Try out spikeinterfaces pipeline for continuous wideband .ncs data

[mvdm]

As for issue #4, but for wideband data.

• [x] Test if we can get this to produce output at all. Yarik has a basic pipeline up and running for .ntt data here: https://github.com/vandermeerlab/data-formats/blob/master/yoh_try_spikeinterfaces.ipynb
• [x] Use a recording site layout file
• [ ] Compare to MClust sorted output. @jgmaz recorded a wideband file in parallel with .ntt files if I understand correctly, any chance you could prioritize spike sorting such a session?

[manimoh]

I got the basic pipeline working for the wideband data using klusta-kwik sorter. See this notebook for more details.

[manimoh]

As a next step, will try to set up a pipeline to run Kilosort on this wideband data.

[manimoh]

Found a problem with the spikesorter implementation of KS2. Opened relevant issue here

[manimoh]

First pass pipeline available here!!

[manimoh]

A missing feature in this whole pipeline is the absence of a way to programmatically exclude certain NCS files (For example: channels that are sampled for LFP purposes) from the spikesorting process. Spikeinterface will throw an error if any NCS file in the folder has a different sampling rate and certain other properties.

This can be fixed by making changes in this file of the neo code base, since spikeinterfaces leverages that for Neuralynx files: https://github.com/NeuralEnsemble/python-neo/blob/master/examples/read_files_neo_rawio.py

I have opened an issue requesting a feature: https://github.com/NeuralEnsemble/python-neo/issues/889

But until then I will fork the python-neo repo and make changes in that, so that people in the lab can use the pipeline more comfortably.

[manimoh]

Update: This working example has been updated to elucidate File exclusion based on changes in a fork of the official python-neo repo.

[manimoh]

Update: https://github.com/vandermeerlab/data-formats/issues/5#issuecomment-754996599