Multi-modal parcellation (MMP) distances calculation¶
Data¶
Data obtained from: https://neuroimaging-core-docs.readthedocs.io/en/latest/pages/atlases.html
A direct link: https://bitbucket.org/dpat/tools/raw/master/REF/ATLASES/HCP-MMP1_UniqueRegionList.csv
Plots¶
3D plot example source: https://jakevdp.github.io/PythonDataScienceHandbook/04.12-three-dimensional-plotting.html
%matplotlib inline
import numpy as np
import matplotlib.pyplot as plt
import pandas as pd
import seaborn as sns
from mpl_toolkits import mplot3d
from scipy.spatial.distance import squareform, pdist
Load the data¶
data = pd.read_csv('HCP-MMP1_UniqueRegionList.csv', index_col=0)
data
Plotting¶
Simply plot all the coordinates first
fig = plt.figure()
ax = plt.axes(projection='3d')
ax.scatter3D(data['x-cog'],
data['y-cog'],
data['z-cog']);
Create masks for the left and the right hemispheres
data[data.index.str.contains('_L')]
left_mask, right_mask = [data.index.str.contains(hemisphere) for hemisphere in ('_L', '_R')]
print('Left mask')
data[left_mask][:3]
print('Right mask')
data[right_mask][:3]
Plotting with masks¶
fig = plt.figure()
ax = plt.axes(projection='3d')
plt.title('Left hemisphere: red\nRight hemisphere: green')
ax.scatter3D(data['x-cog'][left_mask],
data['y-cog'][left_mask],
data['z-cog'][left_mask],
c=data['z-cog'][left_mask],
cmap="Reds");
ax.scatter3D(data['x-cog'][right_mask],
data['y-cog'][right_mask],
data['z-cog'][right_mask],
c=data['z-cog'][right_mask],
cmap="Greens");
ax.set_xlabel('x')
ax.set_ylabel('y')
ax.set_zlabel('z');
Plot labels¶
Preview one of the labels:
data.loc['V1_L']
Plot with selected labels¶
For text annotation examples see: https://matplotlib.org/stable/gallery/mplot3d/text3d.html
labels = ['V1_L', 'V1_R', '1_L', '1_R', 'TGv_L', 'TGv_R', '10pp_L', '10pp_R']
fig = plt.figure()
ax = plt.axes(projection='3d')
ax.view_init(elev=20., azim=-45)
plt.title('Left hemisphere: red\nRight hemisphere: green')
ax.scatter3D(data['x-cog'][left_mask],
data['y-cog'][left_mask],
data['z-cog'][left_mask],
c=data['z-cog'][left_mask],
cmap="Reds");
ax.scatter3D(data['x-cog'][right_mask],
data['y-cog'][right_mask],
data['z-cog'][right_mask],
c=data['z-cog'][right_mask],
cmap="Greens");
# Display labels (texts) to see which ROI is where.
# None is direction of the text. None is en face.
for label in labels:
region = data.loc[label]
ax.text(region['x-cog'],
region['y-cog'],
region['z-cog']*1.2,
label,
None,
color='blue')
ax.set_xlabel('x')
ax.set_ylabel('y')
ax.set_zlabel('z');
It seems that this is presented in radiology notation, i.e., where left is right and right is left. It doesn't really matter here, as we are primarily interested in the distances between the regions.
Calculate the distances¶
Preview all the columns with coordinates.
data[['x-cog', 'y-cog', 'z-cog']]
Prepare an array and data frame with distances¶
distances_array = squareform(
pdist(X=data[['x-cog', 'y-cog', 'z-cog']],
metric='euclidean')
)
distances_df = pd.DataFrame(distances_array, index=data.index, columns=data.index)
distances_df
Visualize the distances¶
# Prepare a bigger figure.
# Axes are unequal because colorbar will be smaller and
# seaborn doesn't compensate for that automatically.
plt.figure(figsize=(16, 13))
# Color bar keyworded arguments are passed so that the colorbar
# will be smaller -- this way the visualization looks better.
ax = sns.heatmap(distances_df,
cbar_kws={"shrink": 0.5})
ax.set_xlabel('Region name')
ax.set_ylabel('Region name');
Save distance values to CSV¶
distances_df.to_csv('MMP_distances.csv')
!ls -1