Diffusion Maps Clustering for Magnetic Resonance Q-Ball Imaging Segmentation
Figure 7
Plots of DTI
and ODF affinity matrices of an axial cropped
slice shown in Figure 6.
The matrices are reordered according to
the second (Fiedler) eigenvector. ThePlots of DTI
affinity matrices are shown in decreasing order of , which takes the values , , , and of the quantity
of elements to cluster. In the DTI
case, the decreasing on the scale parameter leads to a
matrix with highly correlated elements
that is very difficult to cluster. In
the ODF case, the block structure is clear and is better suited to apply a
clustering algorithm.
(a) DTI reordered affinity matrices as the scale-space parameter decreases 0
(b) ODF reordered affinity matrices as the scale-space parameter decreases