The aim of this project is to investigate how different cross-validation schemes influence decoding accuracy on MEG data in source space given the variablitity within the data, as well as investigate the differences between cross-decoding in sensor- and source space.
The data used for this project contains of MEG data from one participant being subject to visual stimuli. The data was recorded over 7 different days with the same stimuli. The stimuli were either animate or inanimate.
The preprocessing of the MEG data is done using MNE-python.
The decoding is done using linear discriminant analysis (LDA). It relies on the scikit-learn library.
├── README.md <- The top-level README for this project.
├── decoding
│ ├── accuracies <- Directory for saving decoding accuracies
│ │ ├── accuracies_LDA_prop.npy
│ │ └── ...
│ ├── plots <- Directory for saving plots
│ ├── cross_decoding.py <- Script running the cross decoding
│ ├── decoder_animacy.py <- Decoder class used for within session decoding
│ ├── decoder_cross.py <- Decoder class used for cross decoding
│ ├── decoding.py <- Script running the within session decoding
│ ├── decoding.py <- Script generating plots of decoding accuracy
│ └── statistics.py <- statistical analysis of decoding accuracy
├── ERF_analysis <- Scripts and information used for ERF analysis
│ ├── plots <- Directory for saving plots
│ └── erf.py <- Generate plots of ERFs and saves the standard deviation of the ERFs needed for the decoding analysis
├── preprocessing <- Scripts for preprocessing of the data
│ ├── check_ica.ipynb <- Plotting of the ICA components
│ └── run_ica.ipynb <- Running ICA on the data
├── source_reconstruction <- Scripts and files used for source reconstruction
│ ├── epochs_2_source_space.py
│ ├── hpi_mri.mat <- Mat file containing the MRI positions of the HPI
│ └── source_space.py <- Setting up source space and BEM
├── subset_data
│ ├── data <- Directory for saving the subset data
│ └── prep_data.py <- Script preparing data subset in source and sensor space
├── event_session_info.py <- Creates event_ids.txt and session_info.py
├── event_ids.txt <- Mapping of the stimuli to the triggers
└── session_info.txt <- Bad channels, ICA noise components, etc. per session
| Do | File | Notes |
|---|---|---|
| Identify bad channels, tmin and tmax | preprocessing/check_raw.ipynb |
Add the channels, tmin and tmax to event_session_info.py. Remember to run the script after changing the values. |
| Run ICA | preprocessing/run_ica.py |
|
| Identify noise components and create epochs | preprocessing/check_ica.ipynb |
Add noise components to event_session_info.py |
| Source reconstruction | source_reconstruction/epochs_2_source_space.py |
| Do | File | Notes |
|---|---|---|
| Prepare data for analysis | subset_data/prep_data.py |
|
| Generate plots of ERFs and save standard deviations | ERF_analysis/erf.py |
| Do | File | Notes |
|---|---|---|
| Prepare data for decoding | subset_data/prep_data.py |
Is not needed if you already have the subset data from the ERF analysis, as both analyses use the same subset data |
| Within session decoding | decoding/decoding_source.py |
|
| Cross decoding | decoding/decoding_across_sessions.py |
Use flag to indicate whether you want to do it in source- or sensor space |
| Statistics | decoding/statistics.py |
|
| Generate plots of decoding accuracy | decoding/plots.py |
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent