Figure 7.

FODs estimated with CSD using varying convolution kernels, estimated voxel-wise using the MC-MDI spherical mean model, see the MC-MDI parameter maps for restricted volume fraction (A) and parallel diffusivity (B). In (C), the colored squares highlight three voxels with different tissue compositions; red for a coherent white matter bundle, green for a crossing geometry, and blue for CSF. In (D), we show the corresponding MC-MDI estimated convolution kernels in these voxels for each b-value of the HCP acquisition scheme. Notice that the white matter kernel is more anisotropic than the crossing one, and the CSF kernel is most isotropic and much smaller than the other two. While CSF areas appear more noisy with this voxel-varying kernel approach, it does appear quite effective at resolving crossing structures.