Bastos Leite, Antonio wrote:
> Let me put thinks clearer, I calculated the average intensity of ASL
> images using large ROIs on DICOM files. I expected the corresponding
> intensity on the 16-bit Analyze files would be: DICOM ROI average
> intensity + 2048 (offset), assuming a scaling factor = 1, but what I
> found was approximately half of that... Therefore, I'm starting to
> suspect (even before trying the floats to confirm...) the scaling factor
> used during conversion could have been 0,5. Does this make any sense?
Hi Antonio,
DICOM can specify slope-intercept rescaling. In the post below, I
wrote some nonsense about window-width/centre, which I now believe is
only used for display purposes:
http://www.jiscmail.ac.uk/cgi-bin/webadmin?A2=ind0609&L=SPM&P=R15963
but I've since found out that DICOMs can (but don't have to -- my T1
structurals didn't, but I believe CT/PET/DTI should) have the
following fields:
Image Rescale Intercept (0028,1052)
Image Rescale Slope (0028,1053)
so the actual intensity is (the stored integer)*slope+intercept.
If the program you are using to convert DICOM->Analyze is
ignoring/mishandling these, then you will probably get garbage results.
It looks to me like SPM5's spm_dicom_convert *does* handle these
fields, so if you use that, then the results should be correct.
Best,
Ged.
|