Vesna Prkovska1, Anna Vilanova1,
Cyril Poupon2, Bart ter Haar Romeny1, Maxime Descoteaux3
1Biomedical Engineering, Eindhoven
University of Technology, Eindhoven, Netherlands; 2NeuroSpin, CEA
Saclay, Gif-sur-Yvette, France; 3Computer Science, Universit de
Sherbrooke, Qubec, Canada
This
work presents a HARDI study of the classification power of different
anisotropy measures. This classification aims towards separating the data
into three compartments: Isotropic,
Gaussian and Non-Gaussian. Afterwards the data can be simplified in the first
two compartments by simpler diffusion models.
To quantify the classification power of the measures, ex-vivo phantom
data is used, and the findings are qualitatively illustrated on real data
under different b-values and gradient sampling schemes. The benefits from the
data simplification are clinically attractive due to the possibility of
significantly decreasing the post-processing time of the HARDI models and
faster, more intuitive visualization.