Skip to content

Latest commit

 

History

History
193 lines (131 loc) · 5.67 KB

README.md

File metadata and controls

193 lines (131 loc) · 5.67 KB
Raw

MINC2 for Julia

CI

Read and write MINC2 files from Julia, reading and writing .xfm transformation files, Manipulations with volumes

Examples

Read a 3D minc volume, calculate mean value

Minc command:

> mincstats -mean mni_icbm152_t1_tal_nlin_sym_09c.mnc
Mean:              29.61005195

Julia version:

using Minc2
using StatsBase

# read volume from minc file, represent it as array of doubles (Float64 in julia world)
icbm=Minc2.read_volume("mni_icbm152_t1_tal_nlin_sym_09c.mnc", store=Float64)
@show mean(Minc2.array(icbm))

Output mean(Minc2.array(icbm)) = 29.61005194874031

Read a 3D volume with real values, and another one with mask labels and show statistics per label

using Minc2
using StatsBase

# read T1w scan
icbm=Minc2.read_volume("mni_icbm152_t1_tal_nlin_sym_09c.mnc", store=Float64)
# read label mask, represent it as array of bytes 
lab=Minc2.read_volume("mni_icbm152_cls_tal_nlin_sym_09c.mnc", store=UInt8)

# calculate image statistics per label
for (i,l) in [1=>"CSF",2=>"GM",3=>"WM"]
    println("$(l):$(mean(Minc2.array(icbm)[ Minc2.array(lab).==i ]))")
end

Output:

CSF:31.263195070792978
GM:64.88399427312729
WM:84.74593912912998

Apply transformation stored in .xfm file to a minc volume, save result

Create transformation:param2xfm -rotations 30 0 0 rotate.xfm

using Minc2

# read T1w scan
icbm=Minc2.read_volume("mni_icbm152_t1_tal_nlin_sym_09c.mnc", store=Float64)

# read transformation
tfm=Minc2.load_transforms("rotate.xfm")

# apply transformation to the volume
transformed_icbm=Minc2.resample_volume(icbm; tfm, order=2, fill=0.0)

# write the resulting volume into file, append history
# the resulting volume will be stored with unsigned short minc data type
# but will preserve floating point vales used fixe-point arithmetics
Minc2.save_volume("transformed_icbm.mnc",transformed_icbm, store=UInt16, history="Julia example")

Integrate jacobians per ROI, based on a transformation in .xfm file

using Minc2

# read ROI labels
rois = Minc2.read_volume("rois.mnc", store=UInt8)

# calculate voxel volume 
start, step, dir_cos = Minc2.decompose(Minc2.voxel_to_world(rois))
voxel_volume = abs(reduce(*,step,init=1))

# read transformation
tfm = Minc2.load_transforms("nonlinear.xfm")

# define sampling parameters for jacobian field
jacobians = Minc2.empty_volume_like(lab, store=Float64)

# calculate per-voxel jacobians
Minc2.calculate_jacobian!(tfm, jacobians)

# Integrate per ROI
for (i,l) in [1=>"ROI-1",2=>"ROI-2",3=>"ROI-3"]
    println("$(l):$( sum((Minc2.array(rois) .== i) .* Minc2.array(jac) )*voxel_volume)")
end

Graphic examples, using Makie

Show MNI-ICBM152 template contours with tissue masks overlays

using CairoMakie
using Minc2

# read T1w scan
icbm=Minc2.read_volume("mni_icbm152_t1_tal_nlin_sym_09c.mnc", store=Float64)
# read label mask, represent it as array of bytes 
lab=Minc2.read_volume("mni_icbm152_cls_tal_nlin_sym_09c.mnc", store=UInt8)


fig = Figure()
gc = fig[1, 1] = GridLayout()

Minc2.draw_outline_with_labels(gc, icbm, lab, 
    labels=Dict([1=>"CSF",2=>"GM",3=>"WM"]),
    nslices = 5)

resize_to_layout!(fig)

save("mni_icbm152_segmentation.png", fig, px_per_unit = 1)

Will produce

MNI-ICBM152

Show MNI-ICBM152 template contours with GM proability map

using CairoMakie
using Minc2

# read T1w scan
anat=Minc2.read_volume("mni_icbm152_t1_tal_nlin_sym_09c.mnc",store=Float64)
# read field
gm=Minc2.read_volume("mni_icbm152_gm_tal_nlin_sym_09c.mnc",store=Float64)

fig = Figure()
gc = fig[1, 1] = GridLayout()

# set background to transparent
Minc2.array(gm)[Minc2.array(gm) .< 1e-6] .= NaN

Minc2.draw_outline_with_heatmap(gc, anat, gm, 
    heat_limits=(0.0,1.0),cmap=:turbo,
    nslices = 5)

resize_to_layout!(fig)

save("mni_icbm152_gm.png", fig, px_per_unit = 1)

Will produce MNI-ICBM152

More examples

See examples directory for more examples:

  • bloor.jl - Apply Blooring kernel to minc volume, equivalent to mincbloor
  • copy.jl - Create a minc volume with the same dimensions as another one, copy voxels
  • draw_heatmap.jl - Show MNI-ICBM152 template contours with GM proability map, as above
  • draw_seg.jl - Show MNI-ICBM152 template contours with tissue masks overlays, as above
  • list_meta.jl - List all metadata stored with minc file, similar to mincheader
  • mriview_dual.jl - Show two minc volumes side-by-side, interactively
  • mriview.jl - Show minc volume, interactively
  • multilabel_overlap.jl - calculate voxel-wise generalized overlap coeffecient , method from http://dx.doi.org/10.1109/TMI.2006.880587
  • resample_grid.jl - apply transformation to _grid file
  • resample_itk.jl - resample nifti file with transformations produces by ANTS (multiple transformations can be applied in one go), equivalent to antsApplyTransforms
  • resample.jl - apply transformation to minc volume, equivalent to mincresample
  • xfm_to_jacobian.jl - calculate jacobian determinant field, for a given .xfm file
  • mincinfo.jl - show information about mincfile(s) and store it in csv file, equivalent to mincinfo

Documentation

Preliminary documentation is available at https://www.ilmarin.info/Minc2.jl/

Documentations for underlying libminc and minc tools is available at https://en.wikibooks.org/wiki/MINC