Promotie C. Zhang: tensor fields

12 december 2017 12:30 - Locatie: Aula, TU Delft - Door: webredactie

Comparative and Ensemble Visualization of Diffusion Tensor Fields. Promotor: Prof.dr. E. Eisemann (EWI).

Scientific visualization of tensor fields is challenging due to the complex and multivariate nature of tensor data. The visualization of multiple tensor fields becomes even more difficult, and still in its infancy. This thesis focuses specifically on the visual analysis of multiple Diffusion Tensor Imaging (DTI) datasets.

DTI is able to measure the diffusion profile of water molecules within each voxel, which is influenced by the underlying fibrous structure of white matter, and models it as the so-called diffusion tensor. The diffusion tensor, mathematically expressed as a 3-by-3 symmetric positive-definite matrix, can be decomposed into scale, shape, and orientation, which are its intrinsic properties. Each of them has a biologically meaningful interpretation for the underlying tissue properties. It is the intrinsic properties of the diffusion tensor that make DTI a uniquely important imaging modality. This thesis makes heavy use of tensor intrinsic properties because neuroscientists can put direct interpretations on them.

To compare two tensor fields in a voxel-wise manner, this thesis proposes a computationally efficient dissimilarity measure to quantify the pair-wise differences between diffusion tensors in terms of tensor intrinsic properties, and a novel checkerboard-style glyph.

To analyze an ensemble of tensor fields, this thesis proposes a representative mean tensor and tensor ensemble variations based on tensor intrinsic properties. An overview + detail visual analysis framework is developed to facilitate the visual exploration of ensembles of tensor fields in the 3D physical and feature space.

In cases where two ensembles of tensor fields are compared, the contradiction between the huge amount of information to be visualized and the limited number of available visual channels becomes much more severe. This thesis resolves this contradiction by carefully combining and extending the checkerboard-style glyph design and the overview + detail framework. This thesis integrates the level-of-detail concept into the glyph representation, which is able to progressively reveal more information as neuroscientists zoom in.

Meer informatie?

Voor inzage in proefschriften van de promovendi kunt u kijken in de TU Delft Repository, de digitale vindplaats van openbare publicaties van de TU Delft. Proefschriften zullen binnen een paar weken na de desbetreffende promotie in de Repository te vinden zijn.