Welcome to Polarspike - a Python library for analyzing and visualizing spike trains. Written by Marvin Seifert
The idea of Polarspike is simple and effective: to provide a simple and intuitive interface for analyzing and visualizing spike trains. At the backend, Polarspike uses Polars and Pyarrow which allows for fast processing of large datasets. On the frontend spike trains are organized in Pandas dataframes (experts in polars can stick to polars also on the frontend), which allows for easy data manipulation and plotting. Polarspike was written with the aim to be able to process the data from multiple multi-electrode array (MEA) recordings without the need to set up a "real" database running on a server. This is archived by allowing the user to load chunks of multiple recordings into memory and analyze them. This way, the user can analyze a potentially unlimited number of recordings, as long as a single stimulus/ recording/ chunk fits into memory. Plotting is done making heavy use of Datashader, this allows for fast plotting of millions of spikes. Polarspike also provides simple interfaces for plotting raster plots and for further data analysis.
Currently, Polarspike is not available on PyPI, so you have to install it from source. To do so, clone this repository
and run pip install . in the root directory of the repository. Alternatively, you can use Python Poetry to install
Polarspike. To do so, run poetry install in the root directory of the repository.
The most challenging step might be to organize your data in a way that Polarspike can read it. Polarspike was tested on datasets using Herding Spikes 2, Kilosort2 and Spyking Circus. The respective Extractors are included in the Extractors module. I am happy to help people to write their own Extractors for other spike sorting software.
On the frontend, Polarspike is organized around two dataframes: One dataframe contains information about the cells (e.g. their location, their type, etc.) and the other dataframe contains information about the stimulus (e.g. the stimulus type, the stimulus parameters, etc.). This makes Polarspike very flexible, as it allows the user to easily filter cells and stimuli based on their properties. The user can also add additional "stimuli" to the stimulus dataframe, e.g. to add a stimulus that is not part of the original recording. Suppose, in the original recording you played a stimulus that lasts 10s and repeats several times but your analysis revealed that the relevant time window is only the first 1 second. Instead of having to re-strucutre your data, you can simply add a new stimulus to the stimulus dataframe that only points to the first second of the original stimulus and load only the respective spikes into memory. (Look at the example notebooks, to get an idea of how this is very useful and flexible). If you need the original 10 seconds stimulus again, you can simply load it when needed.
Often the most important part of data analysis is to explore the data. Polarspike provides interactive features that can be used in Jupyter lab or Jupyter notebook (not sure all widgets will work). For example, you can interact with all dataframes interactively using Panel's tabulator, e.g. to filter cells or stimuli based on their properties. You can also plot spike trains interactively, e.g. to zoom in on a specific time window or to select a subset of cells using either Matplotlib widget or Plotly.
You can download an example dataset from Box. Check out the example notebooks in the examples folder.
Full documentaion is available here
I am happy to receive contributions to Polarspike. I am not a professional software developer, so I am sure there is a lot of room for improvement. If you want to contribute, please open an issue first to discuss the changes you want to make. Thank you. Happy spike-training!