Epileptic EEG classification using Embedded Dynamic Mode Decomposition

Author: Jennifer Hellar
Date: Apr 26, 2022

Datasets

CHB-MIT scalp EEG
(https://physionet.org/content/chbmit/1.0.0/)
Kaggle American Epilepsy Society Seizure Prediction Challenge intracranial EEG
(https://www.kaggle.com/c/seizure-prediction)

Key scripts:

const.py
Dataset constants, key EmDMD and simulation parameters, and functions to obtain directory paths, effective sampling frequencies, and preictal/interictal labelling rules.

emdmd.py
Embedded Dynamic Mode Decomposition (EmDMD) library.

util.py
Common utility library.

Process entire datasets:

extract_all_seg.py [-h] --dataset DATASET
Extracts preictal/interictal segments from all patients in the input dataset.

extract_all_feat.py [-h] --dataset DATASET
Processes all patients from the input dataset to compute EmDMD features.

classify_patient_specific.py [-h] --dataset DATASET --sph SPH --sop SOP [-v VERBOSE]
Classifies each patient in the input dataset and prints CV results.

Process individual patients:

extract_seg.py [-h] --dataset DATASET --patient PATIENT
Extracts and downsamples preictal and interictal segments for the input patient.

extract_feat.py [-h] --dataset DATASET --patient PATIENT [--class CLASS] [--index INDEX]
Computes EmDMD features from the preictal/interictal segment(s) of the input patient.

split_kfolds.py [-h] --dataset DATASET [--patient PATIENT] --sph SPH --sop SOP
Creates training/test cross-validation splits on the computed EmDMD features of the input dataset.

classify.py [-h] --dataset DATASET [-f FOLD]
Classifies the test data of the input dataset to measure performance.

Other scripts:

freq_tb.py
Computes the EmDMD frequencies present in subsets of the data (to detemine r parameter needed).

make_figures.py
Creates the figures present in the paper and provides other misc simulations.