Dev analysis

[jfcabana]

Several impovements and functionnality additions to the analysis module. The code still lacks proper documentation, but the addition made are useful, thus the pull request.

Update analysis.py

• Added an 'apply' (default=True) argument to apply_factor_from_roi method to allow the use of this method to only get and print the scale factor in an ROI, and not applying this factor to the film.
• Changed get_profile_offsets method so that the profile position is taken from the show_results window instead of being at image center.
• Added show_results_ontype method. In the show_results() window, the get_profile_offsets workflow is launched, at the current x/y position, when 'enter' is pressed.
• Added 'plot_isodoses' method in DoseAnalysis class to show a 2D plot of isodoses (film dose and ref dose on same graph).
• Added a tool to be able to draw an interactive ruler in the graphs. Used in the plot_isodoses method in analysis for now. This tool is defined in a new "tools" module.
• Improved analysis report, including a contours plot showing 2D isodoses of film dose and reference dose.
• Added analyse_clusters method, which identifies the clusters of dose on the reference dose map, based on a given threshold relative to maximum dose. For each cluster found, interactive plots are shown to successively get median dose over an roi inside the cluster, and x/y profiles offset and width difference. Theses values are stored in the object under self.clusters_analysis. When publishing pdf report, if the argument plot_clusters_analysis=True, a page is added with details for each cluster. Goal would be to automate this analysis, but not ready for now.

Update imageRGB.py

• Baseclass method plotCB() is replaced by an argument to the plot() method instead, as plotCB() was just a duplicate of plot() with the addition of a line to add the colorbar. plotCB() is left there for now for backward compatibility, until we change to v2.
• Added 'plot_isodoses' method in BaseImage class to perform a contours plot.
• Added a clusters identification in the BaseImage class (detect_clusters method). The method returns a list of center of masses identified in the image, based on a threshold value (applied to the maximum value in the image)

[LuisOlivaresJ]

Looks great! This weekend I going to read the modifications on imageRGB.py

[LuisOlivaresJ]

Looks great! This weekend I going to read the modifications on imageRGB.py

plot_isodoses works on the demo file "A1A_Multi_6cm_001.tif" :smiley:

This is the code to reproduce:

from omg_dosimetry.tiff2dose import Gaf
from omg_dosimetry import tiff2dose
import matplotlib.pyplot as plt

# Load demo LUT
path_to_lut_file = tiff2dose.from_demo_lut()

# Load folder to demo tif file
path_to_tif_folder = tiff2dose.from_demo_image()

# Gaf init.
fit_type = 'rational'
clip = 500
gaf = Gaf(
    path_to_tif_folder,
    lut_file=path_to_lut_file,
    fit_type=fit_type,
    clip=clip,
    img_filt=3,
    crop_edges=0.1,
)

# Plotting
fig, axes = plt.subplots(1, 2)
gaf.dose_opt.plot(ax=axes[0], show=False, colorbar=True, title="Dose")
gaf.dose_opt.plot_isodoses(levels=[350], inline=True, ax=axes[1], show=False, title="Contour")

plt.show()

[LuisOlivaresJ]

Could it be useful to have a method to remove clusters with an area below a given value? Using this can be a solution. I am working on it.

[jfcabana]

Could it be useful to have a method to remove clusters with an area below a given value? Using this can be a solution. I am working on it.

Yes I think it's a good idea!

[LuisOlivaresJ]

Note: I have made a few changes (review my recent commit to analysis.py): mostly formatting changes. Tested and it's functional on my side, but should verify on your side that nothing breaks.

Clusters I have not tested as I haven't dived too deep on this at our center. Perhaps @LuisOlivaresJ could redo the test/example case with the updated "remove small clusters" implementation.

"remove_small_clusters" runs without errors :thumbsup:. Thanks @petruong