Source code for sekupy.io.connectivity

from sekupy.dataset.collections import SampleAttributesCollection
from sekupy.dataset.base import Dataset

from sklearn.preprocessing import LabelEncoder

from sekupy.io.base import add_attributes, load_attributes, load_filelist
from sekupy.io.subjects import add_subjectname
from sekupy.utils.atlas import get_atlas_info

import numpy as np
import logging

logger = logging.getLogger(__name__)

# TODO : Document 



def conn_transformer(data):

    data = np.rollaxis(np.vstack(data), 0, 3)
    parcels = data.shape[0]
    data = data[np.triu_indices(data.shape[0], k=1)].T
    logger.debug(data.shape)

    return data, parcels





def time_transformer(data):
    data = np.vstack(data)
    return data, data.shape[1]





def power_transformer(data):
    data = np.vstack(data)
    return data, data.shape[1]





def load_mat_data(path, subj, folder, **kwargs):

    from scipy.io import loadmat

    meg_transformer = {'connectivity': conn_transformer,
                       'power': power_transformer,
                       'timevarying': time_transformer
                       }

    key = kwargs['mat_key']
    transformer = meg_transformer[kwargs['transformer']]

    # load data from mat
    filelist = load_filelist(path, subj, folder, **kwargs)

    data = []
    info = {'file': []}
    for f in filelist:

        logger.info("Loading %s..." %(f))
        mat = loadmat(f)
        datum = mat[key]
        logger.debug(datum.shape)
        data.append(mat[key])
        info['file'] += [f for _ in range(datum.shape[0])]

    data, info['parcels'] = transformer(data)

    return data, info





def load_mat_ds(path, subj, folder, **kwargs):

    logger.debug(kwargs)
    data, info = load_mat_data(path, subj, folder, **kwargs)

    # load attributes
    attr = load_attributes(path, subj, folder, **kwargs)

    if (attr is None) or (data is None):
        return None

    attr, labels = edit_attr(attr, data.shape)

    logger.debug(data.shape)
    logger.debug(attr)

    ds = Dataset.from_wizard(data, attr.targets, flatten=False)
    ds = add_subjectname(ds, subj)
    ds = add_attributes(ds, attr)

    ds = add_labels(ds, info['parcels'], **kwargs)

    #ds.fa['roi_labels'] = labels
    ds.fa['matrix_values'] = np.ones(data.shape[1], dtype=np.int8)

    ds.sa['chunks'] = LabelEncoder().fit_transform(ds.sa['name'])
    ds.sa['file'] = info['file']

    return ds





def add_labels(ds, parcels, **kwargs):

    info = get_atlas_info(kwargs['atlas'])
    labels = info[4]
    labels = labels[:parcels]

    if kwargs['transformer'] == 'connectivity':
        idx_from, idx_to = np.triu_indices(parcels, k=1)
        nodes_from = [labels[i] for i in idx_from]
        nodes_to = [labels[i] for i in idx_to]
        ds.fa['nodes_1'] = nodes_from
        ds.fa['nodes_2'] = nodes_to
    elif kwargs['transformer'] == 'power':
        ds.fa['nodes'] = labels
    elif kwargs['transformer'] == 'timevarying':
        n_nodes = len(labels)
        idx_from, idx_to = np.triu_indices(n_nodes, k=1)
        nodes_from = [labels[i] for i in idx_from]
        nodes_to = [labels[i] for i in idx_to]
        network_from = []
        ds.fa['nodes_1'] = nodes_from
        ds.fa['nodes_2'] = nodes_to

        pass

    # TODO: Power labels

    return ds





def edit_attr(attr, shape):

    factor = int(shape[0]/len(attr.targets))

    attr_ = dict()
    for key in attr.keys():
        attr_[key] = []
        for label in attr[key]:
            attr_[key] += [label for i in range(factor)]

    return SampleAttributesCollection(attr_), None