well as the cardiovascular morphology. This is a tedious and highly challenging task, even for skilled physicians. Therefore, we aim to alleviate this task by means of comprehensive exploratory and interactive visualization techniques. These techniques incorporate domain knowledge, and provide a more abstract representation of the data that can be steered interactively by the physicians. Prior to visual analysis, sensible abstraction of the high-dimensional data is generally required. We have investigated various approaches to simplify the abundance of information contained in the acquired blood-flow data. On the one hand, we present a segmentation of the luminal geometry, using both direction and speed of the bloodflow velocities. We show that the inclusion of directional information leads to more accurate segmentation results. On the other hand, we abstract the time-resolved blood-flow data using spatiotemporal hierarchical clustering. The resulting cluster tree allows for intuitive level-of-detail selection, using a single user-defined parameter. For sparser detail levels, we use the cluster results in various visualization techniques, providing an abstract overview of the blood-flow data. To facilitate interactive exploration of the four-dimensional blood-flow data, we introduce different probing tools, enabling local analysis of the hemodynamics. The probes enclose a region-of-interest, and serve a basis for various visualizations. The first probing approach focusses on the thoracic arteries, using an automated technique to select vessel cross-sections, perpendicular to the centerline of the vessel. With these cross-sections as a basis, we introduce novel geometry-based blood-flow visualization approaches, such as exploded planar reformats, and flow-rate arrow-trails. In addition, we present improvements on established flow visualization techniques, such as dynamic pathline seeding, and animated pathline highlights. All blood-flow visualizations are combined with an illustrative context to communicate the anatomy. The second probing technique enables exploration throughout the cardiovascular system. To this end, we introduce a virtual probe that resides in the blood-flow field. The virtual probe can be translocated by means of elementary two-dimensional interactions, enabling exploration. Based on the location of the virtual probe, we introduce novel visualization techniques, such as comic-inspired particles, illustrative pathlines, and nested pathtubes. Furthermore, we have investigated approaches to communicate the anatomical context, using volume projections and volume clipping. The results of both probing approaches were evaluated with domain experts, measuring the value of the visualizations, the interaction approaches, and the involved user parameters. The evaluation questionnaires were carried out with several physicians, who are actively involved with advancements in MRI blood flow acquisition, and have in-depth knowledge of diagnosis and treatment of cardiovascular diseases. The feedback obtained from these evaluation studies have yielded valuable insights concerning the presented visualization and interaction techniques. Furthermore, we have extended the use of the virtual probe, visualizing the fourdimensional MRI blood-flow in a similar way as used with color Doppler ultrasound imaging. Ultrasound is an established technique for blood-flow measurements, and the typical red-blue visualizations are familiar to the physicians. We introduce a compound view with different visualizations, inspired by ultrasound imaging, while exploiting the merits of the volumetric MRI blood-flow velocity data. All presented visualization techniques perform in real-time, enabling interactive exploration of the four-dimensional blood-flow data. The renditions update instantaneously when moving the probe, or when the user parameterizes the visualization. Furthermore, real-time interaction with the virtual camera facilitates the visual inspection, providing different viewpoints and enhancing perception of depth in the animated volumetric representations of the blood-flow. To achieve this performance, we have employed modern consumer graphics hardware for our visualizations, enabling parallel processing of the graphics and associated algorithms. Based on the evaluation studies with the involved physicians, we believe that realtime exploration of time-resolved volumetric blood-flow data, by means of illustrative visualizations, facilitates qualitative analysis of the hemodynamic behavior. We were able to present exemplary pathological cases. Time will reveal what new insights can be obtained by means of exploratory qualitative analyses.