Volumetric Visualization of MRI data

Visualization of data is helpful in understanding the data faster and in a more intuitive way; it provides the viewer with an insight to the data in hand. Visualization is classified based on the data characteristics we intend to visualize. The two visualization classifications are: scientific visualization, which focuses mostly on data that has inherent spatial reference, and information visualization, which deals with data with no inherent spatial reference. While the former focuses on realistic renderings of volumes, surfaces and illumination sources; The latter deals with large-scale data which contains mostly non-numerical information such as text.;Scientific medical data such as Computed Tomography or Magnetic Resonance Imaging contains inherent spatial reference data which is embedded in a three-dimensional space. Each point in three-dimensional space represents one or more properties of the measurements or simulations performed at its spatial position. In this thesis, since MRI data is the focus of study and the aim is to have a realistic volume rendering, the scientific visualization technique will be employed.;In the Brian application, the goal is to visualize only the relevant information for the user. Since not all parts of the data are necessary to fulfill a certain task, all the unnecessary data are masked out in the visualization. The goal is to present a method by which unnecessary data can be masked out and visualize data intended for display.;Methods proposed in this thesis, improve the visualization in the classification step. This allows the user to visualize different materials, detect isosurfaces and visualize objects beneath transparent surfaces. With this method, it will be easier to distinguish between different objects and materials in the data and enhance structures, which represent object surfaces. It is worth mentioning that the advantage of selecting the scientific visualization technique over the information visualization is that, it can easily deal with occlusions. It is possible that some parts occlude other areas of the data, hence scientific visualization technique is utilized to avoid the occlusion complication.;A novel approach has been proposed to enhance features of the data in the classification process, which are important for detecting objects beneath surfaces in the human brain. With the proposed method for designing the transfer function, it is possible to select regions of interest or in other words extract properties from the data, which enable the user to classify different materials with a better quality compared to previous proposed methods. In the proposed method, we try to use mean value or local neighboring of each point in the data set to better distinguish different features in the brain. The proposed approach employs transfer function spaces based on statistical properties, data values and gradient magnitudes. The transfer function developed in the mentioned transfer function spaces allows the user to easily distinguish and detect different materials and borders based on the application in hand. This thesis is structured as following which an introduction to the scientific visualization is given, steps of visualization are explained in the visualization pipeline; the classification and proposed methods for improving it are discussed in the proceeding chapters.