Optimisation in strategies in diffusion tensor MR imaging

Abstract

With diffusion MR imaging, the Brownian motion (or self-diffusion) of water molecules is measured. In the corresponding ADC (Apparent Diffusion Coefficient) image, each image element (voxel) represents the average diffusion. In MRI, the diffusion is measured with diffusion gradients along certain directions in space. The diffusion-weighting gradients are often incorporated in an echo planar imaging (EPI) pulse sequence. Depending on the type of tissue being imaged, the measured diffusion may be isotropic, i.e. equal for an directions of the diffusion gradient, as Seen, e.g., in grey matter and the cerebrospinal fluid (CSF).

In contrast, in white matter, the diffusion is higher along the nerve fibres than across because water molecules moving along fibres are not hindered - this is referred to as anisotropic diffusion. In diffusion tensor imaging (DTI), the diffusion, for each voxel, is represented by an ellipsoid with a certain shape and orientation. For e.g. grey matter the ellipsoid is a sphere, whereas for white matter it is elongated in one direction like a cigar. From the diffusion tensor the mean diffusion (size of the ellipsoid) and the degree of anisotropy (shape of the ellipsoid) may be calculated.

In this thesis different strategies to improve the image quality and reliability of the DTI data and diffusion anisotropy maps are presented. These include simulation studies to determine 1) which anisotropy index is most insensitive to noise and 2) which diffusion scheme (i.e. which set of diffusion gradients to use in the scanning process) minimises the variance and bias of the calculated DTI data. The simulation results were also complemented with data from phantoms and volunteers. Additionally, two major types of image artefacts in diffusion weighted single-shot echo planar imaging (DW SS-EPI) and means of correcting them have been investigated. The first artefact is signal dropout, predominantly in the mid-lower part of the brain, due to brain motion.

The second is eddy currents induced by the diffusion gradients that cause the DW images to be distorted differently depending on the direction of the diffusion gradient. The distortions are translation, scaling and shear effects in the phase encoding direction of the image. These distortions, together with patient motion, result in anatomical mismatch between the different DW images used for the calculation of the diffusion tensor data. A new distortion correction method that corrects for this mismatch has been developed. DTI has been performed in a study comparing schizophrenic patients and normal controls with respect to diffusion anisotropy, mean diffusion and morphological differences.