Correlation heatmap of many datasets

Question

I am working with niftis (Neuroimaging format) looking at 3D volumes of the brain.

I want to compare experiments with brain activity.

Therefore I have about 20 experiment files in nifti format.

and 4 brain activity files also in nifti format.

They have the same dimension of the brain and show some values in specific areas/pixels of the brain.

How can I plot a sns.heatmap, that has on the x-axis all experiments, and on the y-axis all activity files. I then want to see a heatmap correlation map, with plotted correlation values in heatmap colors.

I just found sns.heatmap plots for pandas dataframes. Can someone help?

I started with

import nibabel as nib

exp1 = nib.load("exp1.nii")
exp1= exp1.get_fdata()

act1 = nib.load('act1.nii.gz')
act1 = act1.get_fdata()

exp1, und act1 are read in as float64 of dimension (57, 66, 40).

Has someone an idea?

Example data:

act1[20,:,:]
Out[4]: 
array([[ 0.02153977,  0.03817031,  0.10424001, ...,  1.38311863,
         0.86190647,  0.36399585],
       [ 0.02117901,  0.03277704,  0.05148495, ...,  0.47603241,
         0.36187041,  0.16624629],
       [ 0.01256184,  0.02604919,  0.0477814 , ...,  0.13270944,
         0.11138337,  0.0545687 ],
       ...,
       [-1.        , -1.        , -1.        , ..., -1.        ,
        -1.        , -1.        ],
       [-1.        , -1.        , -1.        , ..., -1.        ,
        -1.        , -1.        ],
       [-1.        , -1.        , -1.        , ..., -1.        ,
        -1.        , -1.        ]])

exp1[20,:,:]
Out[8]: 
array([[0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0.23, 0.21, 0.11],
       ...,
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.],
       [0., 0., 0., ..., 0., 0., 0.]])

Answer 1

If you want to plot the correlation matrix using sns/seaborn you need to first extract the BOLD signal in the image(.nii) format that you have.

You can use the nilearn package to extract the BOLD time series signal for a specific atlas of interest (like AAL atlas which consists of 90 ROIs)

Then you can plot the correlation to see the correlations between the ROIs.

Hope this helps