Скачать презентацию Hybrid Visualization for White Matter Tracts using Triangle Скачать презентацию Hybrid Visualization for White Matter Tracts using Triangle

2f72e59c62322f26bc8062443fa7fb4e.ppt

  • Количество слайдов: 24

Hybrid Visualization for White Matter Tracts using Triangle Strips and Point Sprites D. Merhof Hybrid Visualization for White Matter Tracts using Triangle Strips and Point Sprites D. Merhof 1, 2, M. Sonntag 1, F. Enders 1, 2, C. Nimsky 2, P. Hastreiter 1, 2, G. Greiner 1 1 Computer Graphics Group 2 Neurocenter, Dept. of Neurosurgery University of Erlangen-Nuremberg

Outline Motivation Diffusion tensor imaging Fiber tracking 3 D visualization approaches Results Conclusion Acknowledgements Outline Motivation Diffusion tensor imaging Fiber tracking 3 D visualization approaches Results Conclusion Acknowledgements 2 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Motivation White matter tracts Neural structures within the human brain Transmission of motor, sensor Motivation White matter tracts Neural structures within the human brain Transmission of motor, sensor and other impulses Tract systems: Pyramidal tract Corpus callosum Optic tract 3 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Motivation Reconstruction and visualization of white matter tracts of interest for different research areas: Motivation Reconstruction and visualization of white matter tracts of interest for different research areas: Development of the brain Study of brain diseases Neurosurgery 4 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

5 5

Diffusion tensor imaging (DTI) MR imaging sequence (echo planar imaging) Measures diffusion of water Diffusion tensor imaging (DTI) MR imaging sequence (echo planar imaging) Measures diffusion of water molecules White matter consists of long, cylindrical cells Diffusion within white matter is anisotropic Diffusion tensor: Gaussian diffusion per voxel Isotropic diffusion 6 Anisotropic diffusion Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Recostruction - Fiber Tracking Basic Idea: Reduce tensor field to vector field of major Recostruction - Fiber Tracking Basic Idea: Reduce tensor field to vector field of major eigenvector Vector field of main diffusion Perform streamline propagation 7 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Visualization Lines + Fast rendering - Gaps, if lines wider than one pixel - Visualization Lines + Fast rendering - Gaps, if lines wider than one pixel - Width fix, independent of distance Tubes Lines + Significantly improved Visualization - Slow rendering, due to geometric complexity Tubes 8 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Visualization - Idea Rectangular cross section of tube for v ┴ t Use triangle Visualization - Idea Rectangular cross section of tube for v ┴ t Use triangle strips Spherical cross section of tube for v || t Use point sprites Hybrid combination of both techniques 9 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Triangle strips - construction 10 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • Triangle strips - construction 10 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Triangle strips - texturing 1 D texture with luminance and alpha channel Textures for Triangle strips - texturing 1 D texture with luminance and alpha channel Textures for ambient und specular illumination Decreasing value when approaching the border (sine function) channel 1 channel 2 diffuse specular 1 D texture 11 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Flipping artifacts Occur when viewing and segment direction are nearly parallel Use point sprites Flipping artifacts Occur when viewing and segment direction are nearly parallel Use point sprites in this case textured quads, rotated towards camera resemble cross section of tube 12 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Point Sprites - Texturing 2 D-RGB-Texture: draw and illuminate point sprites R channel: alpha Point Sprites - Texturing 2 D-RGB-Texture: draw and illuminate point sprites R channel: alpha channel G / B channel: diffuse / specular illumination 13 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Point Sprites - Rotation Rotate point sprite according to segment direction: Project fiber tangent Point Sprites - Rotation Rotate point sprite according to segment direction: Project fiber tangent t to image plane tp Normalize tp 0 Alpha provides rotation of segment 14 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Point Sprites When v close to t: Staggered point sprites Results in natural appearance Point Sprites When v close to t: Staggered point sprites Results in natural appearance of a tube v Point Sprites 15 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Visualization – Hybrid approach Triangle Strips Represent rectangular cross section Displayed for v t Visualization – Hybrid approach Triangle Strips Represent rectangular cross section Displayed for v t < 0. 98 Point Sprites Represent circular cross section Displayed for v t > 0. 93 16 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Terminating textured quads Problem – Beginning and end of each fiber No tube-like appearance Terminating textured quads Problem – Beginning and end of each fiber No tube-like appearance Solution Terminating textured quads 17 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Results Hybrid visualization approach based on triangle strips and point sprites provides High visual Results Hybrid visualization approach based on triangle strips and point sprites provides High visual quality (similar to tubes) Visual drawbacks of lines are circumvented High performance (due to reduced complexity) Lines 18 Tubes Hybrid Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Results Tracking of whole brain (4 500 fibers, 300 000 segments) Tubes : 16 Results Tracking of whole brain (4 500 fibers, 300 000 segments) Tubes : 16 ∆ / seg. ca. 4. 8 million ∆ Triangle strips : 2 ∆ / seg. ca. 0. 6 million ∆ Graphics card : Nvidia Geforce 6800 Ultra, 256 MB Viewport: 600 x 800 Pixel 19 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Results Clinical relevance: Combination with Volume Rendering Realtime visualization improves inspection of data Improved Results Clinical relevance: Combination with Volume Rendering Realtime visualization improves inspection of data Improved analysis of the data Tumor 20 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

21 21

Conclusion Improved visualization for streamlines based on DT-MRI data Technical aspects Hybrid approach instead Conclusion Improved visualization for streamlines based on DT-MRI data Technical aspects Hybrid approach instead of tubes/lines Rendering performance ≈ Lines Visual quality ≈ Tubes Medical aspects Improved analysis in real-time 22 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Fundings Graduiertenkolleg „Dreidimensionale Bildanalyse und –synthese“ Deutsche Forschungsgemeinschaft (DFG), SFB 603, Project C 9 Fundings Graduiertenkolleg „Dreidimensionale Bildanalyse und –synthese“ Deutsche Forschungsgemeinschaft (DFG), SFB 603, Project C 9 23 Computer Graphics Group + Neurocenter, Dept. of Neurosurgery • University of Erlangen-Nuremberg IEEE Vis 2006

Thank you! Questions? 24 Thank you! Questions? 24