Skip to content

Conversation

@rijobro
Copy link
Contributor

@rijobro rijobro commented Feb 4, 2020

Still in progress.

@KrisThielemans We discussed calculating the transformation matrix and then converting each SIRF index into a STIR index, along the lines of:

https://github.com/rijobro/SIRF/blob/c8578ba65fb44eb95d324ba40a738d9651878ded/src/xSTIR/cSTIR/stir_data_containers.cpp#L360-L367

I realise that we can't access each element of the SIRFImageData individually, so this method won't work. Could we create a STIRImageData, fill it, and then reorient the image? I suppose my question is if there is reorient functionality in STIR?

@KrisThielemans
Copy link
Member

@evgueni-ovtchinnikov is it guaranteed that SIRF iterators run through data in the same order as https://github.com/rijobro/SIRF/blob/c8578ba65fb44eb95d324ba40a738d9651878ded/src/xSTIR/cSTIR/stir_data_containers.cpp#L362-L364? (at least when it's a 3D volume).

we must have some correspondence between indices and order somewhere. Otherwise the geom-stuff doesn't make any sense.

@evgueni-ovtchinnikov
Copy link
Contributor

@KrisThielemans: SIRF's PET image iterators are wrappers for your DiscretisedDensity<3, float>::full_iterator (and the const one).

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants