niftiread(): rescaling phase in Matlab to [-pi,pi]

[rtopfer]

A function is needed to shift+rescale phase images and recast to double from int16,
i.e. to perform what this bit of code in MaRdI does for DICOM's

• info = niftiinfo( phase.nii ) ; : the info struct generally contains .MultiplicativeScaling and .AdditiveOffset fields which are probably needed

• question: Will the conversion formula be consistent across sites/sequences? The last paragraph in Processing Steps - Step 2 on the FSL/FUGUE page suggests it would not; in which case, we might want a site-specific configuration file for field mapping.

• btw: @evaalonsoortiz i noticed acquisitions 09-10 from ACDC108p (your custom sequence) don't have the normal rescaling info in the phase dicom header (so, the converted nifti's don't have it either). I'm not sure if there is a way to consistently determine what the actual phase (in radians) is without that. i.e. if there are wraps, a priori: 0=-pi, and 4095=+pi; but if there aren't any wraps present (e.g. max raw phase = 2000) then i think it might be impossible to determine the correct range.

[evaalonsoortiz]

@rtopfer Hmm ... strange that's missing. What is the header name? Are you sure the range varies per sequence? I thought it only varied according to the software version and/or vendor.

[rtopfer]

RescaleIntercept and RescaleSlope—tags: (0028,1052) and (0028,1053).

No idea whether it actually varies between sequences: I think it's been consistent across the few Siemens GRE sequences we've used but the disclaimer on the FSL/FUGUE page is disconcerting since obviously we want an automated solution for the rescaling and ideally that wouldn't require a massive if/else tree...