Minimal Example

This section provides an overview of a minimal example pipeline setup processing EEG data using the MNE-BIDS package.

It demonstrates

1. More complex pipeline steps.

2. Using the Pipeline_MNE_BIDS_Data class to handle MNE-BIDS data.

3. Running the pipeline (interactively) from a jupyter notebook.

4. Using Sphinx doc to document the pipeline steps.

Folder Structure

The minimal example folder contains the following files:

📂 examples/minimal
├── 📂 steps/                 # Steps
│   ├── 00_conversion.py
│   ├── 01_preprocessing.py
│   ├── 02_continue.py
│   ├── __init__.py           # Marks the directory as a Python package
├── 📂 doc/                   # Sphinx documentation of steps.
│   ├── index.rst
│   └── ...
├── config.py                 # Configuration file for the pipeline
├── Makefile
└── run.ipynb                 # Jupyter Notebook for running the pipeline

(cf. Example Code at Github)

The basic setup is similar to the trivial example with a bit of processing logic added. We will not go into details here, but focus on the data class that is used to pass data between the steps.

Pipeline_MNE_BIDS_Data

The Pipeline_MNE_BIDS_Data is essentially a book keeping device to manage

a dictionary of the form

{
    "1001": {
        "session1": {
            "task1": {
                "01": DATA_RUN_1,
                "02": DATA_RUN_2,
                # ...
            }
        }
    }
}

where DATA_RUN_1, DATA_RUN_2, etc. are paths to the files or data objects.

In many cases we would like to apply the same processing to all (or a certain subset of) runs in the data class. This can be achieved using the apply() method.

For the sake of example, this is demonstrated below.

from lw_pipeline import Pipeline_MNE_BIDS_Data, Config

data = Pipeline_MNE_BIDS_Data(Config())

data.file_paths = {
    "1001": {
        "session1": {
            "task1": {
                "1": 5,
            }
        }
    }
}

# Define a named function to multiply by 5
def multiply_by_5(source, x):
    source = source * 5
    return source

data.apply(multiply_by_5)

print(str(data))

Output

Step multiplyby5 took 0.00 seconds.
PipelineData object handling the following files:
| Subject 1001
|--- Session session1
|----- Task task1
|------- Run 1: 25

A common use case would be to apply a function that is a method in a Pipline_Step subclass.

Note

The second argument x in the defined function is provided a BidsPath object that represents a possibly generated derivative from the applied function.

Note

If the function applied returns an MNE raw object, an MNE epochs object or MNE Annotations the data class will attempt to save the object at the proposed BidsPath location.