baiwenjia/cimas

Cardiac Image Multi-Atlas Segmentation pipeline (CIMAS)

cimas's Issues

Replace "resample" by "mirtk resample-image"

The MIRTK includes the "resample" command from IRTK as resample-image command.

rview binary for MIRTK

Hi @baiwenjia, nice project! I just thought I'd drop a note that users running macOS or Ubuntu may also download the MIRTK "view" binary (i.e., rview with slight changes which can read MIRTK transformation files) without need to enter their personal details. The binary, when copied to the $MIRTK_PREFIX/lib/tools/ directory of the MIRTK installation, can be executed using:

mirtk view [<image>]

Replace "prreg" by "mirtk init-dof"

The MIRTK "register" command replaces the obsolete IRTK prreg binary:

mirtk register target-landmarks.vtk source-landmarks.vtk \
    -dofout out.dof -model Rigid \
    -par "Point set correspondence" "Closest point"

However, when you have a set of landmarks, you are better off with the "mirtk init-dof" command which uses the closed form solution of the least squares problem:

mirtk init-dof out.dof -displacements target-landmarks.vtk source-landmarks.vtk

-displacements, -disp <pset1> <pset2> [<cor12>]
      Create transformation which minimizes the mean squared distance between pairs
      of fiducial point sets <pset1> and <pset2>. When a single <pset> is given,
      it creates a transformation which approximates the displacement vectors stored
      in the "vectors" point set attributes array. Option can be used multiple times.

Replace "surface_smooth" by "mirtk smooth-surface"

You can use the smooth-surface command of MIRTK to perform Laplacian smoothing of surface node positions (or scalar data...).

It appears that

os.system("surface_smooth {0} {0} 500 -relaxation 0.01".format(mesh_fit))

can be replaced by

os.system("mirtk smooth-surface {0} {0} -iterations 500 -lambda 0.01 -combinatorial -exclnode".format(mesh_fit))

(or -distance option instead of -combinatorial to use Laplacian weights inversely proportional to the distance of adjacent nodes). The option -inclnode or -exclnode can be used to specify whether the node that is smoothed itself should contribute to the average or not, respectively.