cuic0808 / eeg-analysis-public

EEG-analysis-HMM

EEG-analysis-public

This pipeline process the raw EEG data per subject in .set, .fdt, preprocess and spectral transform with Matlab script. Then conducted modeling fitting and analysis with Python scripts.

1. Matlab Preprocessing

Main script: yy_auto_process.m

1.1. run with steps to output feature data per trial by sources

includeSteps={'pre','asr','interp','rmEvt','epoch','ica','mara','stratify','gICA','wavelet'};

1.2. run with steps run with steps to output feature data per trial by channels

includeSteps={'wavelet_chan'};

2. Python Processing

2.1. Normalized power envelope by source per individual, output centered feature data (data_process.py)

2.2. Fit HMM with full mean and covariance matrix (hmm.py) from 2.1

2.3. Normalized power envelope by channel per individual, output centered channel data (data_process.py)

2.4. Infer hidden states using decoded trials (from 2.2) and the processed power envelopes by channel (from 2.3). Generate all feature mean by state.(infer_state.py).

2.5. Make topography plots for state and frequency from 2.4. (makeplots_bydata.py)

2.6. Further analysis with hmm_analysis_13.ipynb