Faculty Host: Dimitris Metaxas

Abstract:

Diffusion Weighted MRI adds to conventional MRI the capability of measuring the mobility of water molecules, referred as diffusion. The properties of water diffusion, which can be characterized by the apparent diffusion coefficient (ADC), are used to probe the underlying tissue microstructures.

To improve the accuracy of the estimation of ADC from noisy High Angular Resolution Diffusion-weighted (HARD) MRI, in this work we present a new variational framework for recovery of ADC profiles. The proposed model approximates the ADC profiles by a 4th order spherical harmonic series (SHS), whose coefficients are obtained by solving a constrained minimization problem. The energy functional are designed so that the ADC profiles can be estimated and regularized simultaneously across the entire volume. The estimation is based on the original Stejskal-Tanner equation, while the regularization is achieved by minimizieg a non-standard growth functional for feature preserving. The antipodal symmetry and positiveness of the ADC are also accommodated in the model. Furthermore, we use the coefficients of the SHS and the variance of the ADC profiles from its mean to characterize the diffusion anisotropy. The experiments on both simulated and HARD MRI human brain data indicates the effectiveness of the method in recovery of the ADC profiles and enhancement of the diffusion anisotropy. The characterization of non-Gaussian diffusion based on the proposed method showed a consistency between our results and known neuroanatomy.

Speaker Bio:

Click here