Rt/issue75 shim specs json #92
Conversation
…rom ShimSpecs_Greg.m)
+img/read_nii.m
Outdated
| @@ -0,0 +1,37 @@ | |||
| function [img,info,json] = read_nii( niiFile ) | |||
| %READ_NII Load NIfTI image and header; and, when present, the accompanying .json sidecar | |||
There was a problem hiding this comment.
not sure if it's of any help, but for the records here is how SPM deals with BIDS: https://en.wikibooks.org/wiki/SPM/BIDS
| [ Field, Params, Specs] = ShimOpt.parseinput( varargin ) ; | ||
|
|
||
| if isempty(Specs) | ||
| Shim.System.Specs = ShimSpecs_Greg() ; |
There was a problem hiding this comment.
didn't we agree to use camelCase for variables? so having something like this:
| Shim.System.Specs = ShimSpecs_Greg() ; | |
| shim.System.Specs = ShimSpecs_Greg() ; |
or maybe that would break the code somewhere else and we should do another specific PR for changing formatting?
There was a problem hiding this comment.
baby steps!
We'll get rid of ShimOpt_Greg.m in a separate PR and change the case in the remaining classes.
|
For SMS acquisitions, would it make sense to explicitly store the slice ordering and slice group information in the .json file? This information is in the twix meas.dat header, but I think it's also buried in the DICOM header, though it might not be parsed by normal DICOM header header scripts in Matlab. |
Excellent idea. @rtopfer or @po09i can you look if already done by dcm2niix and if not, open an issue on their repos and document here via new issue. Thx |
rtopfer
force-pushed
the
rt/issue75-ShimSpecs-json
branch
from
5e3a016
to
19f8527
Compare
…vC coils to make them almost identical — 3 new classdefs are not required for the amplifier board upgrade
… original ShimSpecs at this point and *config* should be more clear)
| @@ -0,0 +1,8 @@ | |||
| %b0shim.coil.greg 8-channel AC/DC 3T neck coil | |||
There was a problem hiding this comment.
to look nice on matlab's doc:
| %b0shim.coil.greg 8-channel AC/DC 3T neck coil | |
| %% b0shim.coil.greg | |
| % 8-channel AC/DC 3T neck coil |
i suggest applying this format to the other Contents.m files. Also see shimming-toolbox/shimming-toolbox#134 (comment)
There was a problem hiding this comment.
Not sure if this is the most appropriate thread to post this, but I found a very fast way to interpolate the shim coil profile maps into the grid used by the target volume (for example the EPI volume) to prepare the coil field maps for shim optimization calculations. I'm not sure how Ryan handled this in the original code.
unix('flirt -in coil_field_maps.nii -ref epi_target.nii -applyxfm -usesqform -out coil_field_maps_interp')
This seems to be ~100x faster than my previous approach which used scatteredInterpolant in Matlab. Matlab is extremely slow if you use "scattered" coordinates for the interpolation rather than monotonically increasing points such as those generated by meshgrid. The trick is to use the -usesqform flag in FLIRT, so that it will just apply a coordinate transform/interpolation without rotating or moving the image to register them.
However, I'm not sure if we want to commit to relying on FSL in the GUI. We might want to implement our own fast interpolator using the same methods as FLIRT.
There was a problem hiding this comment.
Thank you @jaystock, I moved the discussion to a specific issue thread shimming-toolbox/shimming-toolbox#159
This PR serves to close #75 — replacing
ShimSpecssubclasses with .json files.This is an initial step toward refactoring the way shim systems are currently represented in the code base: Once completed, a minor refactor of the currently abstract
ShimOptclass should allow us to dispense with several (if not all) of theShimOptchild classes, thereby simplifying the code base and making it very easy to define new shim systems as only the calibration maps and .json file should then be required!