| Diffusion Tensor Imaging (DTI) is the in vivo visualization and analysis of white matter fiber tracts by measuring the anisotropy of water molecule diffusion in the brain tissue. DTI has been increasingly used in clinical imaging. Diffusion weighted images are affected by noise from the human subject and the MRI scanner. This thesis studies the error propagation in the calculation of the DTI invariant anisotropy, mainly the Fractional Anisotropy (FA) using four methods and their comparison in terms of error, filtering and computational efficiency using simulated and human brain data. These methods were Diffusion Tensor, Diffusion Ellipsoid, Hasan and Platonic Variance. The results showed similar trends across the simulated and real data sets. Of the four methods used to calculate FA, the Hasan method without diffusion tensor yielded best computational efficiency, but poor noise robustness, whereas the Platonic Variance method was more robust to noise and provided good computational efficiency. |
