کنفرانس مهندسی برق ایران

صفحه اصلی / سی و دومین کنفرانس بین المللی مهندسی برق

Denoising of the Diffusion Tensor Imaging Data Using k-Space Redundancy

نویسندگان :

Khashayar Esmaeilzadeh¹ Farzaneh Keyvanfard² Abbas Nasiraei Moghaddam³

1- دانشگاه صنعتی امیرکبیر 2- دانشگاه خواجه نصیرالدین طوسی 3- دانشگاه صنعتی امیرکبیر

کلمات کلیدی :

diffusion tensor imaging،denoising filter،spatial frequency domain،fractional anisotropy،cartesian filling

چکیده :

Diffusion tensor imaging is a highly capable, yet noise-sensitive method for obtaining the brain's white matter structure. It comprises repeated acquisitions, each weighted by a diffusion gradient in one specific direction. The acquired data, therefore, has a large amount of redundancy that can be used for its efficient denoising. In this paper, we estimated the pattern for areas in the k-space most affected by the directional diffusion gradient. This pattern was then used for filtrating the acquired data for each direction. In particular, the areas minimally affected by diffusion gradients were replaced by the k-space data that was averaged over all directional acquisitions. The central idea for this filtering is the geometrical constraints on the diffusion caused by fiber orientations. This denoising approach was examined through the resulting reconstructed diffusion images as well as consequent fractional anisotropy (FA) maps, in terms of Signal-to-Noise Ratio and Contrast-to-Noise Ratio (CNR) obtained for different subsampling levels. The results demonstrated a 25.58% increase in SNR and a 27.86% increase in CNR for the FA map when our filter was applied during subsampling (44.44% of each k-space) in the datasets. Additionally, the qualitative assessment showed our filter resulted in a better representation of tracts in the FA map derived from the fully sampled k-spaces.