eeg-diagnosis

This code is released with NME and uses braindecode to test generalization of CNN on EEG anomaly detection as discussed in The NME Scalp EEG Dataset: An Open-Source Annotated Dataset of Healthy and Pathological EEG Recordings for Predictive Modeling

Acknowledgment

The shallow and Deep CNN experiments are built on example provided for BrainDecode by ‪Robin Tibor Schirrmeister here https://github.com/robintibor/auto-eeg-diagnosis-example

Also lstm implementation contain pieces of code from the following package: Kunal Patel et al: https://github.com/kunalpatel1793/Neural-Nets-Final-Project

Requirements

1. Depends on https://robintibor.github.io/braindecode/
2. This code was programmed in Python 3.6 (might work for other versions also).

Run

1. Modify config.py, especially correct data folders for your path..
2. Run with python ./auto_diagnosis.py
3. auto_diagnosis.py defines and train CNN models
4. diagnosis.py gives features from trained deep CNN
5. hybrid_lstm.py trains an LSTM model to classify sequence of features