Scyther

Curvilinear Multiplanar Reformat for Dicom Images 👨‍⚕️ 🖥️

Current version: 0.2.1

This library provides reslicing related functionalities to probe a volume with a surface defined by the extrusion of an arbitrary curve. Powered by VTK and binded to python using pybind11.

Pre-requisite

• Install vmtk : see https://github.com/vmtk/vmtk
• Install Visual Studio 2017+ / a c++ compiler
• Install cmake-gui / ccmake

🏗️ Installation & build

Clone repository:

$ git clone https://github.com/dvisionlab/scyther
$ cd scyther

Clone and install pybind11:

$ git clone https://github.com/pybind/pybind11
$ cd pybind11
$ python setup.py install

Then, from the project root, run cmake:

$ cmake-gui ./src ./build

• set USE_PYHTON_INCLUDE_DIR ON
• set CMAKE_CONFIGURATION_TYPES Release
• set VMTK_DIR as path_to_vmtk_build/Install
• set VTK_DIR as path_to_vmtk_build/VTK_Build
• set ITK_DIR as path_to_vmtk_build/ITK_Build

Configure & generate (alt+c, alt+g) to obtain the .sln file. Open and build with Visual Studio.

In order to check the build and the python binding, open a shell and run:

$ pyhton
>>> import pyCmpr

📂 Modules

• main -> entry point for python binding or c++ stand-alone usage
• cmpr -> functions mapped to python, define the type of cmpr
• stack -> manipulate volume
• geometry -> manipulate slice geometry
• test -> testing code
• render -> visualization tools (using VTK render), useful for debugging

🗺️ Roadmap

• Return image pseudo-header
• Spatial points remapping
• Update readme
• Open issue for vmtk static on Linux

👨‍👩‍👦 Dependencies

• VTK / ITK / VMTK
• Pybind11

Usage example

import json
import pyCmpr as cmpr

# inputs
image_path      = "/path/to/image.nrrd"
frenetTangent   = list of frenet tangents for each spline point, [x,y,z,x,y,z,x,y,z...]
ptn             = list the parallel transport normals for each spline point, [x,y,z,x,y,z,x,y,z...]
seeds_pts       = list the spline points [x,y,z,x,y,z,x,y,z...]
sweep_dir       = sweeping line direction [x, y, z]
sweep_length    = sweeping line length, as float 
resolution      = number of points in the cmpr direction (ie number of columns in the resultiong image), as int 
stack_direction = direction to be used to create the slices stack for cmpr direction, as [x, y, z]
slice_dimension = size of the (squared) axial slice, as float
dist_btw_slices = distance between cmpr slices
n_slices        = number of cmpr slices

# straightened
volume = cmpr.compute_cmpr_straight(image_path, seeds_pts, frenetTangent, ptn, resolution, sweep_dir,
                                      stack_direction, slice_dimension, dist_btw_slices, n_slices, True)
# stretched
volume = cmpr.compute_cmpr_stretch(image_path, seeds_pts, resolution, sweep_dir,
                                      stack_direction, dist_btw_slices, n_slices, True)
# output
volume is an object that contains : 

volume["pixels_cmpr"]       = list of pixel values for cmpr volume
volume["pixels_axial"]      = list of pixel values for axial volume
volume["metadata"]          = list of metadata from original nrrd serie (origin, dimensions, bounds)
volume["dimension_cmpr"]    = dimensions of the resulting cmpr volume, [i,j,k]
volume["dimension_axial"]   = dimensions of the resulting axial volume, [i,j,k]
volume["iop_axial"]         = list of image orientation patient vectors, [1, y1, z1, x2, y2, z2, x1, y1, y1, ...]
volume["ipp_axial"]         = list of image position patient, [x, y, z, x, y, z, ...]