fer-rplazas/feature-extraction-methods

Benchmarking feature extraction methods for electrophysiologically-based brain state detection

The Potential of Convolutional Neural Networks for Identifying Neural States based on Electrophysiological Signals: Code and Figures

This repository accompanies the manuscript

The potential of convolutional neural networks for identifying neural states based on electrophysiological signals: experiments on synthetic and real patient data

Usage:

1. Installing packages & configuring environment:

The python (>=3.10) dependencies are listed in requirements_gpu.yml (a CUDA-capable system is required). We recommend installing them through anaconda or mamba:

conda env create -n <<environment_name>> --file requirements_gpu.yml 

conda activate <<environment_name>>

2. Downloading patient data:

Patient data can be downloaded at https://data.mrc.ox.ac.uk/lfp-et-dbs (DOI: 10.5287/bodleian:ZVNyvrw7R, creating a free account is required). This data should be placed into the patient_data folder.

3. Running code:

The repository is separated into a nbs folder and a code folder:

• The nbs folder includes notebooks with examples and the code used to generate all figures, which can be found here.

• The code folder implements synthetic data generation as well as model implementation and training. It includes two main files (main_patients.py and main_synthetic.py), which train models for patient data and synthetic data tasks respectively. Running these scripts takes some time (a total of over two weeks on a machine with two NVIDIA RTX-3090 GPUs). The scripts take in command-line arguments:

  Example: python main_patients.py --n_jobs 4 --accelerator gpu

  To see all options: python main_synthetic.py --help

  The script outputs are csv files containing performance metrics.

---

License:

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

© The University of Oxford 2019 - 2023