Advanced Visual Medicine: Techniques for Visual Exploration & Analysis

Transcription

1 Advanced Visual Medicine: Techniques for Visual Exploration & Analysis Interactive Visualization of Multimodal Volume Data for Neurosurgical Planning Felix Ritter, MeVis Research Bremen

2 Multimodal Neurosurgical Planning 2

3 Multimodal Neurosurgical Planning Is a resection possible? What type of resection can be performed? What is the risk of the surgical intervention? What is an optimal access path to the lesion? 2

4 Topics to Work on Combining multimodal volume data Enhancement of risk structures Accentuation of spatial relations Reduction of interaction for exploration Visualization of access path to lesion Supporting brain-surface intervention 3

5 Medical Volume Data Common medical volume data: MRI (magnetic resonance imaging), high soft tissue contrast for visualization of anatomical details fmri (functional magnetic resonance imaging), for detection of the brain s activation areas DTI (diffusion tensor imaging), for reconstruction of the brain s nerve tracts CT (computer tomography), just used in special cases due the high radiation exposure, e.g. skull bone infiltrated by tumor tissue PET (positron emission tomography), nuclear imaging technique to detect functional processes 4

6 Multimodal Neurosurgical Planning A tempting assumption: - Combination of all available imaging modalities can identify all relevant structures! Limitations: - Models behind the involved techniques fmri can give hints about functional areas (Accumulation of oxygen) Fiber-Tracking can give hints about axonal pathways (Diffusion of water) - Usefulness of combinations A better interpretation: - Combination of different images may give additional information but may also introduce limitations and additional complexity 5

7 Multimodality Visualization 6

8 Multimodality Visualization 6

9 Multimodality Visualization? 6

10 Multimodal Volume Rendering Visualization of CT & MR data for visualization of implanted electrodes for epilepsy surgery CT, MR & MRA MR (black/white), PET (red) & fmri (yellow and white) Johanna Beyer et al., VIS

11 Multimodal Volume Rendering Combined MRI, fmri visualization Friedemann Rößler et al., SimVis 2006 Werner Jainek et al., EuroVIS

12 Multimodal Volume Rendering MRI, fmri & DTI Alexander Köhn et al., EG 07 9

13 Dealing with Inaccuracies in Multimodal Neurosurgical Planning Accuracy is limited! Sources for limitations are manifold Limited accuracy may become an issue - If the awareness about the limitations gets lost - If the impression of high accuracy is falsely created - If decisions are made, that are not valid based on the given accuracy Limited accuracy turns into inaccuracy Inaccuracy is unavoidable We must be aware of it! 10

14 Sources of Inaccuracies Image registration - Spatial alignment is prerequisite for overlaying different images - Automatic rigid registration 11

15 Sources of Inaccuracies Image registration - Spatial alignment is prerequisite for overlaying different images - Automatic rigid registration 11

16 Sources of Inaccuracies Image registration - Spatial alignment is prerequisite for overlaying different images - Automatic rigid registration Problem: - Spatial deformation 12

17 Sources of Inaccuracies Image registration - Spatial alignment is prerequisite for overlaying different images - Automatic rigid registration Problem: - Spatial deformation 12

18 Sources of Inaccuracies Image registration - Spatial alignment is prerequisite for overlaying different images - Automatic rigid registration Problem: - Spatial deformation Problem: - Different voxel-sizes, slicethickness, interslice-gap 13

19 Sources of Inaccuracies Image registration - Spatial alignment is prerequisite for overlaying different images - Automatic rigid registration Problem: - Spatial deformation Problem: - Different voxel-sizes, slicethickness, interslice-gap 13

20 Extraction of Anatomical Structures T1-weighted ce T1-weighted substraction vessel mask T2-weighted FLAIR extraction brain mask masking masking masking masking 14

21 Risk Structures MRI fmri DTI anatomical data activation areas nerve tracts vessel and brain mask volume rendering fmri visualization fiber tracking 15

22 Multimodal Volume Visualization 16

23 Multimodal Volume Visualization Combined exploration of inner risk structures and brain surface anatomy difficult 16

24 Multimodal Volume Visualization Combined exploration of inner risk structures and brain surface anatomy difficult Simple solution: Dual views - Internal view: visualization of internal structures (risk structures, occluded by skull and brain) - External view: visualization of opaque anatomy (skull / brain). Requires cutting tools for exploration of inner structures. 16

25 Visualization Complexity Not all areas of the brain are of interest, e.g: - Fading of color saturation of fmri areas far away of the ROI and trajectory - Fibers far away are just visualized as outlined silhouettes - Using distance-based selection of transfer function and shader 17

26 Visualization Complexity Not all areas of the brain are of interest, e.g: - Fading of color saturation of fmri areas far away of the ROI and trajectory - Fibers far away are just visualized as outlined silhouettes - Using distance-based selection of transfer function and shader 17

27 Fiber Clustering Group geometrically similar or related fibers acquired by DTI. Improve perception of fiber bundles and connectivity. Improve interaction with fiber bundles. Avoid user-biased quantification results. Klein et al., SPIE

28 Improving Localization Depth perception of individual structures within the brain s hull difficult Visualizing the principal axes of the volume restricted by brain s hull improves localization 19

29 Distance Ring At some orientations one of the axes could be hidden or occluded by other structures Distance Ring indicates location of ROI in view direction - Minimal distance: distance ring is completely open - Maximal distance: distance ring is completely closed 20

30 Distance Ring At some orientations one of the axes could be hidden or occluded by other structures Distance Ring indicates location of ROI in view direction - Minimal distance: distance ring is completely open - Maximal distance: distance ring is completely closed 20

31 Distance Ring Tracing ray through brain mask front [0] position [0..1] back [1] 21

32 Superficial Landmarks Superficial landmarks at surface of skull and brain to support navigation: Relate anatomical structures of head to those landmarks (e.g. nose, ears) Brain structures after opening the skull are known 22

33 Visualization of Contrast Enhanced Structures 23

34 Visualization of Contrast Enhanced Structures 23

35 Visualization of Contrast Enhanced Structures 23

36 Visualization of Contrast Enhanced Structures 23

37 Visualization of Contrast Enhanced Structures 23

38 Visualization of Contrast Enhanced Structures 23

39 Visualization of Contrast Enhanced Structures 23

40 Visualization of Contrast Enhanced Structures 23

41 Automatic Transfer Function Adaptation of Diseased Tissue 24

42 Automatic Transfer Function Adaptation of Diseased Tissue 24

43 Automatic Transfer Function Adaptation of Diseased Tissue 24

44 Automatic Transfer Function Adaptation of Diseased Tissue 24

45 Automatic Transfer Function Adaptation of Diseased Tissue 24

46 Automatic Transfer Function Adaptation of Diseased Tissue 24

47 Automatic Transfer Function Adaptation of Diseased Tissue weighted FLAIR masking 25

48 Automatic Transfer Function Adaptation of Diseased Tissue weighted FLAIR masking 25

49 Automatic Transfer Function Adaptation of Diseased Tissue weighted FLAIR masking 25

50 Automatic Transfer Function Adaptation of Diseased Tissue weighted FLAIR masking 25

51 Two-stage Rendering Pipeline Rieder et al., VCBM

52 Automatic Axis-Aligned Clipping 27

53 Automatic Axis-Aligned Clipping How to reduce interaction required for exploration? 27

54 Automatic Axis-Aligned Clipping How to reduce interaction required for exploration? Dividing rendered volume into octant sectors 27

55 Automatic Axis-Aligned Clipping How to reduce interaction required for exploration? Dividing rendered volume into octant sectors Automatically discard sector located between view point and center of some ROI 27

56 Automatic Axis-Aligned Clipping 28

57 Automatic Axis-Aligned Clipping 28

58 Automatic Axis-Aligned Clipping Partitioning space into 26 different sectors 29

59 Automatic Axis-Aligned Clipping 30

60 Visualization of the Access Path 31

61 Visualization of the Access Path How to support planning of access to a deep-seated lesion? 31

62 Visualization of the Access Path How to support planning of access to a deep-seated lesion? Virtual access path from incision point to target Rieder et al., EuroVIS

63 Visualization of the Access Path How to support planning of access to a deep-seated lesion? Virtual access path from incision point to target Orthogonal cutting plane along the trajectory for detailed exploration Rieder et al., EuroVIS

64 Visualization of the Access Path The virtual access path: simplified cylinder geometry from incision point to ROI for visualization Pincision PROI Fdistance 32

65 Visualization of the Access Path Problem: Incorrect illumination of homogeneous regions (white matter, grey matter) due to ill-defined gradients at cuts Solution: Transfer normals of cutting geometry into cutting surface of volume (consistent shading [Weiskopf et al., TVCG 2003]) 33

66 Visualization of the Access Path 34

67 Brain Peeling 35

68 Brain Peeling How to explore anatomical details in the vicinity of the brain surface with cutting planes? 35

69 Brain Peeling How to explore anatomical details in the vicinity of the brain surface with cutting planes? Peel brain surface step by step: - Compute distance field (DF) from brain mask - Discard voxel if value in DF lower than distance threshold 35

70 Brain Peeling Rieder et al., VCBM

71 Conclusions Transfer function adaptation important for robustness Automated volume clipping facilitates exploration Virtual access path could become integral part of preoperative planning Trajectory aligned cutting plane supports inspection along access path Brain peeling allows the surgeon to inspect anatomical details in the vicinity of the brain surface Awareness about the limitations important! 37

72 Acknowledgments This presentation would have been impossible without the support of my colleagues: Horst Hahn Jan Klein Alexander Köhn Christian Rieder Florian Weiler 38