Our AFQ profiles assume tube-like bundles, which has limitations. We need to explore other options, such as:
- BUAN
- FMQ regression
- k means clustering of nodes from streamlines, based on location + tissue properties
The first suffers from some of the same flaws as AFQ, while the latter two suffer from needing to then correspond the resulting clusters across datasets (likely using some template).