`a`
Inverse Problems and Imaging (IPI)
 

Variational denoising of diffusion weighted MRI

Pages: 625 - 648, Volume 3, Issue 4, November 2009      doi:10.3934/ipi.2009.3.625

 
       Abstract        Full Text (1595.2K)       Related Articles       

Tim McGraw - West Virginia University, Morgantown, WV 26506, United States (email)
Baba Vemuri - University of Florida, Gainesville, FL 32601, United States (email)
Evren Özarslan - National Institutes of Health, Bethesda, MD 20892, United States (email)
Yunmei Chen - University of Florida, Gainesville, FL 32601, United States (email)
Thomas Mareci - University of Florida, Gainesville, FL 32601, United States (email)

Abstract: In this paper, we present a novel variational formulation for restoring high angular resolution diffusion imaging (HARDI) data. The restoration formulation involves smoothing signal measurements over the spherical domain and across the 3D image lattice. The regularization across the lattice is achieved using a total variation (TV) norm based scheme, while the finite element method (FEM) was employed to smooth the data on the sphere at each lattice point using first and second order smoothness constraints. Examples are presented to show the performance of the HARDI data restoration scheme and its effect on fiber direction computation on synthetic data, as well as on real data sets collected from excised rat brain and spinal cord.

Keywords:  Diffusion MRI, denoising.
Mathematics Subject Classification:  Primary: 92C55; Secondary: 62H35.

Received: October 2008;      Revised: August 2009;      Available Online: October 2009.