Is there a way to plot surface images with transparent thresholding? #nilearn

Software Support
, , ,
1

Nilearn version: 12.0
Nilearn environment: Conda environment on Linux (Ubuntu 22.04.5 LTS) with Python 3.9
I am running Nilearn in a .ipynb file (Jupyter notebook) through VS Code.

Hi all,

I am trying to produce some surface maps with transparent thresholding and I am running into errors. I saw with this example that it is possible on the volumetric images (Plotting images with transparent thresholding - Nilearn), but when I try to add the transparency and transparency_range parameters to plot_surf_stat_map, I run into errors. The error is a plot_surf_stat_map does not accept the parameter [transparency/transparency_range].

Based on this issue (Change use of parameter alpha in surface plotting functions and deprecate `bg_on_data` · Issue #3503 · nilearn/nilearn · GitHub) I am suspecting the transparency thresholds feature is not available in nilearn yet for surfaces? I can’t totally tell if this was the same feature or a different feature, and whether or not the feature they were working on was implemented.

With that, I am wondering, is there a way to plot transparent thresholding on a surface map, perhaps using different parameters or methods than what I am attempting with plot_surf_stat_map?

I am also thinking about adding contours to the surface image, as in this volumetric example (Plotting images with transparent thresholding - Nilearn) but I ran into the transparent thresholding roadblock first.

Minimal Example code:

#prepare the surface mesh
big_fsaverage = datasets.fetch_surf_fsaverage('fsaverage')

#some image params
vmin = 0.5
threshold = 3.75
max = 10

image = "stats.nii"

#load in our image
stat_img = nibabel.load(image)

big_texture_left = surface.vol_to_surf(stat_img, big_fsaverage.pial_left, interpolation="nearest",  n_samples=1,radius=0)

plotting.plot_surf_stat_map(
      big_fsaverage.infl_left,
      big_texture_left,
      hemi="left",
      view = "lateral",
      colorbar=True,
      cmap="cold_hot",
      #transparency=image, #this line does not work
      #transparency_range=[vmin, threshold], #this line does not work
      bg_map=big_fsaverage.sulc_left,
      avg_method="median",
      cbar_tick_format="%i",
      vmin = -max,
      vmax = max,
        )

Thanks for any help,
-shabkr

2

the feature is not yet available for nilearn: I have put it on our todo list a while back (extend transparent plotting to non-plotly surface plots · Issue #5215 · nilearn/nilearn · GitHub) but I have not yet got to it.

for plotting contours on surfaces we have a dedicated function: nilearn.plotting.plot_surf_contours - Nilearn

see in this example: Making a surface plot of a 3D statistical map - Nilearn

1 Like
3

Thank you for the response (and for the tips on surface contours!). I’ll work without transparency thresholds for now, and look forward to that update someday.

1 Like