This repository contains training and testing scripts of the signal to signal network model on spikefinder challenge dataset.

Usage

Tested with keras on tensorflow background (on TF1 and TF2) and python 3.8.

1. install python dependencies

   pip install -r requirements.txt
   

2. run trainingscript and save the model in model_dir

   python train_s2s.py model_dir
   

*Optionally set the parameters in config file: config.py

3. evaluate the model stored in model_dir

   python test_s2s.py model_dir/modelname.h5
   

Alternatively, the test script can be edited to predict the spikes given a calcium signal. For this, write a dataset loading function for the given calcium signal in and run the test script until predictions.

If you find this work useful, please cite the following publication:

@article{sebastian2021signal,
title={Signal-to-signal neural networks for improved spike estimation from calcium imaging data},
author={Sebastian, Jilt and Sur, Mriganka and Murthy, Hema A and Magimai-Doss, Mathew},
journal={PLoS Computational Biology},
volume={17},
number={3},
pages={e1007921},
year={2021},
publisher={Public Library of Science San Francisco, CA USA}
}