Deep learning for NeuroImaging in Python.

Randomly bootstraps measures at specific vertices across a group of K subjects, assuming a vertex-to-vertex correspondence between hemispheres.

Init class.

Parameters:

prob : float

the probability of curuption.

n_vertices : int (N, )

the size of the cortical measures.

classmethod groupby(data, by=('texture',), n_neighbors=30, n_components=20, meta=None, weights=None)[source]¶

Regroup subjects based on a combination of metrics.

Parameters:

data : array (M, N)

input data/textures.

by : list of str, default (‘texture’, )

used to determine the metrics.

n_neighbors : int, default 30

the number of neighbors.

n_components : int, default 20

the number of PCA components, used to reduce the input data size.

meta : pandas.DataFrame, default None

the external data.

weights : array, default None

the weight applied to each distance matrix formed from the metric defined in the by parameter.

Returns:

neigh_ind : array (M, n_neighbors)

indices of the nearest subjects in the population.

run(data, group_data, n_samples=1)[source]¶

Applies the group bootstaping.

Parameters:

data : array (N, )

input data/texture.

group_data : array (k, N)

input group data/textures.

n_samples : int, default 1

the number of bootstraping to be performed.

Returns:

data : arr (N, ) or (M, N)

bootsraped input data.