Module connectome.preprocessing.graph_metrics

functions to compute graph metrics for a given brain connectivity dataset based on the bct(py) package

Functions

def explain_graph_metrics() ‑> None

function to explain the different implemented grap metrics

def get_gms_from_pd(data: pandas.core.frame.DataFrame, regions: list = None, cols: list = None, **kwargs) ‑> pandas.core.frame.DataFrame

computes the graph metrics for connectivity data supplied in the form of a DataFrame

Examples:

>>> import numpy as np
>>> import pandas as pd
>>> from connectome.preprocessing.graph_metrics import get_gms_from_pd
>>>
>>> # checking the get_gms_from_pd function
>>> k = 8  # dim of the conn matrix
>>> obs = 10  # observations
>>> conn = pd.DataFrame(
>>>     np.random.normal(
>>>         loc=0.1,
>>>         scale=1.2,
>>>         size=int((k*(k-1)/2)*obs)).reshape(obs, int((k*(k-1)/2))),
>>>         columns=[str(i) + "_" + str(j)
>>>                  for i in range(k)
>>>                  for j in range(i+1, k)])
>>> conn_data_list = [flat_to_mat(conn.iloc[i, :]) for i in range(obs)]
>>> print(get_gms_from_pd(conn, list(conn.columns)))

Args

data

dataFrame containing the conn data

regions

list of names of the regions of the conn matrix

cols

list of columns of the DataFrame data which contain conn data

**kwargs

anything that´s passed to "get_grap_metrics"

Returns

DataFrame containing the computed graph metrics

def get_graph_metrics(conns: list, threshold: float = 0.65, col_names: list = None, use_abs: bool = False, gms: tuple = ('Degrees', 'Modularity', 'Community Structure', 'Clustering Coefficient', 'Characteristic Path Length', 'Node Betweenness', 'Density', 'Component Vectors', 'Transitivity')) ‑> tuple

• computes graph metrics for the given connectivity data

• retransforms the connectivity matrices to pearson correlation before computing the adjacency matrices

  Examples:

>>> import numpy as np
>>> import pandas as pd
>>> from connectome.preprocessing.graph_metrics import get_graph_metrics
>>>
>>> # checking the get_gms_from_pd function
>>> k = 8  # dim of the conn matrix
>>> obs = 10  # observations
>>> conn = pd.DataFrame(
>>>     np.random.normal(
>>>        loc=0.1,
>>>         scale=1.2,
>>>         size=int((k*(k-1)/2)*obs)).reshape(obs, int((k*(k-1)/2))),
>>>         columns=[str(i) + "_" + str(j)
>>>                  for i in range(k)
>>>                  for j in range(i+1, k)])
>>> conn_data_list = [flat_to_mat(conn.iloc[i, :]) for i in range(obs)]
>>> print(get_graph_metrics(conn_data_list))

Args

conns

list of numpy arrays containing the connectivity data

threshold

threshold of correlation to compute adjacency matrices

col_names

colnames of the connectivity matrices in case those are reordered IMPORTANT: refers to the regions - e.g. "1", "4" instead of "1_4", "1_5"

use_abs

whether absolute connectivity values should be used

gms

graph metrics to be computed - can be any of: ['Degrees', 'Modularity', 'Community Structure', 'Clustering Coefficient', 'Characteristic Path Length', 'Node Betweenness', 'Density', 'Component Vectors', 'Transitivity']

Returns

pd.DataFrame containing the computed graph metrics

def is_conn_col(x: str) ‑> bool

checks whether the given column name corresponds to the format of the connectivity data columns, i.e. "1_245", which refers to the connectivity between region 1 and region 245

Examples:

>>> test_names = ["1_2", "asf", "as_asd"]
>>> print([is_conn_col(x) for x in test_names])

Args

x

the provided column name

Returns

True if the column corresponds to the connectivity data format

def pd_to_arrays(data: pandas.core.frame.DataFrame, cols: list = None) ‑> list

extracts the connectivity data from a DataFrame and converts those into matrices / numpy arrays

Args

data

the DataFrame from which the connectivity data should be extracted

cols

the columns of the DataFrame which contain the data of interest

Returns

list of the connectivity matrices / numpy arrays