Anisotropic Acquisition and Analysis for Diffusion Tensor Magnetic Resonance Imaging

Diffusion tensor magnetic resonance imaging (DT-MRI) is a non-invasive imaging method for assessing the characteristics and organization of tissue microstructure. The diffusion tensor provides information about the magnitude, anisotropy, and orientation of water diffusion in biological tissues. In brain white matter, the direction of greatest diffusivity is typically assumed to be parallel to the white matter tracts. The number of DT-MRI applications is rapidly expanding; however, diffusion tensor measurements are also highly sensitive to noise in the raw diffusion weighted (DW) images. Furthermore, the relatively poor spatial resolution of most DT-MRI studies cause partial volume averaging between different tissue regions, which can lead to errors in the estimated DT-MRI measures. Finally, the variance of DT-MRI measures may impair the ability to detect and characterize subtle differences either between regions or subjects. In this thesis, new acquisition and analysis methods for reducing measurement noise effects are investigated. For the case where the diffusion tensor orientation and shape may be estimated a priori, changing the diffusion-weighting with encoding direction may improve the overall accuracy of the diffusion tensor measurements. The variance of DT-MRI measurements is expressed as a function of directional diffusitivities and diffusion weightings. Minimizing the variance using quadratic optimization algorithms leads to an obtainment of anisotropic diffusion weightings. In this study anisotropic diffusion weighting reduced the variance of FA and MD measurements by roughly 50 % in the corpus callosum. Anisotropic Gaussian kernel smoothing was used to reduce the errors and noise for ii the entire regions of DT-MRI data. The anisotropic Gaussian kernels for convolution smoothing are equivalent to the water diffusion distributions described by the diffusion tensor. Further the direction of greatest diffusitivity is often assumed to be parallel to the direction of the local white matter tracts, thus the measured diffusion tensor is a good candidate for anisotropic kernel smoothing. This reduces the partial averaging effects with high levels of smoothing. In voxel based analyses of DT-MRI data, isotropic Gaussian kernel smoothing is often used to blur the individually distinct anatomic features. Anisotropic Gaussian kernel smoothing may reduce the partial volume averaging which will improve anatomic specificity. In this study, anisotropic Gaussian smoothing was applied to DT-MR data from a group of autism subjects to investigate the differences of DT-MRI measurements between the autism and control groups. Anisotropic Gaussian kernel smoothing provides mo re consistent results for the group differences as compared with manual ROI analysis Finally, …