Self-oriented Diffusion Basis Functions for white matter structure estimation

We present an extension to the Diffusion Basis Function Model for fitting the in vivo brain axonal orientations from Diffusion Weighted Magnetic Resonance Images. The standard Diffusion Basis Functions method assumes that the observed Magnetic Resonance signal at each voxel is a linear combination of a static set of basis functions with equally distributed orientations into the 3D unitary sphere. Our proposal, overcomes the limited angular resolution of the original model by adapting the basis orientations using a sophisticated non-linear optimization procedure. The improvements over the standard Diffusion Basis Functions model estimation by our proposal are demonstrated on the synthetic data-sets used on the 2012 HARDI Reconstruction Challenge.