Brain Signal Anaylsis Projects

This repository contains groups of scripts written as a part of fNIRS, EEG classification research for the state-of-the-art deep learning application.

(1) Columbia Data Science Institute (DSI) Conference

"Brain State Classification on Functional Near Infrared Spectroscopy using Convolutional Neural Networks"
Andrew Lee, Jim (Euiyoung) Chung, Xiaofu He

The main project in progress. Our lab is planning finalize the paper soon and the source code will be updated after the paper is published.

(2) EEG Classification for Motor Imagery Tasks using CNN and LSTM

Overview

TensorFlow and Keras implementation of Zhang et al(2018), "EEG-based Intention Recognition from Spatio-Temporal Representations via Cascade and Parallel Convolutional Recurrent Neural Networks" for EEG motar imagery classification on PhysioNet data (https://www.physionet.org/pn4/eegmmidb/). Stacked CNN and RNN were applied on time-distributed sliding windows of raw EEG data.

Acamdemic Reference

directory	title	author	year
BCI4	Filter Bank Common Spatial Pattern (FBCSP) in Brain-Computer Interface	Kai Keng Ang, et al.	2008
BCI4	Filter bank common spatial pattern algorithm on BCI competition IV Datasets 2a and 2b	Kai Keng Ang, et al.	2012
BCI4	Feature Selection for Motor Imagery EEG Classification Based on Firefly Algorithm and Learning Automata	Aiming Liu, et al.	2017
deepLearning	Learning Representations from EEG with Deep Recurrent-Convolutional Neural Networks	Pouya Bashivan, et al.	2016
deepLearning	EEG-Based Emotion Recognition Using Deep Learning Network with Principal Component Based Covariate Shift Adaptation	Suwicha Jirayucharoensak, et al.	2014
deepLearning	A Deep Learning Architecture for Temporal Sleep Stage Classification Using Multivariate and Multimodal Time Series	Stanislas Chambon, et al.	2018
deepLearning	Deep Learning With Convolutional Neural Networks for EEG Decoding and Visualization	Robin T. Schirrmeister, et al.	2017
deepLearning	EEG-based Intention Recognition from Spatio-Temporal Representations via Cascade and Parallel Convolutional Recurrent Neural Networks	Dalin Zhang, et al.	2018
deepLearning	A Deep Learning Method for Classification of EEG Data Based on Motor Imagery	Xiu An, et al.	2014
emotionState	Classifying Different Emotional States by Means of EEG- Based Functional Connectivity Patterns	You-Yun Lee, Shulan Hsieh	2014
emotionState	Emotion Classification Based on Gamma-band EEG	Mu Li, Bao-Liang Lu	2009
preprocessing	Time-series discrimination using feature relevance analysis in motor imagery classification	A.M. Alvarez-Meza, et al.	2014
preprocessing	Comparison of signal decomposition methods in classification of EEG signals for motor-imagery BCI system	Jasmin Kevric, et al.	2015
preprocessing	Classification of EEG Motor imagery multi class signals based on Cross Correlation	D.Hari Krishna, et al.	2016
preprocessing	EEG Signal Processing Techniques For Mental Task Classification	Rajveer Shastri, et al.	2015

Description of files

• All code is written in python3
• Dependencies include MNE (EEG handling package), Keras, requests, urllib3
• Python scripts are adjusted to be run on HPC

Non-Python files:

filename	description
README.md	Text file (markdown format) description of the project.

Python Scripts:

filename	description
eeg_main.py	Import EEG data from data files and start training the graph (added for HPC clusters)
eeg_tensorflow.ipynb	TensorFlow and Keras implementation. Please refer to this notebook for a quick look

Python Modules (HPC version):

filename	description
eeg_preprocessing.py	Preprocessing steps to implement spatio-temporal pattern recognition on raw eeg data. Based on Scikit-learn and MNE pacakges.
eeg_import.py	Functions defined to extract data from .edf file format using MNE package.
eeg_data_downloads.py	Executing this code will generate folders and start downloading PhysioNet data into them.
eeg_eval.py	Evaluation of the model based on the history file. - calls the confusion matrix, loss, and accuracy.
gpu_training.sh	Send the eeg_main.py script to the computational nodes. Enter 'sbatch gpu_trainin.sh' on Habanero HPC.