I habe a question about analysing ADC-maps with VBM.First of all
i created a template with t1-images and then i wanted to process
the ADC maps from the same subjects with following methode:
1. ADC Dicom download
2. Dicoms with DICOM-import to SPM
3. Normalisation:
• Source image: template T1
• Image to write: ADC- maps from DICOM import
• Template image: Epi Template
4. Coregistration
• Reference image: coregistered T1 image
rs....
• Image to reslice : results from normalisation
5. Smoothing: 8mm smoothen
Is my way to analyse them correct?
No.
The procedure you describe can not really be defined as "morphometry",
which is concerned with the study of form or shape - as opposed to
composition or function
( http://medical-dictionary.thefreedictionary.com/morphometry ). It is
therefore more appropriate to refer to it as statistical parametric
mapping (SPM) or voxel-based analysis (VBA).
I'll probably get in trouble for saying this, but in my opinion, many of
the differences among such spatially normalised images are often best
explained by systematic registration error. Depending on the model of
the data used for pre-processing, some of the findings may offer some
useful insights, but in many cases, these will be difficult to
disambiguate from mis-registration.
Your spatial normalisation steps seem a bit dubious. Spatial
normalisation (via the Normalise button) is based on minimising the sum
of squares difference between one of your images and a template image.
The idea is to figure out how to warp the brain of an individual subject
so that it aligns to MNI space.
For this to work, the difference between the template and image that it
is matched with (ie the "source image") should approximately conform to
a field of Gaussian noise. This will not be the case if you try to
minimise the difference between the EPI and T1w template images, and is
very unlikely to give you any kind of estimate of the relative shape of
your individual subjects' brains.
Best regards,
-John
--
John Ashburner <[log in to unmask]>
|