BadChannelRemovalNIRSSCI¶
Remove channels with low scalp coupling correlation from NIRS signals.
This node uses between-channel correlations to identify bad channels. This method uses optical densities as inputs and this requires exactly 2 wavelengths per channel to calculate the correlation coefficient between the two. If there are any channels with unmatched wavelength pairs, those channels will be assigned a correlation coefficient of 0, in which case they will be removed (or marked NaN) from the output data. Before computing SCI (correlation coefficients) the optical densities, this node first applies a band-pass filter (FIR) to avoid systematic factors such as cardiac and Mayer waves. Changing the filter settings may greatly impact the resulting correlations. This node can also output the computed SCI values for each channel, as a signal quality diagnostic. The SCI metric is based on Pollonini et. al. 2016. More Info... Version 0.2.0
Ports/Properties¶
data¶
Data to process.
diagnostics¶
Diagnostic information indicating the per-channel error measure.
threshold¶
Minimum scalp coupling index threshold per channel. This is a value between 0 and 1 (correlation coefficient). Both channels wavelengths in which the optodes are below threshold are removed.
action¶
Action to perform on bad channels. Remove means the channels will be dropped form the data, and mark_nan means that the data of bad channels will be replaced with NaN values (indicating missing data to downstream nodes that can handle it).
axis¶
Axis along which to filter. Rarely anything other than 'time' is used here.
order¶
Filter order. If unspecified, this will be auto-determined based on the tightness of the transition bands and the desired stopband attenuation (using the Kaiser window rule). Note that high orders (e.g., 1000 or greater) will be fairly heavy on the CPU, especially with many channels.
frequencies¶
Transition frequencies. For a lowpass/highpass filter, you can either give the cutoff frequency, or two frequencies to determine the rolloff curve. For a bandpass/bandstop filter, you can give two cutoff frequencies, or 4 frequencies to determine the rolloff curve. The syntax for 'freeform' filters is [(frequency1, gain1), (frequency2, gain2), ...], and you can use the special frequency value -1 as a stand-in for the highest frequency at which a filter can be designed for the given sampling rate (the signal's Nyquist frequency).
mode¶
Filter mode. Low/highpass lets low/high frequencies through, bandpass lets frequencies in a limited band through, and bandstop (a.k.a. notch) removes frequency in a limited band. The freeform mode (rarely used) can design arbitrary filters by specifying frequencies in the spectrum and corresponding desired gain factors.
minimum_phase¶
Design minimum-phase filter. If enabled, a filter with minimum signal delay will be designed, which will distort the signal phases, especially in the transition region. If disabled, a linear-phase filter will be designed, which will not distort the signal at all, but which incurs a delay of half the filter order in samples.
antisymmetric¶
Design antisymmetric filter. This designs rarely-used / special-purpose filters for signal differentiation or 90 degree phase shifting. Make sure you specify the order explicitly and choose and odd number if you use this.
stop_atten¶
Minimum attenuation in stopband. This is the minimum acceptable attenuation, in dB, in the stopband, which is ideally infinitely suppressed, but in practice 30-80 dB are enough, depending on the amplitudes of the signals to attenuate.
convolution_method¶
FIRFilter coefficients can be convolved with the input signal using multiple methods. The standard method performs time-domain convolution. For long signals (>214 samples) using filters with a modest number of coefficients (>24), it can be faster to do the convolution in the frequency domain using the overlap-add method.
cut_preringing¶
Cut off the pre-ringing portion generated by the filter. This is currently only supported in linear-phase mode and offline.
direction¶
Filtering direction. This describes how information is propagated during the filter process. Forward means that an event (e.g., a spike) will result in a ripple that follows the event. Forward-backward means that the ripple will propagate in both directions, and will be symmetric around the event, but this results in effects preceding their causes, which can be problematic when performing causality analysis and is not replicable in real time. Gap-aware filtering additionally respects data boundaries (e.g., where bad time windows have been cut out) and does not propagate information across them.
verbose¶
Print progress including list of channels removed.
set_breakpoint¶
Set a breakpoint on this node. If this is enabled, your debugger (if one is attached) will trigger a breakpoint.
metadata¶
User-definable meta-data associated with the node. Usually reserved for technical purposes.