mne_denoise.dss.variants.narrowband_scan#

mne_denoise.dss.variants.narrowband_scan(data: ndarray, sfreq: float, *, freq_range: tuple[float, float] = (1, 40), freq_step: float = 1.0, bandwidth: float = 2.0, n_components: int = 1, **dss_kws) → tuple[DSS, ndarray, ndarray][source]#

Scan frequencies to find optimal narrowband DSS components.

Sweeps through a frequency range, computing DSS at each frequency. Returns the fitted DSS at the best frequency, along with the full eigenvalue spectrum for visualization.

Parameters:

  • data (ndarray, shape (n_channels, n_times) or (n_channels, n_times, n_epochs)) – Input data.

  • sfreq (float) – Sampling frequency in Hz.

  • freq_range (tuple of float) – (min_freq, max_freq) range to scan. Default (1, 40).

  • freq_step (float) – Frequency step size in Hz. Default 1.0.

  • bandwidth (float) – Bandwidth of bandpass filter at each frequency. Default 2.0.

  • n_components (int) – Number of DSS components to compute at each frequency. Default 1.

  • **dss_kws – Additional keyword arguments passed to DSS.

Returns:

  • best_dss (DSS) – Fitted DSS at the frequency with highest eigenvalue.

  • frequencies (ndarray, shape (n_freqs,)) – Frequencies that were scanned.

  • eigenvalues (ndarray, shape (n_freqs,)) – First eigenvalue at each frequency.

Examples

>>> # Find dominant alpha frequency
>>> best_dss, freqs, eigs = narrowband_scan(data, sfreq=250, freq_range=(7, 14))
>>> print(f"Peak alpha at {freqs[np.argmax(eigs)]:.1f} Hz")
>>> alpha_sources = best_dss.transform(data)

>>> # Plot eigenvalue spectrum
>>> import matplotlib.pyplot as plt
>>> plt.plot(freqs, eigs)
>>> plt.xlabel("Frequency (Hz)")
>>> plt.ylabel("DSS Eigenvalue")