dmalt/speech_meg

DVC repository for MEG overt/covert speech dataset.

It contains conda environment setup and scripts for data preprocessing with MNE-Python, DVC configuration to download the data from GDrive and Python API to load the data directly into a Python project.

Quickstart

To only load the data

1. Clone this repo:

git clone https://github.com/dmalt/speech_meg.git

2. Install DVC and DVC-gdrive:

• with pip:

pip install dvc dvc[gdrive]

• with conda:

conda install -c conda-forge dvc dvc-gdrive

3. From the project root run

dvc pull

4. Complete the authentification step.

At this point DVC will ask for an authentification with your Google account. Follow the link in the terminal. In the opened browser window select the Google account with which the data were shared and click on both checkboxes. If the data were shared with you, the download should start after the authentification.

5. Come back next morning :)

In case of success, the following data will be loaded (18 GB in total):

• raw MEG and audio data @ rawdata,
• data annotations @ rawdata/derivatives/011-annotate_premaxfilt, rawdata/derivatives/031-annotate_postmaxfilt, rawdata/derivatives/032-annotate_speech rawdata/derivatives/033-annotate_covert, rawdata/derivatives/071-annotate_muscles, rawdata/derivatives/101-merge_annotations
• manually marked bad ICA components @ rawdata/derivatives/051-inspect_ica
• aligned audio data @ rawdata/derivatives/081-align_audio
• downsampled and ICA-cleaned MEG data @ rawdata/derivatives/091-resample

All the intermediate files will not be downloaded since they can be recomputed via running the corresponding scripts from rawdata/code/preproc