Nifti图像进入scikit学习

Question

I wanted to use scikit-learn Machine learning variation to do data processing of my neuroimaging data, specifically, fMRI data in Nifti file type.

Nilearn provides the platform. However, I don't understand how the Nitimasker working principle. How it converts 4D fMRI data to 2D data for scikit-learn.

I have 4D data of 1 subject, ie (40, 64, 64, 1452) , a Haxby data. I use Nibabel for accessing the images. If I want to process one planar, [20, :, :, 1] to [20, :, :, 1452] , could I np.flatten it to be the [n_samples,n_features] for scikit-learn platform?

Answer 1

This is not a direct answer but have a look at nilearn which is an extension of scikit-learn for brain imaging data (not sure this is the correct description).

There is an example of the Haxby data