cfd-dicom

Tools for converting CFD data to 4D dicom series

At its most basic level, you can use cfd-dicom.py to convert scalars or integrated quantities to DICOM series to be rendered in Horos. Input data must be specified in the cfd-dicom.py script. In the future, could be modified to do batch jobs or pass args as command line args. Ultimately, DICOM series are stored to ./cfd-dicom/output/{case folder}_voxelsize={voxel_size}_nsteps={n_steps}/. Descriptions of voxel_size, n_steps, and all other necesary inputs are given below.

Usage - cfd-dicom.py args.

For now, these are immeadiately following the if __name__ == '__main__' line.

By default, these are specified to read the example data in cfd-dicom/example_data/

Scalars

1. mesh_path - path to .h5 mesh containing coordinates and topology
2. n_steps* - number of timesteps to sample. Will pick uniformly over all avaliable timesteps. E.g., if there are 1000 steps total and n_steps = 100 every 10th step will be written to DICOM. Eventually, more intelligent sampling schemes could be used, like "bulllet time"
3. voxel_size - isotropic voxel length, assumed to be in the same length scale as the unstructrued mesh, usually mm
4. quantities - velocity, q-criterion, or pathlines. Must be specified as a list. Can specify any combination of u, q, path depending on the DICOM series to be generated. Note: if path is specified, pathline args shown below must also be specified. If you are not interested in pathlines or streamlines, you can stop here and run the script.

Pathlines

5. particle_polydata_path - path to .vtp series containing tracked particle polydata in the same mesh as specified above. Assumes data are tracked particles written out of Paraview's ParticleTracer filter. If you use a naming convention different from Praview's defaults, e.g., ParticleId for particle ids, you will have to change indexing in relevant parts of the code.
6. sigma - Radius of blurring amount. Assumes same length scales as mesh/voxel size above.
7. track_length - Akin to "exposure time" (Steinman, 2000). Larger values will result in longer particle streaks. A track_length of n means that the particle field is imaged every n timesteps. Note: In this code there is a difference between "imaging" the flow (i.e., resetting the "light" on the lens) and actually writing a DICOM snapshot. The former is dependent on track_length and the later on n_steps. What is ultimately written to DICOM is the intersection of frames accounted for in the sampling with n_steps and the snapshot frames resulting from track_length
8. step_stride - Steps over which to compute the pathlines. Akin to frame rate from Steinman, 2000. For example, if step_stride=2, pathlines will be "drawn" between tracked particles at t=0, t=2, t=4, etc as opposed to step_stride=1 where lines will be drawn between each successive timestep (i.e., t=0, t=1, t=2, t=3, etc)

Once these are specified you may run cfd-dicom.py and will have to specify a bounding box for the ROI you wish to voxelize.

The most common errors will probably result from inconsistent naming conventions of the .h5 mesh or time series files and tracked .vtp particles. E.g., sometimes veocity is saved as u (newer) sometimes as Solution/u (older)

Some process work and/or other interesting (but no longer necessary) scripts can be founf in ./extra

Dependencies

pyvista, h5py, numpy, scipy, pathlib, skimage, pydicom

To do

1. Streamlines and/or velocity interpolation
2. There are some redundant/repetitive chunks of code, there is probably a better, more "streamlined ;-)" approach to organization