🧠 In-browser detection of seizure activity from single-channel LFP/EEG brain recordings using deep learning.
Try it online: https://deep-seizure-detect.herokuapp.com
-
Provide the app with a CSV file containing rodent EEG data, where each row is a sequence of 500 samples at 100 Hz. ℹ️ More info - 📄 Example
-
The app will read the file and send it by chunks to our server, where our machine learning algorithm runs. Nothing is stored on our end.
-
Once the whole file has been processed, the app generates a dynamic visualization allowing verification, editing and export of the results.
The model is a convolutional neural net that was built using Keras API with a Tensorflow-backend. It was trained on LFP data from chronically epileptic mice that were generated using intra-hippocampal kainate injections by Dr. Trina Basu.
Built with the Maguire Lab at Tufts University by:
| Matteo Cargnelutti | Pantelis Antonoudiou, PhD |
|---|---|
 |
 |
| Software Development | Data Science |
| Boston, USA / France 🇫🇷 | Boston, USA / Cyprus 🇨🇾 |
| @matteocargnelutti | @pantelisantonoudiou |
This open-source software is distributed under the Apache 2.0 License.
This software was built and made available for research purposes only and is intended for use on rodent data.
"How Deep Learning Solved My Seizure Detection Problems". Commentary by Pantelis Antonoudiou and Jamie Maguire on Epilepsy Currents, Sept 2 2020.
Many thanks to Dr. Trina Basu for allowing us to use some of her data to build the model and test the application.
Please contact us via the issues section.