Homogeneity based segmentation and enhancement of Diffusion Tensor Images - A white matter processing framework

High-quality version: rendezvous.avi (20 MB)
An artistic video created with my colleagues Vesna Prckovska and Tim Peeters for the NSF Visualization challenge 2009. Enjoy it and don't forget to put the speakers on!

High angular resolution diffusion imaging (HARDI) has proven to better characterize complex intra-voxel structures compared to its predecessor diffusion tensor imaging (DTI). However, the benefits from the modest acquisition costs and significantly higher signal-to-noise ratios (SNRs) of DTI make it more attractive for use in clinical research. In this work we use contextual information derived from DTI data, to obtain similar crossing information as from HARDI data. We conduct a synthetic phantom study under different angles of crossing and different SNRs. We compare the extrapolated crossings from contextual information with HARDI data. We qualitatively corroborate our findings from the phantom study to real human data. We show that with extrapolation of the contextual information, the obtained crossings are similar to the ones from the HARDI data, and the robustness to noise is significantly better.

HARDI is a MRI imaging technique that is able to better capture the intra-voxel diffusion pattern compared to its simpler predecessor DTI. However, HARDI in general produces very noisy diffusion patterns due to the low SNR from the scanners at high b-values. Furthermore, it still exhibits limitations in areas where the diffusion pattern is asymmetrical (bifurcations, splaying fibers, etc.). To overcome these limitations, enhancement and denoising of the data based on context information is a crucial step. In order to achieve it, convolutions are performed in the coupled spatial and angular domain. Therefore the kernels applied become also HARDI data. However, these approaches have high computational complexity of an already complex HARDI data processing. In this work, we present an accelerated framework for HARDI data regularizaton and enhancement. The convolution operators are optimized by: pre-calculating the kernels, analysing kernels shape and utilizing look-up-tables. We provide an increase of speed, compared to previous brute force approaches of simpler kernels. These methods can be used as a preprocessing for tractography and lead to new ways for investigation of brain white matter.

The analysis and visualisation of Diffusion Tensor Images (DTI) is still a challenge since it is multi-valued and exploratory in na- ture: tensors, fiber tracts, bundles. This quickly leads to clutter problems in visualisation but also in analysis. In this paper, a new framework for the multi-resolution analysis of DTI is proposed. Based on fast and greedy watersheds operating on a multi-scale representation of a DTI image, a hierarchical depiction of a DTI image is determined conveying a global-to-local view of the fibrous structure of the analysed tissue. The multi-resolution watershed transform provides a coarse to fine partitioning of the data based on the (in)homogeneity of the gradient field. With a transversal cross scale linking of the basins (regions), a hierarchical representation is established. This framework besides providing a novel hierarchical way to analyse DTI data, allows a simple and interactive segmentation tool where dif- ferent bundles can be segmented at different resolutions. We present preliminary experimental results supporting the validity of the proposed method.

In segmentation techniques for Diffusion Tensor Imaging (DTI) data, the similarity of diffusion tensors must be assessed for partitioning data into regions which are homogeneous in terms of tensor characteristics. Various distance measures have been proposed in literature for analysing the similarity of diffusion tensors (DTs), but selecting a measure suitable for the task at hand is difficult and often done by trialand- error. We propose a novel approach to semiautomatically define the similarity measure or combination of measures that better suit the data. We use a linear combination of known distance measures, jointly capturing multiple aspects of tensor characteristics, for comparing DTs with the purpose of image segmentation. The parameters of our adaptive distance measure are tuned for each individual segmentation task on the basis of user-selected ROIs using the concept of Kernel Target Alignment. Experimental results support the validity of the proposed method.

Many different measures have been proposed to compute similarities and distances between diffusion tensors (DTs). These measures are commonly used for algorithms such as segmentation, registration and quantitative analysis of Diffusion Tensor Imaging (DTI) data sets. The results obtained from these algorithms are extremely dependent on the chosen measure. The measures presented in literature can be of complete different nature, and it is often difficult to predict the behavior of a given measure for a specific application. In this chapter, we classify and summarize the different measures that have been presented in literature. We also present a framework to analyze and compare the behavior of the measures according to several selected properties. We expect that this framework will help in the selection of a measure for a given application and to identify when the generation of a new measure is needed. This framework will also allow the comparison of new measures with existing ones.