Source code for yasa.others

"""
This file contains several helper functions to manipulate 1D and 2D EEG data.
"""
import logging
import numpy as np
from scipy.interpolate import interp1d
from .numba import _slope_lstsq, _covar, _corr, _rms

logger = logging.getLogger('yasa')

__all__ = ['moving_transform', 'trimbothstd', 'sliding_window',
           'get_centered_indices']


def _merge_close(index, min_distance_ms, sf):
    """Merge events that are too close in time.

    Parameters
    ----------
    index : array_like
        Indices of supra-threshold events.
    min_distance_ms : int
        Minimum distance (ms) between two events to consider them as two
        distinct events
    sf : float
        Sampling frequency of the data (Hz)

    Returns
    -------
    f_index : array_like
        Filled (corrected) Indices of supra-threshold events

    Notes
    -----
    Original code imported from the Visbrain package.
    """
    # Convert min_distance_ms
    min_distance = min_distance_ms / 1000. * sf
    idx_diff = np.diff(index)
    condition = idx_diff > 1
    idx_distance = np.where(condition)[0]
    distance = idx_diff[condition]
    bad = idx_distance[np.where(distance < min_distance)[0]]
    # Fill gap between events separated with less than min_distance_ms
    if len(bad) > 0:
        fill = np.hstack([np.arange(index[j] + 1, index[j + 1])
                          for i, j in enumerate(bad)])
        f_index = np.sort(np.append(index, fill))
        return f_index
    else:
        return index


def _index_to_events(x):
    """Convert a 2D (start, end) array into a continuous one.

    Parameters
    ----------
    x : array_like
        2D array of indices.

    Returns
    -------
    index : array_like
        Continuous array of indices.

    Notes
    -----
    Original code imported from the Visbrain package.
    """
    index = np.array([])
    for k in range(x.shape[0]):
        index = np.append(index, np.arange(x[k, 0], x[k, 1] + 1))
    return index.astype(int)


# def get_sync_events(data=None, sf=None, detection=None, center='NegPeak',
#                     time_before=0.4, time_after=0.8):
#     """
#     Return the raw data of each detected slow-waves / spindles, after
#     centering to a specific timepoint.
#
#     This function can be used to plot an average template of the
#     detected slow-waves / spindles.
#
#     For more details, please refer to the `Jupyter notebook
#     <https://github.com/raphaelvallat/yasa/blob/master/notebooks/06_sw_detection_multi.ipynb>`_
#
#     Parameters
#     ----------
#     data : array_like or :py:class:`mne.io.BaseRaw`
#         1D or 2D EEG data. Can also be :py:class:`mne.io.BaseRaw`,
#         in which case ``data`` and ``sf`` will be automatically extracted.
#     sf : float
#         The sampling frequency of ``data``.
#         Can be omitted if ``data`` is a :py:class:`mne.io.BaseRaw`.
#     detection : :py:class:`pandas.DataFrame`
#         YASA's detection dataframe returned by the
#         :py:func:`yasa.sw_detect` and
#         :py:func:`yasa.spindles_detect` functions.
#     center : str
#         Landmark of the slow-waves / spindles to synchronize the timing on.
#         Default is to use the negative peak.
#     time_before : float
#         Time (in seconds) before ``center``.
#     time_after : float
#         Time (in seconds) after ``center``.
#
#     Returns
#     -------
#     df_sync : :py:class:`pandas.DataFrame`
#         Ouput long-format dataframe::
#
#         'Event' : Event number
#         'Time' : Timing of the events (in seconds)
#         'Amplitude' : Raw data for event
#         'Chan' : Channel (only in multi-channel detection)
#     """
#     # Check if input data is a MNE Raw object
#     if isinstance(data, mne.io.BaseRaw):
#         sf = data.info['sfreq']  # Extract sampling frequency
#         data = data.get_data() * 1e6  # Convert from V to uV
#
#     # Safety checks
#     assert isinstance(data, np.ndarray), 'data must be a numpy array'
#     assert isinstance(detection, pd.DataFrame), 'detection must be a dataframe'  # noqa
#     assert isinstance(sf, (int, float)), 'sf must be a float or an int'
#     assert center in ['PosPeak', 'NegPeak', 'Peak', 'MidCrossing', 'Start',
#                       'End'], 'center timepoint not recognized'
#
#     multi = False
#     if 'Channel' in detection.keys():
#         chan = detection['Channel'].unique()
#         n_chan = chan.size
#         multi = True if n_chan > 1 else False
#
#     if multi:
#         # Multi-channel (recursive call)
#         assert data.ndim > 1, 'Data must be 2D for multi-channel detection.'
#         df_sync = pd.DataFrame()
#         for c, k in detection.groupby('Channel'):
#             idx_chan = k.iloc[0, -1]
#             df_tmp = get_sync_events(data[idx_chan, :], sf, k.iloc[:, :-2],
#                                      center=center, time_before=time_before,
#                                      time_after=time_after)
#             if df_tmp is None:
#                 continue
#
#             df_tmp['Channel'] = c
#             df_tmp['IdxChannel'] = idx_chan
#             df_sync = df_sync.append(df_tmp, ignore_index=True)
#     else:
#         # Single-channel
#         data = np.squeeze(data)
#         assert data.ndim == 1, 'Data must be 1D for single-channel detection.'  # noqa
#         # Define number of samples before and after the peak
#         assert time_before >= 0
#         assert time_after >= 0
#         bef = int(sf * time_before)
#         aft = int(sf * time_after)
#         # Convert to integer sample indices in data
#         peaks = (detection[center] * sf).astype(int).to_numpy()
#         # Get centered indices
#         idx, idx_nomask = get_centered_indices(data, peaks, bef, aft)
#         # If no good epochs are returned, raise an error and return None
#         if len(idx_nomask) == 0:
#             logger.warning(
#                 'Time before and/or time after exceed data bounds, please '
#                 'lower the temporal window around center; returning None.'
#             )
#             return None
#
#         # Get data at indices and time vector
#         amps = data[idx]
#         time = np.arange(-bef, aft + 1, dtype='int') / sf
#
#         # Convert to dataframe
#         df_sync = pd.DataFrame(amps.T)
#         df_sync['Time'] = time
#         df_sync = df_sync.melt(id_vars='Time', var_name='Event',
#                                value_name='Amplitude')
#     return df_sync


[docs]def moving_transform(x, y=None, sf=100, window=.3, step=.1, method='corr', interp=False): """Moving transformation of one or two time-series. Parameters ---------- x : array_like Single-channel data y : array_like, optional Second single-channel data (only used if method in ['corr', 'covar']). sf : float Sampling frequency. window : int Window size in seconds. step : int Step in seconds. A step of 0.1 second (100 ms) is usually a good default. If step == 0, overlap at every sample (slowest) If step == nperseg, no overlap (fastest) Higher values = higher precision = slower computation. method : str Transformation to use. Available methods are:: 'mean' : arithmetic mean of x 'min' : minimum value of x 'max' : maximum value of x 'ptp' : peak-to-peak amplitude of x 'prop_above_zero' : proportion of values of x that are above zero 'rms' : root mean square of x 'slope' : slope of the least-square regression of x (in a.u / sec) 'corr' : Correlation between x and y 'covar' : Covariance between x and y interp : boolean If True, a cubic interpolation is performed to ensure that the output has the same size as the input. Returns ------- t : np.array Time vector, in seconds, corresponding to the MIDDLE of each epoch. out : np.array Transformed signal Notes ----- This function was inspired by the `transform_signal` function of the Wonambi package (https://github.com/wonambi-python/wonambi). """ # Safety checks assert method in ['mean', 'min', 'max', 'ptp', 'rms', 'prop_above_zero', 'slope', 'covar', 'corr'] x = np.asarray(x, dtype=np.float64) if y is not None: y = np.asarray(y, dtype=np.float64) assert x.size == y.size if step == 0: step = 1 / sf halfdur = window / 2 n = x.size total_dur = n / sf last = n - 1 idx = np.arange(0, total_dur, step) out = np.zeros(idx.size) # Define beginning, end and time (centered) vector beg = ((idx - halfdur) * sf).astype(int) end = ((idx + halfdur) * sf).astype(int) beg[beg < 0] = 0 end[end > last] = last # Alternatively, to cut off incomplete windows (comment the 2 lines above) # mask = ~((beg < 0) | (end > last)) # beg, end = beg[mask], end[mask] t = np.column_stack((beg, end)).mean(1) / sf if method == 'mean': def func(x): return np.mean(x) elif method == 'min': def func(x): return np.min(x) elif method == 'max': def func(x): return np.max(x) elif method == 'ptp': def func(x): return np.ptp(x) elif method == 'prop_above_zero': def func(x): return np.count_nonzero(x >= 0) / x.size elif method == 'slope': def func(x): times = np.arange(x.size, dtype=np.float64) / sf return _slope_lstsq(times, x) elif method == 'covar': def func(x, y): return _covar(x, y) elif method == 'corr': def func(x, y): return _corr(x, y) else: def func(x): return _rms(x) # Now loop over successive epochs if method in ['covar', 'corr']: for i in range(idx.size): out[i] = func(x[beg[i]:end[i]], y[beg[i]:end[i]]) else: for i in range(idx.size): out[i] = func(x[beg[i]:end[i]]) # Finally interpolate if interp and step != 1 / sf: f = interp1d(t, out, kind='cubic', bounds_error=False, fill_value=0, assume_sorted=True) t = np.arange(n) / sf out = f(t) return t, out
def _zerocrossings(x): """Find indices of zero-crossings in a 1D array. Parameters ---------- x : np.array One dimensional data vector. Returns ------- idx_zc : np.array Indices of zero-crossings Examples -------- >>> import numpy as np >>> from yasa.main import _zerocrossings >>> a = np.array([4, 2, -1, -3, 1, 2, 3, -2, -5]) >>> _zerocrossings(a) array([1, 3, 6], dtype=int64) """ pos = x > 0 npos = ~pos return ((pos[:-1] & npos[1:]) | (npos[:-1] & pos[1:])).nonzero()[0]
[docs]def trimbothstd(x, cut=0.10): """ Slices off a proportion of items from both ends of an array and then compute the sample standard deviation. Slices off the passed proportion of items from both ends of the passed array (i.e., with ``cut`` = 0.1, slices leftmost 10% **and** rightmost 10% of scores). The trimmed values are the lowest and highest ones. Slices off less if proportion results in a non-integer slice index. Parameters ---------- x : np.array Input array. cut : float Proportion (in range 0-1) of total data to trim of each end. Default is 0.10, i.e. 10% lowest and 10% highest values are removed. Returns ------- trimmed_std : float Sample standard deviation of the trimmed array, calculated on the last axis. """ x = np.asarray(x) n = x.shape[-1] lowercut = int(cut * n) uppercut = n - lowercut atmp = np.partition(x, (lowercut, uppercut - 1), axis=-1) sl = slice(lowercut, uppercut) return atmp[..., sl].std(ddof=1, axis=-1)
[docs]def sliding_window(data, sf, window, step=None, axis=-1): """Calculate a sliding window of a 1D or 2D EEG signal. .. versionadded:: 0.1.7 Parameters ---------- data : numpy array The 1D or 2D EEG data. sf : float The sampling frequency of ``data``. window : int The sliding window length, in seconds. step : int The sliding window step length, in seconds. If None (default), ``step`` is set to ``window``, which results in no overlap between the sliding windows. axis : int The axis to slide over. Defaults to the last axis. Returns ------- times : numpy array Time vector, in seconds, corresponding to the START of each sliding epoch in ``strided``. strided : numpy array A matrix where row in last dimension consists of one instance of the sliding window, shape (n_epochs, ..., n_samples). Notes ----- This is a wrapper around the :py:func:`numpy.lib.stride_tricks.as_strided` function. Examples -------- With a 1-D array >>> import numpy as np >>> from yasa import sliding_window >>> data = np.arange(20) >>> times, epochs = sliding_window(data, sf=1, window=5) >>> times array([ 0., 5., 10., 15.]) >>> epochs array([[ 0, 1, 2, 3, 4], [ 5, 6, 7, 8, 9], [10, 11, 12, 13, 14], [15, 16, 17, 18, 19]]) >>> sliding_window(data, sf=1, window=5, step=1)[1] array([[ 0, 1, 2, 3, 4], [ 2, 3, 4, 5, 6], [ 4, 5, 6, 7, 8], [ 6, 7, 8, 9, 10], [ 8, 9, 10, 11, 12], [10, 11, 12, 13, 14], [12, 13, 14, 15, 16], [14, 15, 16, 17, 18]]) >>> sliding_window(data, sf=1, window=11)[1] array([[ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]]) With a N-D array >>> np.random.seed(42) >>> # 4 channels x 20 samples >>> data = np.random.randint(-100, 100, size=(4, 20)) >>> epochs = sliding_window(data, sf=1, window=10)[1] >>> epochs.shape # shape (n_epochs, n_channels, n_samples) (2, 4, 10) >>> epochs array([[[ 2, 79, -8, -86, 6, -29, 88, -80, 2, 21], [-13, 57, -63, 29, 91, 87, -80, 60, -43, -79], [-50, 7, -46, -37, 30, -50, 34, -80, -28, 66], [ -9, 10, 87, 98, 71, -93, 74, -66, -20, 63]], [[-26, -13, 16, -1, 3, 51, 30, 49, -48, -99], [-12, -52, -42, 69, 87, -86, 89, 89, 74, 89], [-83, 31, -12, -41, -87, -92, -11, -48, 29, -17], [-51, 3, 31, -99, 33, -47, 5, -97, -47, 90]]]) """ from numpy.lib.stride_tricks import as_strided assert axis <= data.ndim, "Axis value out of range." assert isinstance(sf, (int, float)), 'sf must be int or float' assert isinstance(window, (int, float)), 'window must be int or float' assert isinstance(step, (int, float, type(None))), ('step must be int, ' 'float or None.') if isinstance(sf, float): assert sf.is_integer(), 'sf must be a whole number.' sf = int(sf) assert isinstance(axis, int), 'axis must be int.' # window and step in samples instead of points window *= sf step = window if step is None else step * sf if isinstance(window, float): assert window.is_integer(), 'window * sf must be a whole number.' window = int(window) if isinstance(step, float): assert step.is_integer(), 'step * sf must be a whole number.' step = int(step) assert step >= 1, "Stepsize may not be zero or negative." assert window < data.shape[axis], ("Sliding window size may not exceed " "size of selected axis") # Define output shape shape = list(data.shape) shape[axis] = np.floor(data.shape[axis] / step - window / step + 1 ).astype(int) shape.append(window) # Calculate strides and time vector strides = list(data.strides) strides[axis] *= step strides.append(data.strides[axis]) strided = as_strided(data, shape=shape, strides=strides) t = np.arange(strided.shape[-2]) * (step / sf) # Swap axis: n_epochs, ..., n_samples if strided.ndim > 2: strided = np.rollaxis(strided, -2, 0) return t, strided
[docs]def get_centered_indices(data, idx, npts_before, npts_after): """Get a 2D array of indices in data centered around specific time points, automatically excluding indices that are outside the bounds of data. Parameters ---------- data : 1-D array_like Input data. idx : 1-D array_like Indices of events in data (e.g. peaks) npts_before : int Number of data points to include before ``idx`` npts_after : int Number of data points to include after ``idx`` Returns ------- idx_ep : 2-D array Array of indices of shape (len(idx_nomask), npts_before + npts_after + 1). Indices outside the bounds of data are removed. idx_nomask : 1-D array Indices of ``idx`` that are not masked (= valid). Examples -------- >>> import numpy as np >>> from yasa import get_centered_indices >>> np.random.seed(123) >>> data = np.random.normal(size=100).round(2) >>> idx = [1., 10., 20., 30., 50., 102] >>> before, after = 3, 2 >>> idx_ep, idx_nomask = get_centered_indices(data, idx, before, after) >>> idx_ep array([[ 7, 8, 9, 10, 11, 12], [17, 18, 19, 20, 21, 22], [27, 28, 29, 30, 31, 32], [47, 48, 49, 50, 51, 52]]) >>> data[idx_ep] array([[-0.43, 1.27, -0.87, -0.68, -0.09, 1.49], [ 2.19, 1. , 0.39, 0.74, 1.49, -0.94], [-1.43, -0.14, -0.86, -0.26, -2.8 , -1.77], [ 0.41, 0.98, 2.24, -1.29, -1.04, 1.74]]) >>> idx_nomask array([1, 2, 3, 4], dtype=int64) """ # Safety check assert isinstance(npts_before, (int, float)) assert isinstance(npts_after, (int, float)) assert float(npts_before).is_integer() assert float(npts_after).is_integer() npts_before = int(npts_before) npts_after = int(npts_after) data = np.asarray(data) idx = np.asarray(idx, dtype='int') assert idx.ndim == 1, "idx must be 1D." assert data.ndim == 1, "data must be 1D." def rng(x): """Create a range before and after a given value.""" return np.arange(x - npts_before, x + npts_after + 1, dtype='int') idx_ep = np.apply_along_axis(rng, 1, idx[..., np.newaxis]) # We drop the events for which the indices exceed data idx_ep = np.ma.mask_rows(np.ma.masked_outside(idx_ep, 0, data.shape[0])) # Indices of non-masked epochs in idx idx_ep_nomask = np.unique(idx_ep.nonzero()[0]) idx_ep = np.ma.compress_rows(idx_ep) return idx_ep, idx_ep_nomask