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, SortingAnalyzer) These objects are "lazy" and retrieve data on the fly (no copy!).
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 here.
The main idea is make visible one or several unit and visualy inspect if they should be merge or remove. For this visibility:
- ctlr + double click on a unit in probeview
- check the box visible in the unitlist
- double click on one unit in unitlist unit visible alone
- move one of the roi in the probeview
Views can be reorganized by moving docks by clicking in the title bar of a docks. Any dock (view) can be closed. And can be put back with right click in any title bar of any dock.
Every view has a ? button which open the contextual help. Theses inplace docs are the most important stuff to be read. (but the contains typos)
When some units are visible, the related spike list can be refresh. Then selecting spike per spike can also refersh some views. This enable a very quick and convinient spike per spike jump on traces.
Channel visibility can be handled with one of the roi in the probeview.
By default this tools is a viewer only. But you can turn it into a tools for manual curation using,
the curation=True
option.
This tools supoort the curation format from spikeinterface.
This format enbale to:
- remove units
- merge units
- create manual labels
When this mode is activated a new view is added on top left to maintain the list of removal and merges. The curation format can be exported to json.
The actual main
branch is using the new SortingAnalyzer
object from spikeinterface, so you need at least version 0.101.0 of
spikeinterface and be familiar with the SortingAnalyzer
concept.
In order to use this viewer you will need to know a bit of spikeinterface
You first need to is to get a SortingAnalyzer
object with spikeinterface.
See help here
Note that:
- some extensions are mandatory (unit_location, templates, )
- some extension are optional
- the more extensions are computed the more view are 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)
job_kwargs = dict(n_jobs=-1, progress_bar=True, chunk_duration="1s")
# make the SortingAnalyzer with necessary and some optional extensions
sorting_analyzer = si.create_sorting_analyzer(sorting, recording,
format="binary_folder", folder="/my_sorting_analyzer",
**job_kwargs)
sorting_analyzer.compute("random_spikes", method="uniform", max_spikes_per_unit=500)
sorting_analyzer.compute("waveforms", **job_kwargs)
sorting_analyzer.compute("templates", **job_kwargs)
sorting_analyzer.compute("noise_levels")
sorting_analyzer.compute("unit_locations", method="monopolar_triangulation")
sorting_analyzer.compute("isi_histograms")
sorting_analyzer.compute("correlograms", window_ms=100, bin_ms=5.)
sorting_analyzer.compute("principal_components", n_components=3, mode='by_channel_global', whiten=True, **job_kwargs)
sorting_analyzer.compute("quality_metrics", metric_names=["snr", "firing_rate"])
sorting_analyzer.compute("template_similarity")
sorting_analyzer.compute("spike_amplitudes", **job_kwargs)
With python:
import spikeinterface_gui
# This creates a Qt app
app = spikeinterface_gui.mkQApp()
# reload the SortingAnalyzer
sorting_analyzer = si.load_sorting_analyzer("/my_sorting_analyzer")
# create the mainwindow and show
win = spikeinterface_gui.MainWindow(sorting_analyzer)
win.show()
# run the main Qt6 loop
app.exec_()
Or simpler:
import spikeinterface.widgets as sw
sorting_analyzer = load_sorting_analyzer(test_folder / "sorting_analyzer")
sw.plot_sorting_summary(sorting_analyzer, backend="spikeinterface_gui")
With the command line
sigui /path/for/my/sorting_analyzer
The command line support some otions like --notraces or --curation
sigui --no-traces --curation /path/for/my/sorting_analyzer
To open the viewer with curation mode use curation=True
.
This mode is pretty new and was implemented under kind inducement of friends.
I hope that this could be a fair replacement of phy
.
import spikeinterface_gui
app = spikeinterface_gui.mkQApp()
win = spikeinterface_gui.MainWindow(sorting_analyzer, curation=True)
win.show()
app.exec_()
from spikeinterface.widgets import plot_sorting_summary
sw.plot_sorting_summary(sorting_analyzer, curation=True, backend="spikeinterface_gui")
The curation_dict
can be saved inside the folder of the analyzer (for "binary_folder" or "zarr" format).
Then it is auto-reloaded when the gui is re-opened.
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
orpip install PyQt6
orpip install PyQt5
From pypi:
pip install spikeinterface-gui
From source:
git clone https://github.com/SpikeInterface/spikeinterface-gui.git
cd spikeinterface-gui
pip install .
Samuel Garcia, CNRS, Lyon, France
This work is a port of the old tridesclous.gui
submodule on top of
spikeinterface.
Contrary, to the spikeinterface package, for the developement of this viewer all good practices of coding are deliberately put aside : no test, no CI, no auto formating, no doc, ... Feel free to contribute, it is an open wild zone. Code anarchist are very welcome. So in this mess, persona non grata : pre-commit, black, pytest fixture, ...