Phantoms with attached dimensions

[jonschumacher]

In order to have a physical relation between the size of the phantoms and a possible system matrix simulating the voltages from a the selected phantom, we need to be able to attach dimensions when needed. This can be done by using AxisArrays and Unitful. Every generated phantom should have a constructing function with units which creates the AxisArray version. Dispatch can be performed as described in https://discourse.julialang.org/t/dispatch-on-axisarrays-with-unitful-axis/7733/2. The constructors can maybe be done with https://github.com/jusack/OptionalUnits.jl.

[tknopp]

isn't that this: https://github.com/MagneticParticleImaging/MPIFiles.jl/blob/master/src/RecoData.jl

i.e. we want to have a 1to1 mapping between MDF/ImageData and an in-memory representation.

[jonschumacher]

Maybe I misunderstand you, but this works on a reconstructed image and the phantoms are sort of on the other side. Within the pMPS framework, I would like to be able to scale a phantom such that with a given gradient the correct voltages are simulated.
But ImageMetadata.jl looks interesting. I will have a look into the possibilities it gives.

[tknopp]

A particle concentration image is a particle concentration image, no matter if it is a phantom that is used to synthesize raw data or if it is a reconstructed image.

[jonschumacher]

Hm, true. We can probably copy a lot of that code. Or would you represent a phantom within the in-memory MDF as a reconstruction? I guess if we make use of ImageMeta with the prefix phantom instead of reco we are pretty much compatible with MPIFiles. Or do you have another recommendation?