[Insight-users] Urgent: Strange dicom imageIO

[Mathieu Malaterre]

Miller, James V (GE, Research) wrote:
> Mathieu,
> 
> Defaulting slope/intersept to 1/0 will result in a loss of precision of PET data.
> The slopes in PET are established on a per slice basis to maximize the precision 
> in the 16bits stored on disk.  So if we read in a PET series and write it out with
> a slope of 1/0 we'll loose precision. on some of the slices.
> 
> Just something we should think about.

Ok I see. Well then we need at least some more documentation on that. 
Some users might find it very confusing that this ImageIO is indeed very 
different from the other. The worst part is that some modality support 
32bits precision but not others :)

BTW, how about supporting the slope/intercept in the ImageIO 
(encapsulate the ImageShiftScale).

Mathieu