Intraoperative Adaptation and Visualization of Preoperative Risk by uhb20986

VIEWS: 4 PAGES: 10

									          Intraoperative Adaptation and Visualization
   of Preoperative Risk Analyses for Oncologic Liver Surgery
          Christian Hansen*a , Stefan Schlichtingb , Stephan Zidowitza , Alexander K¨hna ,
                                                                                    o
                 Milo Hindennach  a , Markus Kleemannb , and Heinz-Otto Peitgena

           aMeVis Research - Center for Medical Image Computing, Bremen, Germany;
      b                                                                   u
          University Hospital Schleswig-Holstein, Department of Surgery, L¨beck, Germany


                                                 ABSTRACT
Tumor resections from the liver are complex surgical interventions. With recent planning software, risk analyses
based on individual liver anatomy can be carried out preoperatively. However, additional tumors within the
liver are frequently detected during oncological interventions using intraoperative ultrasound. These tumors are
not visible in preoperative data and their existence may require changes to the resection strategy. We propose
a novel method that allows an intraoperative risk analysis adaptation by merging newly detected tumors with a
preoperative risk analysis. To determine the exact positions and sizes of these tumors we make use of a navigated
ultrasound-system. A fast communication protocol enables our application to exchange crucial data with this
navigation system during an intervention.
    A further motivation for our work is to improve the visual presentation of a moving ultrasound plane within
a complex 3D planning model including vascular systems, tumors, and organ surfaces. In case the ultrasound
plane is located inside the liver, occlusion of the ultrasound plane by the planning model is an inevitable problem
for the applied visualization technique. Our system allows the surgeon to focus on the ultrasound image while
perceiving context-relevant planning information. To improve orientation ability and distance perception, we
include additional depth cues by applying new illustrative visualization algorithms.
    Preliminary evaluations confirm that in case of intraoperatively detected tumors a risk analysis adaptation
is beneficial for precise liver surgery. Our new GPU-based visualization approach provides the surgeon with
a simultaneous visualization of planning models and navigated 2D ultrasound data while minimizing occlusion
problems.
Keywords: Visualization, Intraoperative Imaging, Treatment Planning, Ultrasound Guidance

                                           1. INTRODUCTION
With regard to complex oncological resections, a computer-assisted risk analysis is important for precise and safe
liver surgery. Based upon security margins around each tumor, disturbances of blood supply and drainage within
the remaining liver parenchyma can be computed, quantified and visualized preoperatively.1 However, using
intraoperative ultrasound during oncological interventions, approximately 20% of patients with primary liver
malignancies or metastases show additional tumors.2–4 Although the maximum diameter of these intraoperative
findings is mostly smaller than 15 mm, changes to the resection strategy may be necessary, especially if the
new tumors are adjacent to hepatic vessels. In such cases, an intraoperational tool for automatic risk analysis
adaptation would assist the surgeons in optimizing the resection strategy. We propose a new application, the
Intraoperative Planning Assistant, which allows a risk analysis adaptation by merging intraoperatively detected
tumors with a preoperative risk analysis. To determine the exact positions and sizes of newly found tumors, we
make use of an navigated ultrasound-system.5 Data exchange between Intraoperative Planning Assistant and
navigation system during an intervention is realized using a fast XML-based communication protocol.
   Furthermore, an adequate visualization of the ultrasound plane relative to the preoperative model facilitates
the navigation for the surgeon and is beneficial for precise liver surgery. Major drawbacks of existing concepts
are occlusions of the ultrasound plane by the 3D planning model and fade-out of crucial context information.
   * Corresponding author. E-mail: Christian.Hansen@mevis.de; Tel.: ++49 421-2187722
We introduce a new visualization approach, that allowing a focus-view on the ultrasound image while preserving
context-relevant planning information and depth cues. The paper is organized as follows: First, we review
related work. After describing the applied registration technique, we introduce a new approach for intraoperative
adaptation of risk analysis in the case of intraoperatively found tumors. Furthermore, we introduce requirements
for an adequate, intraoperative visualization and describe a new visualization pipeline. Finally, we present our
results, draw conclusion and discuss future work.

                                           2. RELATED WORK
The work described in this paper extends and combines existing concepts in medical image processing and
visualization. In this section we provide an overview of related work in the field of intraoperative adaptation and
registration, as well as illustrative and intraoperative visualization.

2.1 Intraoperative Adaptation and Registration
An initial concept for intraoperative adaptation of risk analysis for liver surgery was presented by Ritter et al.6
They propose to determine tumor attributes like position and size manually by means of mouse interactions on a
notebook inside the operation room (OR). Their approach is not designed for the use in sterile areas and does not
make use of a navigation system to determine tumor attributes. Since our approach is based on an ultrasound-
based navigation system, the success of an intraoperative risk analysis adaptation correlates with the accuracy of
the applied intraoperative registration technique. Maier-Hein et al.7 developed a motion simulator for CT-guided
liver interventions allowing a real-time adaptation of preoperative calculated tumor positions. Their system is
                                                                                                          u
of particular importance for radio-frequency ablations where liver immobilization is hard to archive. Z¨hlke et
al.8 presented a model-based technique based on Self Organized Maps, using extracted vessel branchings from
preoperative (CT) and intraoperative (tracked 2D Doppler ultrasound) data as input. The method outputs a
rigid registration matrix. Lange et al.9 combined the Iterative Closest Point algorithm and multilevel B-Splines
                                                                                                             u
to achieve a non-rigid registration of preoperative CT/MRI and intraoperative 3D ultrasound. Similar to Z¨hlke
et al. the algorithm is based on the vessel center lines of both image modalities. Papenberg et al.10 introduced
a landmark-based registration approach for ultrasound and CT-data. It combines the distance measure of the
normalized gradient field with a penalizer that forces the deformation to fulfill the landmark condition using an
affine linear deformation model.

2.2 Illustrative Visualization
The rendering of overlapping transparent iso-surfaces like the segmentation result of a vascular tree often gener-
ates a complex output which is difficult to interpret. Selective accentuation of shape features using illustrative
techniques like hatching strokes or silhouettes can reduce visual complexity. Visual complexity can further be
decreased by viewing important objects in focus while less important structures are visualized as context in-
formation or even discarded. In recent years a variety of non-photorealistic methods have been proposed to
increase expressiveness of illustrations, e.g. focus & context techniques, hatching strokes or silhouettes. Detailed
overviews have been given by Strothotte et al.11 and Gooch et al.12 In the field of medical visualization these
illustrative methods were adapted, improved, and combined to allow an effective exploration of medical data
sets.
    Our visualization approach was inspired by a sequence of user studies accomplish by Interrante et al.13 They
evaluated the effect of displaying surface texture in combination with overlapping transparent surfaces and showed
that non-photorealistic techniques are an adequate solution to support evident recognition of layered surfaces.
Corresponding to these studies, Fischer et al.14 introduced a pipeline for illustrative rendering of overlapping
iso-surfaces. Hatching, silhouette extraction, and depth-peeling are combined in such a way that a simultaneous
inspection of inner and outer structures of an object is possible. The pipeline allows a manual definition of focus
regions by specifying a secondary geometry, but is limited to two iso-surface layers. Tietjen et al.15 describes a
focus & context rendering approach for medical visualization that combines silhouettes, direct volume rendering,
and surface shading. Different styles are used to control the appearance of anatomical structures according to
their relevance for the visualization. A context-preserving volume rendering model that automatically enhances
important information in a volume data set using illustrative volume rendering was introduced by Bruckner et
al.16 In order to reduce the opacity in less important data regions, they use a function of shading intensity,
gradient magnitude, distance to the eye point, and previously accumulated opacity. A user-controlled focus
& context technique for volume rendering was presented by Kr¨ger et al.17 Changing attributes like size and
                                                                    u
location of the focus or weight for the context allows the user to control the appearance of a medical visualization
using a point-and-click interface.
2.3 Intraoperative Visualization
Previous publications in the field of medical visualization primary focus on preoperative exploration of medical
data. A detailed overview of these techniques can be found in the book of Preim et al.18 Since the availability of
navigation support in the OR has risen in the last years, intraoperative visualization becomes more interesting.
   One of the first approaches for intraoperative visualization in liver surgery with clinical relevance was pre-
sented by Lange et al.9 Registrated preoperative models of vessels, tumors, the liver surface, and the resection
plane are rendered as colored intersection lines onto the top and perpendicular slice of a 3D ultrasound image.
The current position of these two slices in relation to the preoperative data is presented in an extra viewer.
However,the approach does not consider an adequate visualization in a single view in order to avoid frequent
comparisons.
    Burns et al.19 presented an approach for simultaneous visualization of ultrasound and medical volume data.
To provide a clear view on the ultrasound plane, they removed occluding volume data based on an importance
function. Furthermore, they apply illustrative rendering techniques like silhouette enhancement and Gooch cool-
to-warm shading to provide visual distinction between anatomic structures. Although their aim is similar to
ours, we use segmented iso-surface objects with additional, color-coded planning information (e.g. distance to
relevant structures like tumors or organ surface) for intraoperative visualization. Compared to Burns et al. we
do not clip any planning data, but carefully fade out planning data that occludes the view between camera and
ultrasound plane by applying illustrative rendering techniques.

                                                3. METHODS
In this section we briefly review prior work concerning the underlying risk analyses algorithm for planning
models. Afterwards, we present a new approach for intraoperative risk analyses adaption and describe the
applied registration technique. Furthermore, we define requirements for an intraoperative visualization and
introduce a novel rendering pipeline that allows an adequate visualization of adapted risk analyses together with
intraoperative 2D ultrasound.
3.1 From Voxels to Risk Analyses
The generation of 3D planning models including risk analyses is part of the software assistant MeVisLiverAnalyzer
which integrates all steps to analyze contrast-enhanced CT data for preoperative planning in liver surgery.20 We
briefly describe the important steps for vascular analysis and visualization: Skeletonization with a topology-
preserving thinning algorithm yielding an exact centerline and the radius at each voxel of the skeleton. A graph
analysis transforms the vessel skeleton in a directed, acyclic graph where nodes represent branchings.21 Based
on the assumption of a circular cross-section of vasculature, smooth transitions at branchings and rounded ends
are produced by means of convolution surfaces.22
    Tumors and liver surface are semi-automatically segmented using Live Wire23 and transformed to a triangle
mesh. For the visualization of affected vessels the spatial relations between tumors and vessels have to be
analyzed. Therefore, certain security margins around the tumor are assumed. Computing the difference between
the tumor segmentation mask and the eroded tumor segmentation mask (3x3x3 erosion filter) delivers the border
of the tumors. Distance transformation is applied to all border voxels, thus generating the security margins.1
Finally, a graph traversion identifies affected voxels of depending vascular branches according to the specified
security margins and determines the vertex colors for the convolution surfaces. Based on the eight major branches
of the portal vein (giving a classification similar to Couinaud’s scheme) vascular territories and their volumes
are approximated using a Voronoi Tessellation. An iso-surface renderer generates separate triangle meshes and
sets vertex colors. While vascular territories show the relation between liver parenchyma and supplying/draining
branches, territories at risk show the liver parenchyma of impaired blood in- or outflow. Thus, the calculation of
territories at risk (as the affected vessels) has to be adapted when new tumor findings are made during operation.
3.2 Intraoperative Adaption of Risk Analyses
If additional tumors are found during surgery, tumor size, and position are directly defined in the 2D ultrasound
image on the navigation system’s display which is located in front of the situs. For simplicity, tumors are assumed
to be spherical in shape and it is therefore sufficient to draw a circle around the tumor. After tumor attributes
have been determined, the navigation system sends an XML-request to the Intraoperative Planning Assistant,
containing the new tumor radius and position. In case the request is accepted, the tumor is added to the planning
model and a risk analysis adaptation is performed by the Intraoperative Planning Assistant. To create a new
risk analysis, an Euclidean Distance Transformation is applied to calculate the distances between vessels and
tumors. Affected vessels are identified and relabeled depending on the predefined security margins (Fig. 1a, 1b).




                                (a)                                                  (b)
Figure 1. (a) Preoperative risk analysis for the portal vein vessels and dependent subtrees for different security margins.
Three standard security margins around the tumors are chosen and displayed in red (5mm), yellow (10mm) and green
(15mm). The red sphere on the ultrasound plane shows an intraoperatively detected tumor before adapting the risk
analysis. (b) Adapted risk analysis including relabeled vessels.




                                 (a)                                                 (b)
Figure 2. (a) Preoperative territories at risk that are supplied by the affected vessels. (b) Adapted risk analysis showing
the updated territories at risk. Tumor and risk margins are the same as in Fig. 1.
   Furthermore, a Voronoi Tessellation with respect to the vascular system is performed to approximate the
volume of the parenchyma supplied or drained by the affected vessels (Fig. 2a, 2b). After the computation
has finished, the Intraoperative Planning Assistant sends a response to the navigation system and the updated
planning results are immediately available on the navigation display. Using a multi-scale approach for the
underlying volume data, our algorithm for risk analysis adaptation can choose a trade off between accuracy
and computation time. Figure 3 illustrates the XML-based exchange protocol, defined between Intraoperative
Planning Assistant and navigation system.

                     Intraoperative
                   Planning Assistant                                         navigation system
                                            request planning data
                   allocate planning data
                                             data location of planning data   use tracked ultrasound
                                                                              to look for new lesions



                                             tumor position and radius        define tumor attributes

                                                              acknowledge
                      adapt risk analysis
                                                                              use tracked ultrasound
                                             inform about new planning data

                                             acknowledge                      download new
                                                                              planning data

        Figure 3. Exchange protocol, defined between Intraoperative Planning Assistant and navigation system.


3.3 Intraoperative Registration
In order to calculate the exact position of a newly detected tumor, a registration between the intraoperative
ultrasound images and the preoperative planning data is required. We implemented a registration method that
is robust, executable in real-time and interactively adaptable. It requires the surgeons to define a small set of
corresponding markers in both the preoperative radiological data and the intraoperative ultrasound images using
anatomic landmarks like vessel branchings. By minimizing total squared distances of the corresponding markers,
we compute an affine transformation and apply it to the preoperative data.

3.4 Intraoperative Visualization
In case the ultrasound plane is located inside the liver, occlusion of the ultrasound plane by the planning model is
an inevitable problem for the applied visualization technique. Presenting the ultrasound image without planning
data in a second window results in time-consuming comparisons for the observer because necessary information
is not presented in a single image. Applying transparency or clipping operations on occluding objects removes
important context information and results in a loss of spatial information and clearness. To address these
problems, we define three guiding requirements to ensure clinical applicability of our method:

  1. Diagnostic Usability: While the whole visualization is presented in a single view, ultrasound information
     should always be visible, even if it would be occluded by planning data or parts thereof.
  2. Orientation Aid: Spatial relations between ultrasound plane and planning data should be clearly per-
     ceivable without rotating or translating the camera.
  3. Error Identification: Since recent intraoperative registration methods make a compromise between real-
     time and error-prone computation, a clear hint to perceive registration errors should be provided.

   The aim of visualization approach is to provide an adequate presentation of planning models and ultrasound
data in a single view. Therefore, we combine three extended rendering techniques in a GPU-accelerated rendering
pipeline (Fig. 4). This pipeline consists of the following four steps in which the first three steps directly address
the mentioned requirements:
3.4.1 Focus, context and mask image generation
The main objective of our visualization approach is to provide the surgeon with a focus-view on the ultrasound
plane and a context-view on the planning model in order to ensure diagnostic usability. Since the focus lies
on the ultrasound plane, we create a mask by first rendering this plane with disabled texturing onto an off-
screen texture M. For the generation of the focus and context views, we render the colored and shaded planning
models onto two off-screen textures F and C. For F, the models are transparent. Correct order-independent
transparency is achieved via depth-peeling.24 Using the mask M to combine F and C would lead to an abrupt
change of rendering styles at the border of the ultrasound plane. This causes a wrong spatial interpretation due
to receiving the impression that the ultrasound plane occludes the vessels. We found that a gradual transition
between focus and context images avoids this misinterpretation. To achieve this, the mask M is of low resolution.
When sampling from M, we use fast bi-cubic interpolation as described by Hadwiger et al.25 to obtain a smooth
gradient. Thus, we avoid several blurring passes which are required when a higher resolution is utilized.

3.4.2 Silhouette and hatch stroke generation
We found that transparent layered surfaces in combination with shape-accentuating strokes give a context-
preserving view of the planning data while providing orientation aid and minimizing occlusion problems. Fur-
thermore, we suggest varying line width of silhouettes depending on the distance to the ultrasound plane to
support depth perception. In a first step, we render the depth values Dstored of the vessels into an off-screen
texture. While D is the relative distance between the current vertex and the ultrasound plane, and R describes
the diameter of the bounding sphere, we use a vertex shader to translate all vertices of the model by the factor
DR in opposite direction to their normal vector. This causes a shrinking of the model depending on the distance
to the ultrasound plane. The corresponding fragment shader computes the difference of the current depth value

                                                              Orientation Aid

             Discretized Z-Buffer           Discretized Z-Buffer
                                                                   NPR-Image                       Colored NPR-Image
                                             of shrinked model

                                                                           N    If (N=black)
                                                                                Nα+C(1-α)
                                                                                      αC
                   Dstored (D                  Z
                              stored-Z)*σ
                                                       clipping


                        Diagnostic Usabilty                                                         Error Identification
                                                                           C
                                                                           F Cα+F(1-α)         +
                                                                                  α

                                                                                                     ultrasound image &
                                                                                                    intersection contours

                Context image                      Focus image
                  (opaque)                         (transparent)


                             Rescale &
                          cubic interpolation
            ultrasound plane
            as 16x16 texture




    Figure 4. Pipeline overview showing intermediate rendering results together with the associated requirements.
Z and Dstored which delivers a distance encoding silhouette in image space precision, i.e., the thickness decreases
with increasing distance to the ultrasound plane. We extend the method by applying hatching strokes within
the same rendering step, using the hatching algorithm introduced by Ritter et al.26

3.4.3 Intersection contour generation
This is done by first cutting an image representation of the ultrasound plane out of a 3D binary vessel mask
(applying a multiplanar reformation). Then a marching squares algorithm is used to find all iso-contours in the
image. Thus, the registration result can be compared with the current ultrasound image (error identification).

3.4.4 Compositing in screen space
A fragment program starts by computing the output color colM odel using a convex combination of the current
F and C samples. The value sampled from the mask image M is used as the weight α. Recall that we use bi-
cubic interpolation when sampling from M to achieve a smooth transition between F and C. Next, the program
computes the contribution from N to the current fragment color. If the sample of N is black the program outputs
colN instead of colM odel , where colN is determined by blending black with the color of C using α. Figure 6 shows
a resulting image generated with the technique described above.

                                                   4. RESULTS
So far, preoperative risk analyses for oncologic liver surgery were of limited value during interventions in case
of intraoperatively found tumors. For the first time, we provide surgeons with an intraoperational tool for risk
analyses adaptation that can be integrated into a surgical workflow. Preliminary evaluations (Fig. 5) confirm
that in case of newly discovered tumors an adaptation of the preoperative risk analysis is beneficial for precise
liver surgery. Our application, in combination with a navigated ultrasound-system, offers crucial decision support
and is easy to use. Using an XML-based communication protocol, the provided service can be utilized by other
systems as well.
    Furthermore, we present methods for simultaneous visualization of preoperative planning models and nav-
igated 2D ultrasound. Our new GPU-accelerated visualization approach is guided by requirements of clinical
applicability and provides the surgeon with a focus-view on a moving ultrasound plane within a complex, inter-
weaving 3D planning model (Fig. 6). Unlike previous visualization approaches, we minimize occlusions of the
ultrasound plane by the 3D planning model and avoid fade-out of crucial context information. With a prototype
implementation of our pipeline on the developer platform MeVisLab∗ we achieve interactive framerates (16fps
on a GeForce 7900 GT) with a 2 GHz CPU using a model with 203k vertices (four depth-peel passes and one
anti-antialiasing pass). We expect real-time framerates after optimization of the code.

                                5. DISCUSSION AND FUTURE WORK
Preliminary evaluations with ex-vivo pig livers have been performed. The purpose was to prove the feasibility
of our concept in collaboration with liver surgeons as well as an analysis of the surgical workflow for the intra-
operative risk analysis adaptation. For the near future we plan to evaluate our system under realistic conditions
during oncological interventions with a focus on accuracy. Therefore, we will replace the spherical approximation
of tumors by a tumor segmentation algorithm applied on intraoperative 3D ultrasound data using a predefined
sphere as the region of interest. Since our system uses a rigid registration approach, which does not provide an
adequate accuracy for clinical use, we want to extend our system by an intensity-based elastic registration.10
Regarding clinical applicability, a trade-off between registration accuracy and computation time needs to be
found.
    Concerning our new visualization approach, we expect additional clues from a quantitative user study which
is currently performed in cooperation with clinical partners. Furthermore we will extend the approach with an
automatic viewpoint selection algorithm27 that computes optimal camera positions for a 3D scene with respect
to clinical parameters, like newly found tumors.
  ∗
      For more information visit www.mevislab.de
                               (a)                                                       (b)
Figure 5. (a) Intraoperative Planning Assistant showing the risk analysis for a pig’s liver with artificial tumors before
applying a risk analysis adaptation. (b) Preliminary evaluations in the OR using the Intraoperative Planning Assistant
and the ultrasound-based navigation system for laparoscopic liver surgery.




Figure 6. Simultaneous visualization of preoperative planning data and intraoperative 2D ultrasound. The close-up view
on the right hand side shows the highlighted intersection areas and depth-accentuating hatching strokes. While silhouettes
encode the distance to the ultrasound plane, a gradual transition from opaque to transparent rendering (around the focused
ultrasound plane) avoids misinterpretation of spatial relations.
                                        ACKNOWLEDGMENTS
This work was funded by the Federal Ministry of Education and Research (SOMIT-FUSION project FKZ
01—BE03C). The authors thank Mathias Markert, Stefan Weber (Technical University Munich, Germany),
                                                 u
Armin Besirevic, Volker Martens (University of L¨beck, Germany), Holger Bourquain, Horst Hahn, Olaf Kon-
rad, Guido Prause, and Felix Ritter (MeVis Research, Bremen, Germany) for their valuable ideas and support
                                                                          e
in performing this work. We also express gratitude to Wolfram Lamad´ (RBK Stuttgart, Germany) and all
involved physicians of the FUSION-project for fruitful discussion and clinical advice.

                                              REFERENCES
 1. B. Preim, H. Bourquain, D. Selle, K. Oldhafer, and H.-O. Peitgen, “Resection proposals for oncologic liver
    surgery based on vascular territories,” in International Journal of CARS, pp. 1280–1283, 2002.
 2. W. Bloed, M. S. van Leeuwen, and I. B. Rinkes, “Role of intraoperative ultrasound of the liver with improved
    preoperative hepatic imaging,” European Journal of Surgery, vol. 166, pp. 691 – 695, 2000.
 3. M. Cohen, M. Machado, and P. Herman, “The impact of intra operative ultrasound in metastases liver
    surgery,” Arq Gastroenterol., vol. 42(4), pp. 206–12, 2005.
 4. J. Ellsmere, R. Kane, R. Grinbaum, M. Edwards, B. Schneider, and D. Jones, “Intraoperative ultrasonog-
    raphy during planned liver resections: why are we still performing it?,” Surg Endosc., pp. 353–358, 2007.
 5. P. Hildebrand, S. Schlichting, V. Martens, A. Besirevic, M. Kleemann, U. Roblick, L. Mirow, C. B¨rk,      u
    A. Schweikard, and H. Bruch, “Prototype of an intraoperative navigation- and documentation system for
    laparoscopic radiofrequency ablation first experiences,” European Journal of Surgical Oncol., in press, 2007.
 6. F. Ritter, M. Hindennach, W. Lamade, K. Oldhafer, and H.-O. Peitgen, “Intraoperative adaptation of
    preoperative risk analysis in oncological liver surgery,” in Proceedings of CURAC, 2005.
                         u                               u
 7. L. Maier-Hein, S. M¨ller, F. Pianka, A. Seitel, B. M¨ller-Stich, C. Gutt, U. Rietdorf, G. Richter, H. Meinzer,
    B. Schmied, and I. Wolf, “In-vitro evaluation of a novel needle-based soft tissue navigation system with a
    respiratory liver motion simulator,” in SPIE Medical Imaging, vol. 6509, pp. 16–1 – 16–12, 2007.
         u
 8. D. Z¨hlke, “Transform learning - registration of medical images using self organization,” in Workshop on
    Self-Organizing Maps, Bielefeld, Germany, pp. CD–ROM, 2007.
                                    u
 9. T. Lange, S. Eulenstein, M. H¨nerbein, H. Lamecker, and P.-M. Schlag, “Augmenting intraoperative 3D
    ultrasound with preoperative models for navigation in liver surgery,” in Proceedings MICCAI, Lecture Notes
    in Computer Science 3217, pp. 543–541, 2004.
10. N. Papenberg, T. Lange, J. Modersitzki, P. M. Schlag, and B. Fischer, “Image registration for ct and
    intraoperative ultrasound data of the liver,” in SPIE Medical Imaging, in press, 2008.
11. T. Strothotte and S. Schlechtweg, Non-Photorealistic Computer Graphics. Modelling, Rendering, and Ani-
    mation. San Francisco: Morgan Kaufmann, 2002.
12. B. Gooch and A. Gooch, Non-Photorealistic Rendering. Natick, MA, USA: AK Peters, Ltd., 2001.
13. V. Interrante, H. Fuchs, and S. M. Pizer, “Conveying the 3d shape of smoothly curving transparent surfaces
    via texture,” IEEE Transactions on Visualization and Computer Graphics, vol. 3, no. 2, pp. 98–117, 1997.
14. J. Fischer, D. Bartz, and W. Straßer, “Illustrative Display of Hidden Iso-Surface Structures,” in Proceedings
    of IEEE Visualization, pp. 663–670, 2005.
15. C. Tietjen, T. Isenberg, and B. Preim, “Combining Silhouettes, Surface,and Volume Rendering for Surgery
    Education and Planning,” in IEEE/Eurographics Symposium on Visualization (EuroVis), pp. 303–310, 2005.
                                                      o
16. S. Bruckner, S. Grimm, A. Kanitsar, and E. Gr¨ller, “Illustrative context-preserving exploration of volume
    data,” IEEE Transactions on Visualization and Computer Graphics, vol. 12, no. 6, pp. 1559–1569, 2006.
          u
17. J. Kr¨ger, J. Schneider, and R. Westermann, “Clearview: An interactive context preserving hotspot visual-
    ization technique.,” IEEE Transactions on Visualization and Computer Graphics, vol. 12, no. 5, pp. 941–948,
    2006.
18. B. Preim and D. Bartz, Visualization in Medicine. Theory, Algorithms, and Applications. Morgan Kauf-
    mann, 2007.
                                                             o
19. M. Burns, M. Haidacher, W. Wein, I. Viola, and E. Gr¨ller, “Feature emphasis and contextual cutaways for
    multimodal medical visualization.,” in EuroVis, pp. 275–282, 2007.
20. M. Hindennach, S. Zidowitz, A. Schenk, H. Bourquain, and H.-O. Peitgen, “Computer assistance for fast
    extraction and analysis of intrahepatic vasculature from contrast-enhanced ct-volume data for preoperative
    planning in liver surgery.,” in International Journal of CARS, vol. 2, pp. 451–452, 2006.
21. D. Selle, B. Preim, A. Schenk, and H.-O. Peitgen, “Analysis of vasculature for liver surgical planning,” in
    IEEE Transactions on Medical Imaging, vol. 21, pp. 1344–1357, 2002.
22. S. Oeltze and B. Preim, “Visualization of vasculature with convolution surfaces: method, validation and
    evaluation.,” IEEE Transactions on Medical Imaging, vol. 24, no. 4, pp. 540–548, 2005.
23. A. Schenk, G. Prause, and H.-O. Peitgen, “Local cost computation for efficient segmentation of 3d objects
    with live wire,” in SPIE Medical Imaging:Image Processing, vol. 4322, pp. 1357–1364, 2001.
24. C. Everitt, “Interactive order-independent transparency,” tech. rep., NVIDIA Corporation, 2001.
25. C. Sigg and M. Hadwiger, “Fast third-order texture filtering,” in GPU Gems II, Addison Wesley, pp. 313–
    329, 2005.
26. F. Ritter, C. Hansen, V. Dicken, O. Konrad, B. Preim, and H.-O. Peitgen, “Real-time illustration of vascular
    structures,” IEEE Transactions on Visualization and Computer Graphics, vol. 12, no. 5, pp. 877–884, 2006.
         u
27. K. M¨hler, M. Neugebauer, C. Tietjen, and B. Preim, “Viewpoint Selection for Intervention Planning,” in
    IEEE/Eurographics Symposium on Visualization (EuroVis), pp. 267–274, 2007.

								
To top