Skip to content

samuelgarcia/spikeinterface-gui

 
 

Repository files navigation

spikeinterface-gui

GUI for spikeinterface objects without data copy.

This is a cross platform interactive viewer to inspect the final results and quality of any spike sorter supported by spikeinterface (kilosort, spykingcircus, tridesclous, mountainssort, yass, ironclust, herdingspikes, hdsort, klusta...)

This interactive GUI offer several views that dynamically refresh other views. This allows us to very quickly check the strengths and weaknesses of any sorter output.

Contrary to other viewers (like phy), this viewer skips the tedious and long step of copying and reformatting the entire dataset (filtered signal + waveform + PCA) to a particular format or folder organisation. This gui is built on top of spikeinterface objects (Recording, Sorting, WaveformExtractor) These objects are "lazy" and retrieve data on the fly (no copy!). And contrary to phy, this is a view only tool : no manual curation at the moment (split/merge/trash have to be done outside).

This viewer internally use Qt (with PySide6, PyQT6 or PyQt5) and pyqtgraph. And so, this viewer is a local desktop app (old school!!). There is a web based viewer work-in-progress here.

screenshot

Launch

In order to use this viewer you will need to know a bit of spikeinterface

Step 1 : extract waveforms

You first need to "extract waveform" with spikeinterface See help here

Note that:

  • not all waveform snippets are extracted (See max_spikes_per_unit) only some of them
  • this step is cached to a folder (and can be reloaded)
  • this step can be run in parallel (and so is quite fast)
  • optionally PCA can be computed and displayed

Example:

import spikeinterface.full as si
recording = si.read_XXXX('/path/to/my/recording')
recording_filtered = si.bandpass_filter(recording)
sorting = si.run_sorter('YYYYY', recording_filtered)

# extract waveforms 
# sparsity is important because it makes everything faster!!!
waveform_folder = '/path/for/my/waveforms'
job_kwargs = dict(n_jobs=10, chunk_duration='1s', progress_bar=True,)
we = si.extract_waveforms(
    recording_filtered, sorting, waveform_folder,
    max_spikes_per_unit=500,
    ms_before=1.5, ms_after=2.5,
    sparse=True,
    **job_kwargs
)
# compute the noise level a faster opening in sigui
si.compute_noise_levels(we)

# optionally compute more stuff using the spikeinterface.postprocessing module
# principal components, template similarity, spike amplitudes
# This will enable us to display more views
si.compute_principal_components(we,
    n_components=3,
    mode='by_channel_local',
    whiten=True)
si.compute_template_similarity(we,  method='cosine_similarity')
si.compute_spike_amplitudes(we, **job_kwargs)

Step 2 : open the GUI

With python:

import spikeinterface_gui
# This creates a Qt app
app = spikeinterface_gui.mkQApp() 
# reload the waveform folder
we = si.WaveformExtractor.load_from_folder(waveform_folder)
# create the mainwindow and show
win = spikeinterface_gui.MainWindow(we)
win.show()
# run the main Qt6 loop
app.exec_()

With the command line

sigui /path/for/my/waveforms

Install

For beginners or Anaconda users please see our installation tips where we provide a yaml for Mac/Windows/Linux to help properly install spikeinterface and spikeinterface-gui for you in a dedicated conda environment.

Otherwise,

You need first to install one of these 3 packages (by order of preference):

  • pip install PySide6 or
  • pip install PyQt6 or
  • pip install PyQt5

From pypi:

pip install spikeinterface-gui

From source:

git clone https://github.com/SpikeInterface/spikeinterface-gui.git
cd spikeinterface-gui
pip install .

Author

Samuel Garcia, CNRS, Lyon, France

This work is a port of the old tridesclous.gui submodule on top of spikeinterface.

About

GUI for spikeinterface objects

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • Python 100.0%