/Neural_Decoding_from_Calcium_Imaging_Data

Primary LanguageJupyter NotebookMIT LicenseMIT

Neural_Decoding_from_Calcium_Imaging_Data

1. 50 ROIs Input (GRU, Keras)

Demo Code: https://github.com/atakehiro/Neural_Decoding_from_Calcium_Imaging_Data/blob/main/1.RNN_Keras/1.Basic_Code/Keras_GRU_Behavior_Analysis.ipynb

2. Image Sequence Input (CNN-GRU, fastai)

Demo Code

  1. Npz file creation: https://github.com/atakehiro/Neural_Decoding_from_Calcium_Imaging_Data/blob/main/2.CNN-RNN_PyTorch/Behavior_Analysis_3_channel_image_to_npz_data.ipynb

  2. Mean and std calculation for normalization: https://github.com/atakehiro/Neural_Decoding_from_Calcium_Imaging_Data/blob/main/2.CNN-RNN_PyTorch/Behavior_Analysis_Calculate_Mean_and_Std.ipynb

  3. Stage 1: https://github.com/atakehiro/Neural_Decoding_from_Calcium_Imaging_Data/blob/main/2.CNN-RNN_PyTorch/Fastai_Behavior_Analysis_Stage1_Efficientnet-B0.ipynb

  4. Stage 2: https://github.com/atakehiro/Neural_Decoding_from_Calcium_Imaging_Data/blob/main/2.CNN-RNN_PyTorch/Fastai_Behavior_Analysis_Stage2_Efficientnet-B0-GRU-nonFix.ipynb

Environment

OS: Ubuntu 16.04 or 18.04

GPU: NVIDIA GeForce RTX2080 or RTX3090

Programming Language: Python 3.6

Software: Anaconda

Dataset

The dataset is available in Zenodo.

https://zenodo.org/records/10408334

Zenodo DOI: 10.5281/zenodo.10408334

Publication

Ajioka T, Nakai N, Yamashita O, Takumi T (2024) End-to-end deep learning approach to mouse behavior classification from cortex-wide calcium imaging. PLoS Comput Biol 20(3): e1011074. https://doi.org/10.1371/journal.pcbi.1011074

Author

Takehiro Ajioka

E-mail: tajioka@m.u-tokyo.ac.jp

Affiliation

Department of Physiology, Kobe University School of Medicine