image-x-institute/mri_distortion_toolkit

Characterisation and reporting of geometric distortion in MRI

mri_distortion_toolkit's Issues

harmonics need to be scaled for gradient strength!!

Ok this is a pretty dumb thing to not realise, but at the moment my harmonics are not accounting for different gradient strengths.

So, I will implement a scaling to 1 mT/m in the calculate Bz step, then in the distortion correction step, we can scale the encoding fields according to the harmonics

Reports needs to scale harmonics according to gradient strength

Figure out a way to have a 'phantom summary' file in e.g. json format, that could be read by the MarkerVolume class

or should phantom code read from the json file?

Fix spherical harmonics analysis docs

add some B0 reporting to the reports module

If we have B0 harmonics, add

• Vrms in different volumes
• pk-pk in different volumes

add doc string coverage tests

Not getting good results segmenting the tutorial CT data

I don't think the math is quite right here.. I will try to fix.. I want to add an iterative approach anyway

marker_size_lower_tol=0.9, marker_size_upper_tol=1

# Set up min and max marker volumes based on expected number of markers
n_voxels_median = np.median(np.array(n_voxels))
voxel_min = (1-self._marker_size_lower_tol) * n_voxels_median
voxel_max = (1+self._marker_size_upper_tol) * n_voxels_median

This code is currently giving this for the CT data:

Median marker volume: 748.0
Expected marker volume: 74.79999999999998 to 1496.0

(Too many markers found at the moment)

run darg lint

darglint -s sphinx TopasOpt

incorporate gradient strength scaling into 'calculate harmonics wrapper'

smoother interface between MarkerVolumeCentroids and dicom data

We should probably just have a 'save' and 'load' function?

Can delete docs because it is built into a new branch with github actions

why are markdown emojis not working in the docs

add spherical harmonic theory to the docs

add test alignment step

add a test to ensure the alignment can recover a known rigid misalignment

add this to utilities

def calculate_harmonics(forward_volume, back_volume, n_order=8):
    # match the markers
    matched_markers = MatchedMarkerVolumes(ct_volume, forward_volume, sorting_method='radial', ReferenceMarkers=11,
                                              WarpSearchData=True, ReverseGradientData=back_volume)
    matched_markers.MatchedCentroids.to_csv('MatchedMarkerVolume.csv')

    # calculate B fields
    B_fields = ConvertMatchedMarkersToBz(matched_markers.MatchedCentroids, forward_volume.dicom_data)

    # calculate harmonics
    # B0
    B0_data = B_fields.MagneticFields[['x', 'y', 'z', 'B0']]
    B0_data = B0_data.rename(
        columns={"B0": "Bz"})  # spherical harmonics code expects to receieve one field called Bz
    B0_Harmonics = SphericalHarmonicFit(B0_data, n_order=n_order, r_outer=150)
    # Gx
    GradXdata = B_fields.MagneticFields[['x', 'y', 'z', 'B_Gx']]
    GradXdata = GradXdata.rename(
        columns={"B_Gx": "Bz"})  # spherical harmonics code expects to receieve one field called Bz
    G_x_Harmonics = SphericalHarmonicFit(GradXdata, n_order=n_order, r_outer=150)
    G_x_Harmonics.harmonics.to_csv('G_x_harmonics.csv')
    # Gy
    GradYdata = B_fields.MagneticFields[['x', 'y', 'z', 'B_Gy']]
    GradYdata = GradYdata.rename(
        columns={"B_Gy": "Bz"})  # spherical harmonics code expects to receieve one field called Bz
    G_y_Harmonics = SphericalHarmonicFit(GradYdata, n_order=n_order, r_outer=150)
    G_y_Harmonics.harmonics.to_csv('G_y_harmonics.csv')
    # G_z
    GradZdata = B_fields.MagneticFields[['x', 'y', 'z', 'B_Gz']]
    GradZdata = GradZdata.rename(
        columns={"B_Gz": "Bz"})  # spherical harmonics code expects to receieve one field called Bz
    G_z_Harmonics = SphericalHarmonicFit(GradZdata, n_order=n_order, r_outer=150)
    G_z_Harmonics.harmonics.to_csv('G_z_harmonics.csv')

    return B0_Harmonics, G_x_Harmonics, G_y_Harmonics, G_z_Harmonics

remove double name def in utilities.convert_spherical_harmonics

and add a numpy case to the tests, which would have caught this...

Update URL to report

Auto export to slicer

At the moment, the code automatically exports to slicer (makes the slicer_centroids.mrk.json file and overwrites the old one) each time the segmentation is run.

I think this should be a user run function? Like run the segmentation, then if it looks right, export to slicer? You could overwrite a 'correct' segmentation with a wrong one if you're playing around with the parameters

visualise dicom data in reports?

https://plotly.com/python/visualizing-mri-volume-slices/

build and publish docs automatically

https://github.com/peaceiris/actions-gh-pages

when we have a large load region, we want to figure out the optimal threshold without including the load...

Fix logger message

In marker analysis
logger.warning('failed to extract phase encoding
should be frequency

add a test of CT read in

As noted in #72 this case is not currently covered, just need to add on CT slice to test data to do it.

Typo in tutorial

In the marker matching section

'# ground truth centroids ground_truth_volume = MarkerVolume(data_loc / 'CT' / 'MR.mrk.json', verbose=False)`

MR.mrk.json should be slicer_centroids.mrk.json

add a test that all examples run

Add read me to examples

figure out how many azimuthal/ elevation points we need to fit harmonics of a given order

also add a check; if user requests order=10 that their data can support that

Diagram should point to code docs

Add to utilities?

def enumerate_subfolders(data_loc):
    data_dict = {}
    subfolders = [ f.path for f in os.scandir(data_loc) if f.is_dir() ]
    for i, dir in enumerate(subfolders):
        folder_name = os.path.split(dir)[1]
        data_dict[str(i)] = folder_name
    for key in data_dict.keys():
        print(f"'{key}': '{data_dict[key]}',")

should harmonics be exactly the same for forward/reverse direction switch

Add phase encode type to MarkerVolume

the calculate harmonics wrapper might make more sense to accept the field data,

see here
https://github.com/ACRF-Image-X-Institute/MRI_DistortionCorrection/blob/main/examples/end_to_end.py

write convert_spherical_harmonics

Error when segmenting CT

Need to skip this when segmenting CT because there's no bandwidth tag

self._calculate_gradient_strength()

Specify coverage options in setup.cfg

https://coverage.readthedocs.io/en/6.3.2/config.html

Add export dicom data step to tutorials

need to put the 'get MR data' function inside methods somehow....

add # pragma: no cover to all plotting routines

run through examples pypi code

why is the sign of B0 harmonics sometimes reversed...

https://github.com/ACRF-Image-X-Institute/MRI_DistortionQA/blob/main/docsrc/Harmonic_sanity_checks.md

Test: recon data with only gradient harmonics should give zero distortion

add 'returns' to all docstrings in utilities

investigate switching order of 'forward' and 'reverse' datasets in Marker matching

make B0 table in report go next to B0 figure

Pretty sure the elekta distortion tests are mixing up R and d

add some checks into the marker matching step

• if max distortion exceeds some threshold
• if the distortion field is not smoothly varying

Change field analysis to harmonic analysis

add notes on gradient strength scaling to docs

add 'gradient_strength' to expected dicom_data in calc fields

add 'systemic error' to perturb markers