search

PeyracheLab / miniscoPy

A package to analyse calcium imaging data recorded with the Miniscope.
GNU General Public License v2.0
32 stars 9 forks source link

readme

miniscoPy

(This package is not maintained anymore. Compatibility issues between python package might arise.)

A package to analyse calcium imaging data recorded with the Miniscope.

Yet, the architecture of this package is almost the same as the CaImAn package. If needed, it is thus very easy to insert CaImAn functions. Besides, the visualization tools of miniscoPy have been kept at its bare minimum purposefully.

Overview

To take a glimpse at the package, you can look at two Jupyter notebooks :

Getting Started

Prerequisites

This installation protocol has been tested with a freshly installed Ubuntu 18.04 LTS. It is not guaranted if you have conflicting packages already installed. In this case, it may work better within a Miniconda environment (indications coming soon). First start with cloning and installing packages:

git clone https://github.com/peyrachelab/miniscoPy
cd miniscoPy
sudo apt-get install python3-pip

Most packages can be installed in one single command line.

pip3 install --user numpy scipy matplotlib jupyter pandas cython scikit-learn scikit-image sklearn h5py opencv-python tqdm future psutil

The most troublesome to install is pyav. This package allows for the fast loading of the avi files generated by the Miniscope recording software. According to the PyAV installation protocol and for Ubuntu >= 14.04 LTS, dependencies can be installed with:

sudo apt-get install -y python-dev pkg-config

sudo apt-get install -y libavformat-dev libavcodec-dev libavdevice-dev libavutil-dev libswscale-dev libavresample-dev libavfilter-dev

pip3 install av

If it does not work, the best strategy is to start a conda environment and do:

conda install av -c conda-forge

If it still doens't work (and you cannot stand Conda environment and have already recompiled a linux kernel), the ultimate strategy is to compile the lib in a Conda environment and then to link it by hand.

Brian2 and xgboost are additional packages required to run the Jupyter notebook using simulated data, they are easy to install:

pip3 install --user brian2 xgboost

Installing

To install miniscoPy:

python3 setup.py build_ext -i
python3 setup.py install --user

Building and testing on Windows (experimental)

In order to build miniscoPy on Windows you will need Microsoft Visual Studio 2015 or 2017 installed (Community Edition is OK) and Miniconda(preferably) or Anaconda.

First make sure you have conda-forge channel added to your conda installation:

conda config --show channels

If not - add the conda-forge channel by executing:

conda config --add channels conda-forge

Then create new conda environment and activate it:

conda create -n miniscoPy
activate miniscoPy

Install minimal set of packages required for data processing (without IPython and Jupyter):

conda install av scipy pandas cython scikit-learn scikit-image h5py opencv tqdm future pyyaml psutil

CD to a directory where files of this repository located and build C extention:

python setup.py build_ext -i

Check your environment by executing all test scripts located in the "testbench" directory:

python -m unittest discover testbench

Finally, process sample data:

python main_cnmf_e.py

Deployment

CNMF-E has a lot of free parameters and two recording sessions from the same animal may be best analyzed with slightly different parameters. Thus, to keep track of the parameter set for each session, miniscoPy looks for a parameter file in the recording folder in yaml text format. An example of such file can be found in _miniscoPy/example_movies/parameters.yaml_.

The yaml file is loaded as a dictionary at the beginning of an analysis in a Python session:

import yaml
parameters = yaml.load(open('/example_movies/parameters.yaml', 'r'))

The Miniscope recording software saves several avi files in the recording folder. Before starting the analysis pipeline, it is required to list all the avi files in a recording folder:

import glob
files = glob.glob('/example_movies/*.avi')

Then, the first step is the motion correction algorithm :

from miniscopy.base.motion_correction import *
data, video_info = normcorre(files, None, parameters['motion_correction'])

The second step is CNMF-E :

from miniscopy import CNMFE
cnm = CNMFE(data, parameters['cnmfe'])
cnm.fit()

A minimal working example can be found in main_cnmf_e.py. Another example with generated data can be found in main_test_hd_generated.py.

Built With

Besides, codes from several toolboxes are used :

Version

Troubleshooting

Send us an email or post an Issue.

Author

License

This project is licensed under the GNU License - see the LICENSE.md file for details.

Acknowledgments