| Image | Maxium Projection | Epiproj |
|---|---|---|
 |
 |
 |
This is a set of three ITK based filters to compute the depth image of a surface in a volume and to project the surface along the depth axis.
Those filters can be used independently or together. Here combined Epiproj - Epithelium Projection - to project in 2D an apical fluorescence signal of an epithelium.
The Epiproj is a 2-step approach for computing the depth image of the signal (epiprojDepthMapGenerator) and projecting the signal (epiprojDepthMapProjector) based on the dept image.
Insight ToolKit (ITK) += 5.0
CMake += 3.0
All filters are higly inspired from the itkProjectionImageFilter for dimension managment and process threading. The core filter is the itkVolumeToDepthMapFilter which compute a DepthMap from a given Volume. The itkMultiscaleVolumeToDepthMapFilter integrate the itkVolumeToDepthMapFilter into a multiscale process to compute the depth map at each scale step of the volume using the depth map at the lower scale to initialise the current scale computation. Finaly the itkDepthMapProjectionFilter project a given volume using a given depth.
ITK filter that take a nD image (optimised for 3D, 4D+ was not evaluated) and compute a depth map of the signal localisation as output. The signal localisation detection m_Peak can be define as:
- the maximum intensity along the projection dimension (value = 0)
- the first relevant intensity peak along the projection dimension (value = 1)
The second definition require an additional value m_Tolerance to define the relevantness of the intensity Finaly, an initialisation depth map can be provided to speed up the computation.
An overlayer of the itkVolumeToDepthMapFilter that use a multiscale pyramide to compute the depth map. Here the filter first compute the map at the lower scale of the pyramide and use the output as initialisation for computing the map at the next scale level. THe multiscale approach allows a speed up of the process and is also used to controle the specificity of the process to small high scale structure such has small holes in the surface.
This filter will apply a maximum or average projection of a volume around a provided corresponding depth map.
Usage example of each filters can be found in their respective test:
- itkVolumeToDepthMapFilterTest.cxx
- itkMultiscaleVolumeToDepthMapFilterTest.cxx
- itkDepthMapProjectionFilterTest.cxx
Using the three ITK filter, two scripts can be found in the epiproj folder.
Those script allows the computation of a depthmap from a volume and the projection
of the volume using the computed depthmap.
Variouse parameters, smoothing, and preprocessing option were added to the script to
provide a usable two-steps projection program.
git clone https://github.com/StRigaud/Epiproj.git
mkdir epiproj/build
cd epiproj/build
ccmake -DCMAKE_BUILD_TYPE=Release ..
make
make test
All tests should pass, and compiled executable can be found in "bin" folder at the root of the project.
Usage: ./epiprojDepthMapGenerator
InputFileName (string) - path to input file.
OutputFileName (string) - path to output file.
Sigma (float) - smoothing parameters.
Options:
Type (string) - Computation on maximum (max) or variance (var) intensity.
Level (int) - Number of scaling level. (=5)
Peak (int) - Detecting peak. (=0)
Tolerance (float) - Intensity ratio (=0.1).
Delta (int) - Degree of freedom per step. (=1)
The options allows different detection type and higly depend on the data and the output expected. Type is a signal pre-processing option to compute the depthmap on. The default value (max) means that no alteration of the signal is performed, the variance option apply a local 2d variance filter. Peak allows to guide the intensity detection to the maximum peak (default behaviour) or to the first peak detected. The peak relevantness are then defined by the Tolerance value, not used if detecting maximum peak. Finaly the Delta is the ± freedom to explore at each scale step. A low value will not allow the algorithm to get too far away that what he detected a low scale, on the contrary a too high value will make it to adapt too much to every imperfection of the signal. See filter itkDepthMapProjectionFilter documentation for further details on the algorithm.
Usage: ./epiprojDepthMapProjector
InputFileName (string) - path to input file.
DepthFileName (string) - path to depth map file.
OutputFileName (string) - path to output file.
Options:
Median (int) - Median radius kernel. (=0)
Type (string) - Projection type, maximum (max), average (avg) intensity.
upperRange (int) - Upper range band. (=1)
lowerRange (int) - Lower range band. (=1)
shift (int) - Depth shift. (=0)
The options allows different projection. Median is a radius size of a pre-processing median filter applied to the signal before projection. A 0 value will skill the filter. Type will apply a projection type, maximum or average. While maximum will yield the best contrast result, the average may be relevant for quantification purposes. The upperRange and lowerRange are the number of z-plan upper and lower the depthmap you defined to be part of the projection band. Finaly the shift is z-axis translation operation to be applied to the depthmap before projection. See filter itkVolumeToDepthMapFilter and itkMuliscaleVolumeToDepthMapFilter documentation for further details on the algorithm.
The depthmap can be compute on a pre-processed signal, this allows to apply specific filter that change the dinamic of the signal. The most common use is with the local variance for example. From the raw signal, we compute the local 2D variance that we use as input to compute the depthmap. The depthmap is then used to project the raw signal. Because we used the local variance filter, we remove any uniforme layer of signal allowing here to focus the depthmap on the cell signal.
| Image | Local Variance | Depth Map | Projection |
|---|---|---|---|
|
 |
 |
|
@software{rigaud_2020,
author = {Rigaud Stephane},
title = {{Epiproj : A multiscale epithelium projection software}},
month = jun,
year = 2020,
publisher = {Zenodo},
version = {V2},
doi = {10.5281/zenodo.3878789},
url = {https://github.com/StRigaud/Epiproj.git}
}
This work was supported and tested by the Polarity, Division, and Morphogenesis Team at Institut Curie, Paris, France.