RippleNet

Description

This repository contains files for RippleNet, a recurrent neural network with Long Short-Term Memory (LSTM) layers for detecting sharp-wave ripples in single-channel LFP signals measured in hippocampus CA1.

Author: Espen Hagen (https://github.com/espenhgn)

LICENSE: https://github.com/CINPLA/RippleNet/blob/master/LICENSE

Citation

RippleNet and its application is now described in a peer-reviewed publication, which can be cited as:

RippleNet: A Recurrent Neural Network for Sharp Wave Ripple (SPW-R) Detection Hagen, E., Chambers, A.R., Einevoll, G.T. et al. RippleNet: a Recurrent Neural Network for Sharp Wave Ripple (SPW-R) Detection. Neuroinform (2021). https://doi.org/10.1007/s12021-020-09496-2

BitTex format:

@Article{Hagen_2021,
  author    = {Espen Hagen and Anna R. Chambers and Gaute T. Einevoll and Klas H. Pettersen and Rune Enger and Alexander J. Stasik},
  journal   = {Neuroinformatics},
  title     = {{RippleNet}: a Recurrent Neural Network for Sharp Wave Ripple ({SPW}-R) Detection},
  year      = {2021},
  month     = {jan},
  doi       = {10.1007/s12021-020-09496-2},
  publisher = {Springer Science and Business Media {LLC}},
}

The older preprint can be cited as:

RippleNet: A Recurrent Neural Network for Sharp Wave Ripple (SPW-R) Detection
Espen Hagen, Anna R. Chambers, Gaute T. Einevoll, Klas H. Pettersen, Rune Enger, Alexander J. Stasik
bioRxiv 2020.05.11.087874; doi: https://doi.org/10.1101/2020.05.11.087874

BibTex format:

@article {Hagen2020.05.11.087874,
	author = {Hagen, Espen and Chambers, Anna R. and Einevoll, Gaute T. and Pettersen, Klas H. and Enger, Rune and Stasik, Alexander J.},
	title = {RippleNet: A Recurrent Neural Network for Sharp Wave Ripple (SPW-R) Detection},
	elocation-id = {2020.05.11.087874},
	year = {2020},
	doi = {10.1101/2020.05.11.087874},
	publisher = {Cold Spring Harbor Laboratory},
	abstract = {Hippocampal sharp wave ripples (SPW-R) have been identified as key bio-markers of important brain functions such as memory consolidation and decision making. SPW-R detection typically relies on hand-crafted feature extraction, and laborious manual curation is often required. In this multidisciplinary study, we propose a novel, self-improving artificial intelligence (AI) method in the form of deep Recurrent Neural Networks (RNN) with Long Short-Term memory (LSTM) layers that can learn features of SPW-R events from raw, labeled input data. The algorithm is trained using supervised learning on hand-curated data sets with SPW-R events. The input to the algorithm is the local field potential (LFP), the low- frequency part of extracellularly recorded electric potentials from the CA1 region of the hippocampus. The output prediction can be interpreted as the time-varying probability of SPW-R events for the duration of the input. A simple thresholding applied to the output probabilities is found to identify times of events with high precision. The reference implementation of the algorithm, named {\textquoteright}RippleNet{\textquoteright}, is open source, freely available, and implemented using a common open-source framework for neural networks (tensorflow.keras) and can be easily incorporated into existing data analysis workflows for processing experimental data.Competing Interest StatementThe authors have declared no competing interest.},
	URL = {https://www.biorxiv.org/content/early/2020/05/12/2020.05.11.087874},
	eprint = {https://www.biorxiv.org/content/early/2020/05/12/2020.05.11.087874.full.pdf},
	journal = {bioRxiv}
}

Clone

These codes can be downloaded using git (www.git-scm.com):

cd <Repositories> # whatever download destination
git clone https://github.com/CINPLA/RippleNet
cd RippleNet

Some binary files like .h5 and .pkl may be tracked using Git LFS (https://git-lfs.github.com)

dependencies

python>=3
numpy
scipy
ipympls
matplotlib
h5py
pandas
seaborn
notebook
jupyter
tensorflow>=2.0
tensorflow-gpu (optional)

Dependencies can be installed in your existing Python environment using the requirements.txt file and the pip utility:

pip install -r requirements.txt

To install an Anaconda Python (www.anaconda.com) environment with these dependencies, issue

conda env create -f environment.yml
conda activate ripplenet

This will not install tensorflow-gpu for hardware acceleration by default.

Binder

You may mess around with the RippleNet notebooks on MyBinder.org:

Retraining networks is not recommended (no GPU access)!

Files and folders:

README.md: This file
LICENSE: License file
environment.yml: Conda environment file
RippleNet_training_bidirectional.ipynb: Jupyter notebook for training bidirectional RippleNet
RippleNet_training_unidirectional.ipynb: Notebook for training unidirectional RippleNet
RippleNet_manuscript_figures.ipynb: Notebook for generating figures 2-7 in Hagen E. et al. (2020)
RippleNet_timeseries_prediction.ipynb: Notebook for generating figures 8-11 in Hagen E. et al. (2020)
RippleNet_interactive_prototype.ipynb: Notebook with user-interactive detection and rejection of ripple events
trained_networks/
- ripplenet_*directional_random_seed*.h5: trained RippleNet instances of uni- or bidirectional types
- ripplenet_*directional_best_random_seed*.h5: best-performing model on validation set during training
- ripplenet_*directional_history_random_seed*.csv: training history (.csv format)
- ripplenet_*directional_history_random_seed*.pkl: training history (.pickle format)
ripplenet/
- common.py: shared methods and functions
- models.py: function declarations for tensorflow.keras models
data/
- train_00.h5: Training data set (mouse)
- train_tingley_00.h5: Training data set (rat)
- validation_00.h5: Validation data set (mouse)
- validation_00.h5: Validation data set (rat)
- test_00.h5: Test data set (mouse)
- m4029_session1.h5: Test data set (mouse, continuous)

CINPLA / RippleNet