I am currently working on a study of patients who went through
surgery and had removed regions of their brain. SPM8 does a good
job on segmenting and normalizing all the patients but one, who
has a big resection. In this patient segmention is not done
properly and I cannot find out why.SPM does agreat job with
other patients with a resection as big as this one´s.
¿Is there another way of normalizing for particular cases like
this?
I think you've just been lucky with the images of resected brains where
the algorithm has appeared to work. When algorithms fail to do what is
hoped, it is usually difficult to predict where and when this will
occur. An explanation for why it didn't work is simply that the
segmentation model is not that accurate for resected brains - which is
not very satisfactory. Another potential explanation would be that the
algorithm has got caught in a local optima, from which it could not
escape.
In order to make it work reliably on resected brains, the model would
need to be changed. There's a vast literature on image segmentation in
the presence of lesions of one form or another, and some of these
approaches may be suitable for your data. Unfortunately, in most cases,
the descriptions in the papers would need to be coded up.
Best regards,
-John
--
John Ashburner <[log in to unmask]>
|