image analysis, and given exemplar textures from medical textbooks and other medical illustrations, the MGTS framework is fully automatic and does not require any user interaction during either model-based part or texture synthesis part. There are several parameters in MGTS that can be set by the user, such as the thresholds for transition regions and the size of the neighborhoods.
Figure 6.20: Solid textured models (the parotid, masseter, scm, mandible, thyroid) integrated into the model-guided rendering framework
As indicated in the introduction of this chapter, the goal of the MGTS framework is eventually to be able to fill a space containing many organs with synthesized solid textures to create an illustrated look for a particular patient. In Figure 6.20 solid textured models (the parotid, masseter, scm, mandible, thyroid) are visualized inside the CT data using the Model Guided Rendering (MGR) framework (Merck (2009)) to identify these organs.
Figure 6.21: (Top row) solid textured models are visualized inside volume rendering. (Bottom row) a clipping plane is applied to the 3D data. The left image illustrates that without textures it is hard to distinguish the anatomical organs and the regions inside them. The right image shows that with solid textures mapped onto the 2D slices, the organs can be easily identified.
distinguish the organs inside the CT data, the textures mapped onto a 2D slice help the user to distinguish them in a 2D slice view of CT data. In Figure 6.21 solid textures are mapped onto a 2D slice to visualize and distinguish the organs on that slice.
Figure 6.22 demonstrates the usage of texture in radiation dose visualization. Here the textures help identification of organs whose surfaces are washed with dose colors. For the applications in which color is used to encode a separate data set, like dose in radiation treatment planning, texture can be used to identify the models and their shapes.
Figure 6.22: (Left) Solid textured models under dose color wash in 3D. (Right) Dose is mapped onto the textured model.
Figure 6.23: Solid textured lung model. The texture is synthesized for illustrating the emphysema of the lungs assuming that it has this disease. Sample illustration is available in “http://www.virtua.org/”
an organ since image intensity information inside the organ should be investigated for determining these regions. However, the framework has capability of synthesizing tex- tures that express a pathology (e.g., sickness) in the organ. In Figure 6.23 the textbook illustration on the top shows the emphysema of the lungs for a typical patient. A 2D exemplar texture is cropped from this illustration and used for synthesizing lung texture for a real patient’s lung model. The synthesized textures on and inside the model are shown in the two images below the illustration.
The MGTS framework can also be used for synthesizing complex objects that consists of many substructures, such as the kidney. Figure 6.24 illustrates some of the textured structures inside a patient’s kidney.
Figure 6.24: Solid textures are synthesized for the layers of kidney and for the pyramids inside the kidney.
et al. (2009)) because it can generate consistent and detailed progressively changing solid textures from 2D exemplar textures. Different solid textures can be produced for different models using the proposed framework, and they can easily be integrated with the CT image data. In addition, the illustrations can be used to highlight the models inside the CT data for different application. The drawback of the framework, however, is the cost in both computation and memory, as it explicitly computes and stores a dense 3D array of voxels covering the entire target model. In addition, the guidance information is based on the model, not the image-intensity pattern. In the future these drawbacks can be dealt with, and new approaches can be added to the system.
Chapter 7
Conclusion
Traditional medical illustration has focused on presenting information in an effective, efficient and attractive way to understand the anatomy of a typical patient. Its main goals are to record and disseminate medical knowledge by reducing visual overload. Computer-supported medical illustrations use different visual elements, such as shading, cutting, deformation, and annotation to achieve that goal.
While in scientific visualization the objective is to provide detailed, objective infor- mation by mapping data to color value, in illustrative visualization the goal is to provide aggregated or abstracted information; sometimes subjective (e.g., personal ideas / the- ories) by mapping data to colors, textures, symbols, styles, patterns, sketches, etc. The focus of this dissertation is to provide one of these means, texture, to obtain illustrative visualizations of anatomic objects. The main goal of the model-guided texture synthesis (MGTS) framework is to maintain the global comprehension from traditional medical il- lustrations while visualizing the shape of the anatomical structures of individual patient data. There are methods that try to create illustrative visualization of patient data, such as methods in Lu & Ebert (2005) which use a color-transfer function for isotropic texture cubes. However, none of these methods consider the variation of materials inside the anatomical models.
ance information obtained from model-specific parameterization and region-specific tex- tures in order to synthesize region-specific anisotropic textures. Texture generation requires only limited manual intervention, which consists of setting some parameters. MGTS is designed to be based on deformable model-based segmentation technologies, which have been under development at UNC. It can be applied to a broad variety of patient-specific, segmented anatomical models. One can interact with it (e.g., using model-based clipping planes to see the interior regions of a textured organ) in a 3D visualization environment, for example in the Model Guided Rendering framework pro- posed in Merck (2009).
This chapter has four main sections. First, there is a review of the dissertation’s thesis statement and claims in the context of the methods and materials presented in the previous chapters. Second, there is a discussion of some of the problems of the methods and framework presented along with some limitations of my study. Third, there are suggestions of directions for future research and for method improvements. Fourth, there are implementation details about each component of MGTS.