I have a dicom series folder with MR data.
With pydicom or simpleITK i can read this data, and now i want to do the following:
1) MPR reformatting. I want to create the other 2 primary projections for the data, i.e., if my dataset was taken in the axial projection, then i want to have sagittal and coronal and so on.
2) I have some points that are stored in the world coordinates. I want to recompute the world coordinates to my image coordinates in the specific plane.
I know how to do it using a medical imaging software, for instance, MeVisLab or Slicer, but how to do it in python to keep the processing in one language/ without switching between software packages? Are there any libraries with examples?
Thank you.
Tetyana ,
You can use "nmrglue" package to manipulate 3D MR data. Hope it will work.
Thank you! But it is not for MRI data (namely, images), but for NMR data, which is completely different :)
If your image data is stored as a 3D NumPy array, then the simplest approach for MPRs is to swap the axes using the transpose() method, e.g.:
import numpy as np
a = np.arange(128*96*64).reshape((128,96,64)) # 64 slices 128 x 96
mpr1 = a.transpose((0,2,1)) # -> 96 slices 128 x 64
mpr2 = a.transpose((1,2,0)) # -> 128 slices 96 x 64
(cave: image data may be mirrored or upside-down, this can be dealt with e.g. mpr1 = mpr1[:,::-1,:] for the second axis).
Coordinate transformations are affine transformations - you need something like the affine matrix in NIfTI data to translate between different coordinate systems.
Thank you, Olaf! If my sequence is isotropic, e.g., 1x1x1 mm^3, then simple rearrangement works. But i guess, i need some interpolation in the case, for instance, 1x2.5x4mm or so?
2D interpolation (e.g. with scipy.interpolate) is one possibility. A simpler solution is to set the correct aspect ratio of the pixels in the viewer, e.g.
import matplotlib.pyplot as plt
plt.imshow(a[:,:,z], aspect=dy/dx) # pixel size (dx, dy)
plt.show()
The nipy ecosystem http://nipy.org/ might be helpful for you. Nibabel for loading and conversion of image volumes, and nipy http://nipy.org/nipy itself for a multitude of image operations, including resampling and mapping of coordinate systems.
> how to do it in python to keep the processing in one language/ without switching between software packages?
All feature of 3D Slicer (www.slicer.org) is accessible from Python. No need to use any other languages or software packages. You can also fully customize and extend the application to do what you need and only what you need, get FDA approval, use it on patients, etc.
Thank you, Andras! Can you please be more specific how to access the Slicer's functionality from Python? I mean, can i import the slicer's functions? Is there any examples on that?
Slicer has its own Python interpreter instance where you import your modules. VTK, SimpleITK, numpy, Qt, etc. are already available. See this tutorial for a quick overview: https://www.slicer.org/wiki/Documentation/Nightly/Training#Slicer4_Programming_Tutorial
If you have any questions, post them to https://discourse.slicer.org
Ah, i see, but it limits me to use only those Python libraries that are available in Slicer.
For now, you have to build Slicer on your computer if you want to install any Python packages. Alternatively, you can use virtual Python environment for accessing any Python packages - this is how modules that use deep learning work.
We expect to switch to Python3 in about a year and then you can pip install any package without the need for custom builds or virtual environments.
If you want it back in DICOM, then it gets a fair bit more complicated (mainly because I am not sure how well DICOM writers in ITK or other libraries actually work in respecting the original acquisition so to speak). Otherwise the Resample image filter from SimpleITK will have all you need.
https://itk.org/SimpleITKDoxygen/html/classitk_1_1simple_1_1ResampleImageFilter.html
Tetyana Ivanovska
Can you help me to read MRI dicom data With pydicom or simpleITK?
Main use of pydicom is to read metadata, not to reconstruct 3D volume, so if yo need to load a 3D MRI then you can use SimpleITK (there are lots of examples, just google for "simpleitk dicom read"). If you need to work with time sequences, 4D data, DTI, DWI MRI, etc. then you need something more sophisticated, such as dcm2niix or 3D Slicer.
If you want to visualize or interact with your 3D data, or you want to import and export DICOM, or you want to work with DICOM segmentation objects, then probably the best tool to use is 3D Slicer, as it has all these features nicely integrated and preconfigured. It provides a virtual Python environment where you can install all the Python packages you need (with pip install) and you can use convenient GUI, or use the built-in Jupyter notebook server, or run Python scripts.
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Programming Languages
- Python