Plot 3D nifiti on surface using nilearn

Hi everyone,

I have a 3D nifty file (brain map) that I would like to plot on a surface.

I am using the following code however, texture variable contains NaNs and the plotting function fails.

Any tip how to overcome this?

Code:

import numpy as np, os, sys
from nilearn import plotting, datasets; from nilearn.surface import load_surf_data
import numpy as np, os, warnings, matplotlib.pyplot as plt
from matplotlib.ticker import FormatStrFormatter
from matplotlib import ticker
from mpl_toolkits.mplot3d import Axes3D
from nilearn import surface

hemisphere = 'left'
view = 'medial'
black_bg = True
colorbar= True
cmap = 'hsv'

# Load the surface 
fsaverage = datasets.fetch_surf_fsaverage('fsaverage')

# The stat map
stat_img = '/Users/loukas/Desktop/stat_map.nii'

# The texture
texture = surface.vol_to_surf(stat_img, fsaverage.pial_left)

fig = plt.figure(dpi=300)
ax = fig.add_subplot(111, projection='3d')
plt.gcf().axes[0].yaxis.set_major_formatter(ticker.FormatStrFormatter("%f"))
# Plotting
plotting.plot_surf_roi(fsaverage["infl_{}".format(hemisphere)], roi_map = texture, \
	hemi = hemisphere, view = view, bg_map = fsaverage["sulc_{}".format(hemisphere)], \
	bg_on_data = True, darkness = 0.6, output_file = 'mapped_signal.png', \
	cmap = cmap, colorbar = colorbar, black_bg=black_bg, axes= ax, figure=fig, )
#for ax in plt.gcf().axes:
#	ax.yaxis.set_major_formatter(ticker.FormatStrFormatter(">%.4f<"))
plotting.show()
plt.close()

Data: Easyupload.io - Upload files for free and transfer big files easily.
Pass: makis

A quick and dirty fix is to smooth slightly (e.g. fwhm=1mm) prior to projection. Internally, the code replaces NaNs with 0’s.
There is proabbly a better way to do that, though…
My 2c,

Hi @makaros622

I had a look at this today, and it turns out plot_surf_stat_map can handle textures with nans while plot_surf_roi cannot…
I don’t see a good reason for this, so I opened the following PR to fix the issue: handle nans in plot_surf_roi by NicolasGensollen · Pull Request #2743 · nilearn/nilearn · GitHub

Hope this helps!

Great. Thanks. It is indeed strange that only the first function can handle nans