port STIR TOF support to SIRF

[KrisThielemans]

we're finalising the STIR TOF PR UCL/STIR#304

This should normally require very few changes to SIRF. I think the only places are where we expose getting the projection data as TOF adds another dimension.

I think this only affects PETAcquisitionData.dimensions, as_array and show. @evgueni-ovtchinnikov could you please confirm?

Ideally, we should prepare for this in 2.0 by already letting these functions return the TOF dimension (although it's of course size 1). That way, we can enable TOF support in 2.1, without breaking backwards compatibility.

Of course, related is what dimensions the data should be to support PET data for the future, in particular dynamic/gated data.

[evgueni-ovtchinnikov]

will SegmentBySinogram<float>::full_iterator work in the same way?

[NikEfth]

Hi, I fixed a bug that I introduced on my last commit. And few other minor. I will push tonight. There is an open issue with some hdf5 stuff that leaked in. @kris, advice on this? Cheers

…

On Sat, 23 Feb 2019, 10:25 Kris Thielemans, ***@***.***> wrote: we're finalising the STIR TOF PR UCL/STIR#304 <UCL/STIR#304> This should normally require very few changes to SIRF. I think the only places are where we expose getting the projection data as TOF adds another dimension. I think this only affects PETAcquisitionData.dimensions, as_array and show. @evgueni-ovtchinnikov <https://github.com/evgueni-ovtchinnikov> could you please confirm? Ideally, we should prepare for this in 2.0 by already letting these functions return the TOF dimension (although it's of course size 1). That way, we can enable TOF support in 2.1, without breaking backwards compatibility. Of course, related is what dimensions the data should be to support PET data for the future, in particular dynamic/gated data. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#315>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AIlA_pBdaiTdSEfJ-oJXF2fccqIN7-fNks5vQRb-gaJpZM4bN3hQ> .

[KrisThielemans]

will SegmentBySinogram<float>::full_iterator work in the same way?

yes, but currently we need another loop over tof-bins (not just views/segments), i.e. our Segment/Sinogram/Viewgram` classes are the same as before (just also track the TOF bin). Very easy, but obviously it cannot be done on SIRF master yet. would need a separate branch for this.

@ALEXJAZZ008008 will help with this.

[KrisThielemans]

@NikEfth let's keep STIR discussions to the STIR PR. But we cannot really remove hdf5-related commits/files, unless we go with git filterbranch, and screw up everybody's repository (which we don't want to do!). It is probably a matter of first merging the non-TOF GE Signa stuff.

[KrisThielemans]

See #359 for preparation

[KrisThielemans]

Probable list of things that need to change

• cstir.cpp cSTIR_getAcquisitionsDimensions needs to get number of TOF bins from STIR
• STIR.py PETAcquisitionData.show() has sinogram selection. Would need TOF bin selection.
• similar for MATLAB (won't do this anymore)
• stir_x.cpp: estimate_randoms will need an extra loop over tof_bins (simply accumulating in the same non-TOF array)
• stir_data_container.cpp
  • PETAcquisitionData::norm() etc needs extra loop over tof bins (done in fixes STIR TOF AcquisitionData algebra #1208)

[KrisThielemans]

another:

• AcquisitionData.rebin() needs an extra argument num_tof_bins_to_combine. At the moment this will default to 1 (i.e. don't change TOF bins). This is in fact the only thing currently supported by STIR anyway.

[KrisThielemans]

randoms still to do