Visual Computing for Medicine

1. Visual Computing for Medicine • Research Proposal in Germany (DFG) for establishing a priority programme • (co-authored with T. Ertl, H. Hagen, H.-C. Hege, G. Scheuermann, two medical doctors, a psychologist) • Major research areas: • Integration of information from different modalities • Perceptual and cognitive evaluation of visualization techniques • Integration of simulation and visualizaton for complex treatment planning • Intraoperative visualization (integrate intraoperative findings, update information from the planning stage) Bernhard Preim - 2007

2. Integration of Simulation and Visualizaton • Plenty of examples in medicine: • Implant design and placement benefits from biomechanical simulations • Simulation of thermal distribution in therapies, such as Radio-frequency ablation for tumor destruction • Blood flow simulations • Geometric models for the simulation as well as visualizing the results are challenging vis. problems. • Many research papers, e.g. Annals of Biomedical Engineering. No participation of Vis-Researchers!! Bernhard Preim - 2007

3. Integrating Simulation and Visualization: Treatment Planning for Cerebral Aneurysms • From: http://www.cfdrc.com/ Bernhard Preim - 2007

4. Introduction • Formation and development of vascular diseases (stenosis, aneurysmis, atherosclerosis, …) strongly depend on the local blood flow • Placement of stents and bypass surgery attempts to correct the blood flow in a defined way. • Flow speed influences the lumen. • Wall shear stress and vortiticity influence the formation and development of aneurysms (source: many studies and papers, e.g. in the American Journal of Neuroradiology) Bernhard Preim - 2007

5. Introduction • Differences between flexible and inflexible stents as well as between different stent geometries are investigated with CFD analysis. • From: [LaDisa06] Bernhard Preim - 2007

6. Introduction: cerebral aneurysms • Goals of Computer Support: • Improved understanding of cerebro-vascular disesases • Precise diagnostis based on personalized flow-models • Decisions on treatment options • Optimization, in particular minimally-invasive endovascular procedures • What has to be done? • Generation of patient individual vascular models as a basis for flow simulations • Manipulation of simulation models and –parameters to predict the consequences of treatment options Bernhard Preim - 2007

7. Introduction Risk of rupture is high • Different flow patterns related to aneurysms of very similar shape, size and location. From: [Cas07] Bernhard Preim - 2007

8. Basics Region of high wall shear stress (often only 1% of the overall size, often 5-10% of the overall wall shear stress) • Parts of a sac aneurysm, Source: [Ceb05] Maximum diameter Diameter of the neck • Localization: often in the Circle of Wilis, • Source: Wikipedia Bernhard Preim - 2007

9. Basics • Missing artery • Vessel anatomy differes strongly from patient to patient • Strong influence on the resulting flow • Circle of Wilis is complete and of normal topology in 40% of the patients only • Quelle: [Ceb01] Bernhard Preim - 2007

10. Image Acquisition • High-end modality rotation angiography • Example for a specific sequence: • Rotation angiography with 15 images per second which are acquired during a rotation (180-200 degrees) • Acquisition time: 8 seconds. • Reconstruction of a regular voxel grid with at least 300 slices • Optimal resolution: 0.1x0.1x0.3 mm Bernhard Preim - 2007

11. Model Generation • Steps: • Segmentation of relevant vessels • Generation and smoothing of surface models • Inserting clipping planes to restrict the simulation area • Improvement and refinement of the models • Meshing of inner structures Bernhard Preim - 2007

12. Our Project Bernhard Preim - 2007

13. Validation Images Courtesy: G. Janiga, Flow Simulation Department Bernhard Preim - 2007

14. Model Generation • Our Method (so far): • Segmentation with advanced region growing • Model generation with MPU implicit-variant (Schumann, EuroVis, 2007) • Postprocessing step to enhance triangle quality and support tesselation with tetrahedra Aneurysm, 1sec, Δ 53,948 Aneurysm, 5.3sec, Δ 61,324 • [Schumann et al., 2007] Bernhard Preim - 2007

15. weight SupportofQ(x) Q(x)=0 (local approximation with quadrics) Distance Q(x)>0 Q(x)<0 f(x)=0 Weighted average of local approximations Visualization with MPU Implicits Bernhard Preim - 2007

16. Model Generation Image Courtesy Ragnar Bade Bernhard Preim - 2007

17. Model Generation • Removal of small triangles • Edge Collapse and • Diagonal Swapping to avoid long edges • Movement of vertices in the center of their neighbors Bernhard Preim - 2007

18. First Results Images Courtesy Ragnar Bade Bernhard Preim - 2007

19. Simulation of Flow • Important simulation parameters: • Blood flow is considered as pulsatil flow (specific velocity profiles are considered) • Laminar or turbulent flow. Usually, the behavior is considered as laminar, which is mostly correct. • Blood is a Non-Newtonian fluid (Viscosity is not constant) • Vessel walls are elastic and change over time depending on the flow. This effect is largely ignored. Major results: • Direction and magnitude of blood flow • Wall shear stress Bernhard Preim - 2007

20. First results • Assumptions: • Inlet: A, B, C (Peak v= 80 cm/s) • Outlet: D, E, F, G - Non-Newtonian behavior • - Laminar flow (Reynolds number: ~1000, turbulent behavior > 2500) Images Courtesy: G. Janiga, Flow Simulation Department Bernhard Preim - 2007

21. Results: Wall Shear Stress Images Courtesy: G. Janiga, Flow Simulation Department Bernhard Preim - 2007

22. Visualization of Simulation Results • Scalar values (velocity, wall shear stress) and vector values (direction of flow) • Visualizations in cross-sections and 3D-visualizations. • Color coding and isolines for the display of scalar values • Streamlines for blood flow simulation • All values are time-dependent! Visualization at selected points in time or animation over the heart cycle. Bernhard Preim - 2007

23. Visualization of Simulation Results • Research Goals: • Select „right“ points in time, „right“ cross-sections • Generate appropriate animations efficiently • Use smooth vessel visualization in combination with the visualization of the results. Bernhard Preim - 2007

24. Concluding Remarks CFD analysis enable an evaluation of blood flow patterns. On an individual basis, these may (in the long term) be used for treatment planning. Visualization challenges are: • Geometric reconstruction of simulation models, • Visualization of simulation results along with the vascular anatomy • To provide appropriate interaction techniques. Bernhard Preim - 2007

25. Concluding Remarks • More general: • There is plenty of problems in medical research and treatment planning where simulation and visualization should be closely connected. • So far, the only the output options of FEM tools are used. • How can we contribute: • Willingness to understand the underlying problems • Willingness to tackle also non-vis problems, such as image analysis and user interface development Bernhard Preim - 2007

26. Contributions Desired • www.medvis-book.de • Interesting staff: • links to software, • videos, • PDF-documents, • e.g. on Illustrative Medical Visualization, DTI, (Logarithmic) Transfer Functions, Volume Rendering, Normal Estimation Schemes, …) Bernhard Preim - 2007

27. Acknowledgement • M. Skalej,Ö. Gürvit (Neuroradiology) R. Bade D. Thevenin (Fluid Simulation) • T. Bölke, G. Rose (Medical Technology) • Christian Hege and Stefan Zachow for discussions on simulation and visualization as well as grid generation. Bernhard Preim - 2007