Vries, G. de and Wijbenga, S. (2003) 3D Blood-Vessel Analysis and Visualization. Master's Thesis / Essay, Computing Science.
|
Text
Infor_Ma_2003_TWijbenga_GdeVries.CV.pdf - Published Version Download (2MB) | Preview |
Abstract
Innovative scanning technologies such as computed X-ray tomography (CT), magnetic resonance imaging (MRI) or positron emission tomography (PET) empower radiologists to obtain 3D information of the inner human. This information is represented by a set of 2D gray scale images stacked upon each other. Analyzing those image sequences for diagnosis and therapy purposes using 2D image processing systems is a hard and time-consuming task, especially in the case of bloodvessels or other filamentous structures, which are the subject of this report. Not only structures which are not interesting for the observer are present in the image; thin elongated structures (like bloodvessels) which are perpendicular to the viewing direction of those 2D "slices" are hardly visible. This indicates the need for alternative ways of visualizing data with such structures. An example which shows the advantage of 3D imaging (a volume) over 2D imaging (a slice) is given in figure 1.1. Virtual reality applications offer the possibility to visualize the data in a more intuitive way. By looking at a 3D image, physicians can recognize topological coherency in a much faster and more natural way. The benefits of virtual reality has been shown in many applications ranging from architectural design to flight simulations, but there is a difference between these and medical applications. Virtual reality mostly uses (textured) polygons to visualize a virtual environment. Data from CAD systems, like in architectural design, already consists of polygons. Applications that use natural data that is converted to polygon data depend heavily on simplification algorithms to keep polygon count within a reasonable range. By simplifying data, information can be distorted or lost for the sake of reasonable frame rates. This must not be done in medical applications because a diagnosis can depend on these small details. Current advances in constructing high performance computers and implementing 3D algorithms in hardware will overcome these limitations, but even constructing a inefficient polygon set from medical data has some difficulties. Data obtained from MRI, CT or PET consists, as earlier noted, of a cube of scalar values. Only a 'brightness' is known. If one wants to create a triangle mesh (or other graphical primitives) out of it, more information is needed. Enhancement of curvi-linear, dendritic or other filamentous details has many applications in medical image analysis. In this thesis, we study one of those medical applications: the extraction and visualization of blood vessels in 3D angiography datasets. We explain in chapter 2 the concept of filtering, and show that blood vessels can be extracted more easily using filtering. Also a data structure (Maxtree) is described, which is excellently suitable for fast filtering. Chapter 3 is about segmentation of the vessels (dividing the voxels in two classes, vessel voxels and non-vessel voxels). This can be useful for quantitative analysis. Several segmentation methods are described, and a few are implemented. Also, new variants of exsisting strategies are implemented, and the results are discussed. We extended an existing segmentation method which has not been used before in this context (Rapid Automatic Threshold Selection), and developed it further. Those modifications and extensions gave results which were promising enough to lead to a submitted publication for ICIP 2003 (International Conference on Image Processing). In chapter 4 we will give a quick survey on different visualisation techniques and combine direct volume rendering techniques with the Maxtree. We show that with the Maxtree, if combined with the right visualization method, filtering and visualization is possible in a speed at which interactively working is possible. A discussion of the results and conclusions are given in chapter 5. We would like to thank everybody who helped us and inspired us during the research, especially our supervisors Michael Wilkinson and Michel Westenberg.
| Item Type: | Thesis (Master's Thesis / Essay) |
|---|---|
| Degree programme: | Computing Science |
| Thesis type: | Master's Thesis / Essay |
| Language: | English |
| Date Deposited: | 15 Feb 2018 07:30 |
| Last Modified: | 15 Feb 2018 07:30 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/8875 |
Actions (login required)
 |
View Item |