If you are referring to a binary mask, I don't think there is any
consensus in thresholding PVE at particular probability. You can hard
segment (1 or 0) a voxel into one of the three compartments (GM, WM, and
CSF), and I often use either 0.5 or maximum belonging probability as the
In VBM studies, people often try to avoid partial volume effect by
thresholding unmodulated images at 0.8~0.95.
Hope this helps,
Mazerolle, Erin wrote:
> I'm wondering if people usually threshold the PVE maps before using
> them as a mask (e.g., a gray matter mask) and if so, how is the
> threshold selected?