I suspect that this is because the hippocampus is by default labelled as “cortical” grey matter (really more pragmatically described as “non-penetrable grey matter”), meaning that ACT does not permit streamlines to be within that region (whether due to candidate seed points being within that region or due to streamlines projecting toward that region).
If you were to use the
-sgm_amyg_hipp option in
5ttgen, then the hippocampi and amygdalae would be explicitly segmented and labelled as “sub-cortical” (“penetrable”) grey matter, meaning that streamlines can both be seeded within those regions, and when streamlines enter such regions from the white matter, the streamlines can project into those structures, rather than being terminated immediately at the tissue interface.
I’ve never been sure what best to do here: While these two structures would ideally be treated the same way as other non-cortex grey matter nuclei, doing so with the
5ttgen fsl script in particular can lead to some unintended consequences due to the way in which intensity-based and surface-based segmentations are combined.