| By the way, what losses will be incurred by entering a time series into the
| segmentation?
The segmentation algorithm describes the tissue classes as multivariate normal
distributions, which requires a mean vector of length n and a covariance matrix
of size n*n, where n is the number of images in the time series. With 84
images, there are 3654 unknowns (3570 for the symmetric covariance matrix and 84
for the mean) to compute in order to describe each of the distributions, which
makes processing very slow and unstable.
In order to identify grey matter from an EPI dataset, you are best off using just
a single image. As the algorithm stands, the other images don't provide any
useful information for partitioning into different tissue types.
All the best,
-John
|