def test_plot_bursts(tsig_burst):
times = create_times(N_SECONDS, FS)
bursts = detect_bursts_dual_threshold(tsig_burst, FS, (0.75, 1.5), F_RANGE)
plot_bursts(times,
tsig_burst,
bursts,
save_fig=True,
file_path=TEST_PLOTS_PATH,
file_name='test_plot_bursts.png')
def detect_bursts_df_amp(df, sig, fs, f_range, amp_threshes=(1, 2),
n_cycles_min=3, filter_kwargs=None):
"""Determine which cycles in a signal are part of an oscillatory
burst using an amplitude thresholding approach.
Parameters
----------
df : pandas DataFrame
Dataframe of waveform features for individual cycles, trough-centered.
sig : 1d array
Time series.
fs : float
Sampling rate, Hz.
f_range : tuple of (float, float)
Frequency range (Hz) for oscillator of interest.
amp_threshes : tuple (low, high), optional, default: (1, 2)
Threshold values for determining timing of bursts.
These values are in units of amplitude (or power, if specified) normalized to
the median amplitude (value 1).
n_cycles_min : int, optional, default: 3
Minimum number of cycles to be identified as truly oscillating needed in a row in
order for them to remain identified as truly oscillating.
filter_kwargs : dict, optional
Keyword arguments to :func:`~neurodsp.filt.filter.filter_signal`.
Returns
-------
df : pandas DataFrame
Same df as input, with an additional column to indicate
if the cycle is part of an oscillatory burst.
"""
# Detect bursts using the dual amplitude threshold approach
sig_burst = detect_bursts_dual_threshold(sig, fs, amp_threshes, f_range,
min_n_cycles=n_cycles_min, **filter_kwargs)
# Compute fraction of each cycle that's bursting
burst_fracs = []
for _, row in df.iterrows():
fraction_bursting = np.mean(sig_burst[int(row['sample_last_trough']):
int(row['sample_next_trough'] + 1)])
burst_fracs.append(fraction_bursting)
# Determine cycles that are defined as bursting throughout the whole cycle
df['is_burst'] = [frac == 1 for frac in burst_fracs]
df = _min_consecutive_cycles(df, n_cycles_min=n_cycles_min)
df['is_burst'] = df['is_burst'].astype(bool)
return df
# the lower amplitude threshold. # # **Other Parameters** # # - `avg_type`: used to set the average for normalization to either 'median' or 'mean' # - `magnitude_type`: used to set the metric for thresholding, to 'amplitude' or 'power' # ################################################################################################### # Settings for the dual threshold algorithm amp_dual_thresh = (1, 2) f_range = (8, 12) # Detect bursts using dual threshold algorithm bursting = detect_bursts_dual_threshold(sig, fs, amp_dual_thresh, f_range) ################################################################################################### # # You can plot detected bursts using # :func:`~.plot_bursts`. # ################################################################################################### # Plot original signal and burst activity plot_bursts(times, sig, bursting, labels=['Simulated EEG', 'Detected Burst']) ################################################################################################### # # The graph above shows the bursting activity in red.
# --------------- # # Now that we have a sense of the main rhythmicity of this piece of data, lets # explore using NeuroDSP to investigate whether this rhythm is bursty. # ################################################################################################### # Burst settings amp_dual_thresh = (1., 1.5) f_range = (peak_cf - 2, peak_cf + 2) ################################################################################################### # Detect bursts of high amplitude oscillations in the extracted signal bursting = detect_bursts_dual_threshold(sig, fs, amp_dual_thresh, f_range) ################################################################################################### # Plot original signal and burst activity plot_bursts(times, sig, bursting, labels=['Raw Data', 'Detected Bursts']) ################################################################################################### # Measure Rhythmicity with Lagged Coherence # ----------------------------------------- # # So far, in an example channel of data, we have explored some rhythmic properties of # the EEG data. We did so by finding the main frequency of rhythmic power, and checking # this frequency for bursting. # # Next, we can try applying similar analyses across all channels, to measure
def get_burst_epochs(data, fs, f_range, amp_thresh_dual):
bursting = burst.detect_bursts_dual_threshold(data, fs, f_range, amp_thresh_dual)
bursting_stats = burst.compute_burst_stats(bursting, fs)
return bursting, bursting_stats
# (default: median) amplitude of the whole time series. Two thresholds are # defined based off of this normalized amplitude. In order for a burst to be # detected, the amplitude must cross the higher amplitude threshold. The burst # lasts until the amplitude then falls below the lower amplitude threshold. # # **Other Parameters:** # * The average for normalization can be set to either the mean or median # by modifying the `average_method` keyword argument. # * Power can be used instead of amplitude by modifying the `magnitude_type` # keyword argument. # # Detect bursts using 'deviation' algorithm amp_dual_thresh = (1, 2) f_range = (8, 12) bursting = detect_bursts_dual_threshold(sig, fs, f_range, amp_dual_thresh) # Plot original signal and burst activity plot_bursts(times, sig, bursting, labels=['Simulated EEG', 'Detected Burst']) ################################################################################################### # # The graph above shows the bursting activity in red. The algorithm was # used with thresh=(1, 2), so any time point with more than 3 times the # median magnitude in the alpha range (8-12 Hz) was marked as bursting activity. ################################################################################################### # # Burst detection applied to real recordings # ------------------------------------------
