Reliability Comparison of Accelerated High Angular Resolution Diffusion Imaging for White Matter Tract

Abstract

Purpose: Accelerated high angular resolution diffusion imaging (HARDI) deals with distortion problem from echo planar imaging (EPI) and longer scan time from unaccelerated HARDI However whether the reconstructed data can keep the information of white matter tract is an open question This study is to compare the similarity of the quantified tract features between the fully sampled and accelerated datasets Method: In this article single-shot and multi-shot EPI are used for data acquisition In k-space acceleration the two-fold and four-fold datasets underwent the SENSE and fast HARDI reconstruction respectively The acceleration in d-space contains half a third and a quarter of gradient directions within the four-fold acceleration To assess the fidelity of the accelerated data it will be compared with the fully sampled dataset including 128 gradient directions In this way two methods were prepared for comparison: track-volume based and tractographic based comparison Track-volume coverage and diffusion indices involve in track-based comparison The diffusion indices contain the fractional anisotropy (FA) the mean diffusivity (MD) and the generalized fractional anisotropy (GFA) Tractography based comparison different from above measures the morphology distance between streamlines directly after paired matching The white matter tracts for measurement are the corpus callosum (CC) the inferior fronto-occipital fasciculus (IFO) and uncinate fasciculus (UNC) The box-plot makes the results present at a glance for comparison of different subjects accelerations and fibers Result: In track-volume coverage the percentage of true positive was about 80~90% 40~80% and 50~80% in the CC IFO and UNC respectively Although the percentage was lower in the IFO and UNC the region of true positive occurred at backbone of the tractography The correlation coefficient value of diffusion indices was at least up to 0 7 in the four-fold acceleration but was down to 0 4 because of different distortion level in the two-fold acceleration The morphology distance was only 1 to 2 voxels per point in the center of a streamline approximately with little difference in inter-subject and intra-subject Conclusion: On the balance of image and tractography quality reserving a quarter of ky-line and a third of diffusion direction is acceptable at most in this reconstruction In clinical the scan time can be reduced to one-third compared with commonly used the two-fold SENSE

Date of Award	2016 Dec 14
Original language	English
Supervisor	Tzu-Cheng Chao (Supervisor)