CNN-ripple plugin for Open Ephys

CNN-ripple is a 1D convolutional neural network (CNN) operating over high-density LFP recordings to detect hippocampal SWR both offline and online. It works with recordings from several types of recordings (e.g. linear arrays, high-density probes, ultradense Neuropixels).

This plugin allows the use of CNN-ripple as a real-time detection tool as part of the Open Ephys platform.

You can check our results in our manuscript (Navas-Olive, Amaducci et al. eLife 2022)

Installing the plugin

Download the latest release for your operating system from https://github.com/PridaLab/CNNRippleDetectorOEPlugin/releases.

For Linux:

1. Copy the CNNRippleDetectorOEPlugin.so file to Open Ephys plugins folder.
2. Download Tensorflow 2.3.0 libraries into CNNRippleDetectorOEPlugin/libs/bin/x64. You need to extract the content of the lib folder from this TAR file.

For Windows (default path for plugins: C:\ProgramData\Open Ephys):

1. Copy the CNNRippleDetectorOEPlugin.dll file to Open Ephys plugins folder.
2. Download Tensorflow 2.3.0 libraries into Open Ephys shared folder. You need to extract the .dll file from the lib folder from this ZIP file.

Running Open Ephys with the plugin

For Linux:

1. Open a terminal and load Tensorflow dynamic libraries using

export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:<your_path>/CNNRippleDetectorOEPlugin/libs/bin/x64/

2. Run Open Ephys on the same terminal.

For Windows:

1. Run Open Ephys.

Plugin configuration

The plugin receives 8 channels as input. You can use the ChannelMap plugin to select the channels.

• File: selector for the CNN model .pb file. Can be found in the CNNRippleDetectorOEPlugin/model directory.
• Pulse duration: duration of the TTL pulse sent when a ripple is detected (in milliseconds).
• Timeout: recovery time after a pulse is sent (in milliseconds).
• Calibration: calibration time before the experiment to setup the signals normalization (in seconds). One minute is usually enough.
• Threshold: probability threshold for the detections. Between 0 and 1.
• Drift: number of standard deviations above which the signal is considered to be dominated by extreme offset drift and the CNN will not predict.
• Output: output channel for TTL pulses.

Compiling the plugin from source

1. Clone this repository in the same directory where plugin-GUI is located.
2. Download Tensorflow 2.3.0 libraries into CNNRippleDetectorOEPlugin/libs/bin/x64. You need to extract the content of the lib folder from this TAR file (for Linux) or this ZIP file (for Windows).
3. Go to the Build directory and execute to create the CMake project:

For Linux:

cmake -G "Unix Makefiles" -DCMAKE_BUILD_TYPE=Release  ..

For Windows:

cmake -G "Visual Studio 16 2019" -A x64 -DCMAKE_CXX_FLAGS="/EHsc /utf-8" ..

4. To compile and install the plugin:

On Linux, in the same Build directory:

make
make install

On Windows, open Visual Studio 16 2019 and compile and install the solution (see notes below on how to install Visual Studio).

5. Install and run the plugin as indicated above.

How to install Visual Studio 2019 and Open Ephys from source

1. Clone Open Ephys GUI repository.

2. Download and install Visual Studio 2019.

3. Run Resources/DLLs/FrontPanelUSB-DriverOnly-4.4.0.exe to install DAQ drivers if needed.

4. Install C++ compiling tools for Visual Studio 2019 and create a new project so everything gets properly configured.

5. Download and install CMake. Select "Add CMake to the system PATH".

6. Add the line add_definitions(-DWIN32) into line 107 of CMakeLists.txt.

7. Open PowerShell, go to folder plugin-GUI/Build and run CMake.

cmake -G "Visual Studio 16 2019" -A x64 -DCMAKE_CXX_FLAGS="/EHsc /utf-8" ..

-A indicates the target architecture, which must be specified accordingly.

8. Open open-ephys-GUI.sln, which can be found in the Build directory, with Visual Studio 2019.

9. Compile the solution.