A Computational toolbox for large scale Calcium Imaging data Analysis and behavioral analysis.
Recent advances in calcium imaging acquisition techniques are creating datasets of the order of Terabytes/week. Memory and computationally efficient algorithms are required to analyze in reasonable amount of time terabytes of data. This project implements a set of essential methods required in the calcium imaging movies analysis pipeline. Fast and scalable algorithms are implemented for motion correction, movie manipulation, and source and spike extraction. CaImAn also contains some routines for the analyisis of behavior from video cameras. In summary, CaImAn provides a general purpose tool to handle large movies, with special emphasis on tools for two-photon and one-photon calcium imaging and behavioral datasets.
-
Handling of very large datasets
- Memory mapping
- Parallel processing in patches
- Frame-by-frame online processing [5]
- OpenCV-based efficient movie playing and resizing
-
Motion correction [6]
- Fast parallelizable OpenCV and FFT-based motion correction of large movies
- Can be run also in online mode (i.e. one frame at a time)
- Corrects for non-rigid artifacts due to raster scanning or non-uniform brain motion
-
Source extraction
- Separates different sources based on constrained nonnegative matrix Factorization (CNMF) [1-2]
- Deals with heavily overlapping and neuropil contaminated movies
- Suitable for both 2-photon [1] and 1-photon [3] calcium imaging data
- Selection of inferred sources using a pre-trained convolutional neural network classifier
- Online processing available [5]
-
Denoising, deconvolution and spike extraction
-
Behavioral Analysis [7]
- Unsupervised algorithms based on optical flow and NMF to automatically extract motor kinetics
- Scales to large datasets by exploiting online dictionary learning
- We also developed a tool for acquiring movies at high speed with low cost equipment [Github repository].
We recently incorporated a Python implementation of the OnACID [5] algorithm, that enables processing data in an online mode and in real time. Check the script demos_detailed/demo_OnACID_mesoscope.py or the notebook demo_OnACID_mesoscope.ipynb for an application on two-photon mesoscope data provided by the Tolias lab (Baylor College of Medicine).
-
Installation on Mac (Suggested PYTHON 3.5)
- Download and install Anaconda (Python 2.7 or Python 3.5) http://docs.continuum.io/anaconda/install
git clone https://github.com/flatironinstitute/CaImAn cd CaImAn/ conda env create -f environment_mac.yml -n caiman source activate caiman (ONLY FOR PYTHON 2) conda install numpy==1.12 (ONLY FOR PYTHON 2) conda install spyder=3.1 conda install -c conda-forge tensorflow keras python setup.py build_ext -i
Some possible issues when running in parallel mode (dview is not None) because of bugs in Python/ipyparallel/numpy interaction, sometimes CaImAn hangs. In this case, we suggest to use dview = None. In the near future this should be solved, and currently the dev branch solves the issue (you need to recreate the environment using the instructions in the dev README).
-
Installation on Linux
- Download and install Anaconda (Python 2.7 or Python 3.5) http://docs.continuum.io/anaconda/install
git clone https://github.com/flatironinstitute/CaImAn cd CaImAn/ conda env create -f environment.yml -n caiman source activate caiman (ONLY FOR PYTHON 2) conda install spyder=3.1 python setup.py build_ext -i
- To make the package available from everywhere and have it working efficiently under any configuration ALWAYS run these lines before starting spyder:
export PYTHONPATH="/path/to/caiman:$PYTHONPATH" export MKL_NUM_THREADS=1 export OPENBLAS_NUM_THREADS=1
-
Installation on Windows
(Python 3)
- Download and install Anaconda (Python 3.6) http://docs.continuum.io/anaconda/install,
- GIT (https://git-scm.com/) and
- Microsoft Build Tools for Visual Studio 2017 https://www.visualstudio.com/downloads/#build-tools-for-visual-studio-2017
- reboot.
git clone https://github.com/flatironinstitute/CaImAn cd CaImAn git pullstart>programs>anaconda3>anaconda prompt
conda env create -f environment_mac.yml -n caiman activate caiman conda install -c conda-forge tensorflow keras python setup.py build_ext -i conda install numba jupyter notebook --NotebookApp.iopub_data_rate_limit=1.0e10
(Python 2.7) not supported in Windows
- Installation on Linux (Windows and MacOS are problematic with anaconda at the moment)
- create a new environment (suggested for safety) and follow the instructions for the calcium imaging installation
- Install spams, as explained here. Installation is not straightforward and it might take some trials to get it right
-
Notebooks : The notebooks provide a simple and friendly way to get into CaImAn and understand its main characteristics.
-
you can find them in directly in CaImAn folder and launch them from your ipython Notebook application:
-
to launch jupyter notebook :
source activate CaImAn conda launch jupyter (if errors on plotting use this instead) jupyter notebook --NotebookApp.iopub_data_rate_limit=1.0e10
-
-
demo files are also found in the demos_detailed subfolder. We suggest trying demo_pipeline.py first as it contains most of the tasks required by calcium imaging. For behavior use demo_behavior.py
- /!\ if you want to directly launch the python files, your python console still must be in the CaImAn directory.
- All diffs must be tested before asking for a pull request. Call 'nosetests' program from inside of your CaImAn folder to look for errors. For python3 on MacOS nosetests does not work properly. If you need to test, then type the following from within the CaImAn folder:
cd caiman/tests
ls test_*py | while read t; do nosetests --nologcapture ${t%%.py}; done;- This test will run the entire CaImAn program and look for differences against the original one. If your changes have made significant differences you'll be able to be recognise regressions by this test.
- Andrea Giovannucci, Flatiron Institute, Simons Foundation
- Eftychios A. Pnevmatikakis, Flatiron Institute, Simons Foundation
- Johannes Friedrich, Flatiron Institute, Simons Foundation
- Erick, Cobos, Baylor College of Medicine
- Valentina Staneva, University of Washington
- Ben Deverett, Princeton University
- Jérémie Kalfon, University of Kent, ECE paris
A complete list of contributors can be found here.
The following references provide the theoretical background and original code for the included methods.
[1] Pnevmatikakis, E.A., Soudry, D., Gao, Y., Machado, T., Merel, J., ... & Paninski, L. (2016). Simultaneous denoising, deconvolution, and demixing of calcium imaging data. Neuron 89(2):285-299, [paper], [Github repository].
[2] Pnevmatikakis, E.A., Gao, Y., Soudry, D., Pfau, D., Lacefield, C., ... & Paninski, L. (2014). A structured matrix factorization framework for large scale calcium imaging data analysis. arXiv preprint arXiv:1409.2903. [paper].
[3] Zhou, P., Resendez, S. L., Stuber, G. D., Kass, R. E., & Paninski, L. (2016). Efficient and accurate extraction of in vivo calcium signals from microendoscopic video data. arXiv preprint arXiv:1605.07266. [paper], [Github repository].
[4] Friedrich J. and Paninski L. Fast active set methods for online spike inference from calcium imaging. NIPS, 29:1984-1992, 2016. [paper], [Github repository].
[5] Giovannucci, A., Friedrich J., Kaufman M., Churchland A., Chklovskii D., Paninski L., & Pnevmatikakis E.A. (2017). OnACID: Online analysis of calcium imaging data in real data. NIPS 2017, to appear. [paper]
[6] Pnevmatikakis, E.A., and Giovannucci A. (2017). NoRMCorre: An online algorithm for piecewise rigid motion correction of calcium imaging data. Journal of Neuroscience Methods, 291:83-92 [paper], [Github repository].
[7] Giovannucci, A., Pnevmatikakis, E. A., Deverett, B., Pereira, T., Fondriest, J., Brady, M. J., ... & Masip, D. (2017). Automated gesture tracking in head-fixed mice. Journal of Neuroscience Methods, in press. [paper].
The implementation of this package is developed in parallel with a MATLAB toobox, which can be found here.
Some tools that are currently available in Matlab but have been ported to CaImAn are
Python 3 and spyder If spyder crashes on MacOS run
brew install --upgrade openssl
brew unlink openssl && brew link openssl --force
SCS:
If you get errors compiling scs when installing cvxpy you probably need to create a link to openblas or libgfortran in /usr/local/lib/, for instance:
sudo ln -s /Library/Frameworks/R.framework/Libraries/libgfortran.3.dylib /usr/local/lib/libgfortran.2.dylib
Debian fortran compiler problems: If you get the error gcc: error trying to exec 'cc1plus': execvp: No such file or directory in Ubuntu run or issues related to SCS type
sudo apt-get install g++ libatlas-base-dev gfortran libopenblas-dev
conda install openblas atlas
If there are still issues try
export LD_LIBRARY_PATH=/path_to_your_home/anaconda2/lib/
If that does not help, try
conda install atlas (only Ubuntu)
pip install 'tifffile>=0.7'
conda install accelerate
conda install openblas
The code uses the following libraries
- NumPy
- SciPy
- Matplotlib
- Scikit-Learn
- ipyparallel for parallel processing
- opencv for efficient image manipulation and visualization
- Tifffile For reading tiff files. Other choices can work there too.
- cvxpy for solving optimization problems (for deconvolution, optional)
- Spams for online dictionary learning (for behavioral analysis, optional)
For the constrained deconvolution method (deconvolution.constrained_foopsi) various solvers can be used, some of which require additional packages:
'cvxpy': (default) For this option, the following packages are needed:
'cvx': For this option, the following packages are needed:
In general 'cvxpy' can be faster, when using the 'ECOS' or 'SCS' sovlers, which are included with the CVXPY installation. Note that these dependencies are circumvented by using the OASIS algoritm for deconvolution.
Documentation of the code can be found here. Moreover, our wiki page covers some aspects of the code.
Special thanks to the following people for letting us use their datasets for our various demo files:
- Weijian Yang, Darcy Peterka, Rafael Yuste, Columbia University
- Sue Ann Koay, David Tank, Princeton University
- Manolis Froudarakis, Jake Reimers, Andreas Tolias, Baylor College of Medicine
If you use this code please cite the corresponding papers where original methods appeared (see References above), as well as the following abstract:
Giovannucci, A., Friedrich, J., Deverett, B., Staneva, V., Chklovskii, D., & Pnevmatikakis, E. (2017). CaImAn: An open source toolbox for large scale calcium imaging data analysis on standalone machines. Cosyne Abstracts.
Please use the gitter chat room for questions and comments and create an issue for any bugs you might encounter.
This program is free software; you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation; either version 2 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.
You should have received a copy of the GNU General Public License along with this program. If not, see http://www.gnu.org/licenses/.