Using MRT instead of CT for radiotherapy planning would reduce dose administered to patient before actually stating the curative treatment. However, using a distorted image for therapy planing can have fatal consequences (especially for proton & ion beam therapy). Therefore making sure the obtained images depict the patients geometry accurately is crucial.
The aim of the project is to develop a tool that calculates the distortion of an MRT image by comparing it to an CT Image. More information here: https://github.com/cyberspeck/warp-thesis/raw/master/latex2/bachelor-thesis.pdf
The ultimate goal would be a script that automatically finds all rods (of a phantom designed for this purpose) in a 3D CT & MRT image, compares them and creates a 3D vector field of the measured distortion not only of the position, but also the rod profile.
Steps: 0) way to import DICOM images and convert into useful format for further computing using python https://pyscience.wordpress.com/2014/10/19/image-segmentation-with-python-and-simpleitk/ http://insightsoftwareconsortium.github.io/SimpleITK-Notebooks/
-
compare CT & MRT image of 1 rod in 2D a) calculating the bending and displacement of the rod in MRT image relative to CT (important to keep in mind: CT shows plastic and filling, MRT only filling) http://stackoverflow.com/questions/18435003/ndimages-center-of-mass-to-calculate-the-position-of-a-gaussian-peak http://docs.scipy.org/doc/scipy-0.14.0/reference/generated/scipy.ndimage.measurements.center_of_mass.html
-
calculating the distortion of the circular rod profile a) using dice-coefficient (aka "sorensen index") (see also Medpy: http://pythonhosted.org/MedPy/_modules/medpy/metric/binary.html) b) calculating & plotting waveform along x-axis of image “waveform” = histogram showing signal intensity (ordinate) along image coordinate from left to right (abscissa) c) measure along various angles (e.g. bootom left to upper right, etc → different angles)
-
imporving scirpts: a) 2D → 3D b) automatically finding multiple rods (could be done similar to this: https://blancosilva.wordpress.com/2010/12/15/image-processing-with-numpy-scipy-and-matplotlibs-in-sage/) c) working with the uncropped aligned CT & MRT images d) use python to align images, get inspired here: https://itk.org/Wiki/SimpleITK/Tutorials/MICCAI2015 and https://github.com/InsightSoftwareConsortium/SimpleITKTutorialMICCAI2015 or http://insightsoftwareconsortium.github.io/SimpleITK-Notebooks/ (notebooks 60-67) or https://aaltodoc.aalto.fi/handle/123456789/20891 and https://aaltodoc.aalto.fi/bitstream/handle/123456789/20891/master_Tapani_Karoliina_2016.pdf?sequence=1&isAllowed=y or http://kevin-keraudren.blogspot.co.at/2014/12/medical-image-analysis-ipython-tutorials.html and https://github.com/curiale/Medical-Image-Analysis-IPython-Tutorials
Required packages (recommended to use anaconda): install anaconda create virtual environment conda create -n ENV2 python=2.7 spyder ipykernel source activate ENV2 pip install SimpleITK pip install numpy pip install scipy pip install matplotlib pip install scikit-image conda install notebook ipykernel ipython kernel install --user
Separate Matplotlib figure windows pop up for Spyder 2.3/3.0: Tools > Preferences > Ipython Console > Graphics > Graphics Backend > Backend: “automatic”