The set of images used for creating the TPM.nii were pretty old and
the contrast between GM and WM in this region was not very high.
Also, much of the GM in the central sulcus has a different composition
from GM elsewhere in the cortex, with an intensity a bit closer to
that of WM. Both these things combine to make the TPM (which was
constructed by segmenting these images) a bit less accurate in this
region, but probably not as inaccurate is in eg thalamus or putamen.
What is needed is an expertly labelled dataset that can be used for
training tissue classifiers, which contains MRI scans and labelled
tissues. Until a dataset like this is available based on some
consensus of which regions the experts consider to be GM, all attempts
define it automatically can be based only on heuristics. Taking an
operational definition, GM is what the algorithms say is GM. Unless a
bit of expert human knowledge is plugged in to help supervise them,
algorithms won't necessarily reproduce what investigators hope they
will.
Best regards,
-John
On 18 November 2010 11:16, Tetiana Dadakova <[log in to unmask]> wrote:
> Dear SPM list,
>
> I am new to SPM and I have a question concerning the TPM.nii template,
> which I use for segmentation.
> When I look at this template, I can see some "artifacts" (pointed with
> red arrows on attached picture).
> Could you please explain me, if this is really an artifact or
> anatomically meaningful region?
>
> Thank you for your time.
> Tetiana.
>
|