Welcome

SpikeSort is a flexible spike sorting framework implemented completely in Python. It features manual and automatic clustering, many data formats and it is memory-efficient. Truely Open Source, BSD-licensed.

Eavesdropping on neurons in action

Spike sorting is a common pre-processing step for the analysis of neuroscientific data. The goal of the procedure is to detect the times at which neurons generated action potential, also called spikes.

The times at which spikes arrive encode information about sensory stimuli and their perception. Therefore, correct identification of occurrences of spikes elicited by different neurons is necessary to understand the function and disfunction of nervous system.

Spike sorting done right

Spike sorting is based on extracellular recordings of small electric potential associated with spikes by microelectrodes placed very close to the cell. In practice, such electrodes pick up activity of multiple neurons active simultaneously. The goal of spike sorting is to discriminate the responses of indvidual cells from others.

Each cell has its own characteristic spike waveshape, so that cells can be easily differentiated just by comparing their spikes. Spikes of similar shapes and amplitudes are grouped together using automatic or manual clustering procedure.

If you want to know more read an in-depth article on spike sorting in Scholarpedia or simply go directly to the tutorials in the documentation of SpikeSort and start sorting.

Start sorting in matter of minutes

Run included Python script:

$ export DATAPATH='where/your/data/is/located'
$ python -i cluster_beans.py

and you can start sorting:

>>> browser.show()                         # sort and show results
>>> cluster = base.features['LabelSource']
>>> cluster.recluster(1, 'gmm',2 )         # split cell 1 into 2 clusters
>>> cluster.delete_cells(2, 3)             # delete unwanted cells

Features

• user-friendly and customisable,
• interactive command-line interface in Python,
• graphical-user interface and visualization widgets,
• k-means and gaussian mixture models clustering algorithms and manual cluster cutting,
• support for multi-channel data (for example, from tetrodes),
• support for binary datasets and HDF5 files (support for other formats can be easily implemented),
• based on established scientific libraries: NumPy, PyTables, matplotlib, scikit-learn

License

SpikeSort is free and Open Source. Its liberal license (BSD) allows for non-commercial and commercial use.

Contribute

Found a bug? Have a good idea for improving SpikeSort? Need help to analyse your data? Head over to github page and create a new ticket or fork. If you just want to chat with developers, send us an e-mail.