Toolkit module

The toolkit module includes tools to process SI objects throughout your analysis.


The toolkit.preprocessing sub-module includes preprocessing steps to apply before spike sorting. Preprocessors are lazy, meaning that no computation is performed until it is required (usually at the spike sorting step). This enables one to build preprocessing chains to be applied in sequence to a RecordingExtractor object. This is possible because each preprocessing step returns a new RecordingExtractor that can be input to the next step in the chain.

In this code example, we build a preprocessing chain with 2 steps:

  1. bandpass filter

  2. common median reference (CMR)

import spikeinterface.toolkit as st

# recording is a RecordingEctractor object
recording_f = st.bandpass_filter(recording, freq_min=300, freq_max=6000)
recording_cmr = st.common_reference(recording_f, operator="median")

After preprocessing, we can optionally save the processed recording with the efficient SI save() function:

recording_saved ="preprocessed", n_jobs=8, total_memory="2G")

In this case, the save() function will process in parallel our original recording with the bandpass filter and CMR and save it to a binary file in the “preprocessed” folder. The recording_saved is yet another RecordignExtractor which maps directly to the newly created binary file, for very quick access.


After spike sorting, we can use the toolkit.postprocessing sub-module to further post-process the spike sorting output. Most of the post-processing functions require a WaveformExtractor as input. Available postprocessing tools are:

  • compute principal component scores

  • compute template similarity

  • compute template waveform metrics

  • get amplitudes for each spikes

  • compute auto- and cross-correlogram

Quality Metrics

Quality metrics allows to quantitatively assess to goodness of a spike sorting output. The toolkit.qualitymetrics sub-module includes functions to compute a large variety of available metrics (‘sort’ - spike times based; ‘rec+sort’ - based on waveforms; ‘pc’ - based on PC scores):

  • firing rate (sort)

  • ISI violation ratio (sort)

  • presence_ratio (sort)

  • amplitude_cutoff (rec+sort)

  • snr (rec+sort)

  • isolation_distance (pc)

  • l_ratio (pc)

  • d_prime (pc)

  • nearest_neighbor (pc)

For more information about quality metrics, check out this excellent documentation from the Allen Institute.