Open ashgillman opened 4 years ago
ashgillman
commented
4 years ago @KrisThielemans I have listed here as "prior image" rather than "prior" (confusing) or "side information" (is this convention?).
Does it make sense that a PLS prior or a kernel prior would depend on a prior image?
All of these are up for debate and happy to change.
danieldeidda
commented
4 years ago A number of parameters have names that could be improved. With the next major release breaking backwards compatibility, it could be an opportune time to make these changes.
Where possible, old function names should remain (maybe tagged as deprecated for removal in 5.0?) so that other libraries, e.g. SIRF, can update at a more convenient time.
This issue was spun off from a discussion in #422.
Variables to be renamed: (TODO, update this list with PLS and others)
* [ ] HKEM `sigma_m` -> `prior image intensity Gaussian scaling` * [ ] HKEM `sigma_dm` -> `prior image distance Gaussian scaling` * [ ] HKEM `sigma_p` -> `current estimate intensity Gaussian scaling` * [ ] HKEM `sigma_dp` -> `current distance intensity Gaussian scaling` * [ ] HKEM `number of neighbours` -> `neighbourhood diameter` * [ ] HKEM `number of non-zero feature elements` -> `feature vector length`
The major difference between sigma_m/dm and sigma_p/dp is that one is for a constant prior and the other is for the iterative prior. I think iterative should be there to make things generic. In this way it doesn't matter if we use a PET image as "anatomical". I agree though that if someone is approaching the KOSMAPOSL just from reading the paper might be a little confused. However, we can always specify names using comments in the parameter file.
A number of parameters have names that could be improved. With the next major release breaking backwards compatibility, it could be an opportune time to make these changes.
Where possible, old function names should remain (maybe tagged as deprecated for removal in 5.0?) so that other libraries, e.g. SIRF, can update at a more convenient time.
This issue was spun off from a discussion in #422.
Variables to be renamed: (TODO, update this list with PLS and others)
sigma_m->prior image intensity Gaussian scalingsigma_dm->prior image distance Gaussian scalingsigma_p->current estimate intensity Gaussian scalingsigma_dp->current distance intensity Gaussian scalingnumber of neighbours->neighbourhood diameternumber of non-zero feature elements->feature vector length