Cross-Session on Multiple Datasets

This example shows how to perform a cross-session analysis on two MI datasets using a CSP+LDA pipeline

The cross session evaluation context will evaluate performance using a leave one session out cross-validation. For each session in the dataset, a model is trained on every other session and performance are evaluated on the current session.

# Authors: Sylvain Chevallier <sylvain.chevallier@uvsq.fr>
#
# License: BSD (3-clause)

import warnings

import matplotlib.pyplot as plt
import seaborn as sns
from mne.decoding import CSP
from sklearn.discriminant_analysis import LinearDiscriminantAnalysis as LDA
from sklearn.pipeline import make_pipeline

import moabb
from moabb.datasets import BNCI2014_001, Zhou2016
from moabb.evaluations import CrossSessionEvaluation
from moabb.paradigms import LeftRightImagery


warnings.simplefilter(action="ignore", category=FutureWarning)
warnings.simplefilter(action="ignore", category=RuntimeWarning)
moabb.set_log_level("info")

Loading Dataset

Load 2 subjects of BNCI 2014-004 and Zhou2016 datasets, with 2 session each

subjects_list = [1, 2]

datasets = [Zhou2016(), BNCI2014_001()]

for d in datasets:
    # replace the subject list
    d.subject_list = subjects_list

Choose Paradigm

We select the paradigm MI, applying a bandpass filter (8-35 Hz) on the data and we will keep only left- and right-hand motor imagery

paradigm = LeftRightImagery(fmin=8, fmax=35)

Create Pipelines

Use the Common Spatial Patterns with 8 components and a Linear Discriminant Analysis classifier.

pipeline = {}
pipeline["CSP+LDA"] = make_pipeline(CSP(n_components=8), LDA())

Get Data (optional)

To get access to the EEG signals downloaded from the dataset, you could use dataset.get_data(subjects=[subject_id]) to obtain the EEG under an MNE format, stored in a dictionary of sessions and runs. Otherwise, paradigm.get_data(dataset=dataset, subjects=[subject_id]) allows to obtain the EEG data in sklearn format, the labels and the meta information. The data are preprocessed according to the paradigm requirements.

# X_all, labels_all, meta_all = [], [], []
# for d in datasets:
#     # sessions = d.get_data(subjects=[2])
#     X, labels, meta = paradigm.get_data(dataset=d, subjects=[2])
#     X_all.append(X)
#     labels_all.append(labels)
#     meta_all.append(meta)

Evaluation

The evaluation will return a DataFrame containing a single AUC score for each subject / session of the dataset, and for each pipeline.

overwrite = True  # set to True if we want to overwrite cached results

evaluation = CrossSessionEvaluation(
    paradigm=paradigm, datasets=datasets, suffix="examples", overwrite=overwrite
)
results = evaluation.process(pipeline)

print(results.head())

     score      time  ...  pipeline                  codecarbon_task_name
0  0.85113  0.241677  ...   CSP+LDA  80dadd6b-380c-45f6-8d61-d90fc1c68530
1  0.93640  0.256716  ...   CSP+LDA  87af0bad-31b6-40da-b1c7-f61d8e27f314
2  0.93960  0.204371  ...   CSP+LDA  15a9b178-fc36-48d3-a6a1-3df2f83d305a
3  0.90800  0.181730  ...   CSP+LDA  59114b58-552b-4fc3-b634-f9e4721e00f6
4  0.77037  0.170870  ...   CSP+LDA  f7fdeef5-1484-4c94-aceb-43291b1139b0

[5 rows x 13 columns]

Plot Results

Here we plot the results, indicating the score for each session and subject

sns.catplot(
    data=results,
    x="session",
    y="score",
    hue="subject",
    col="dataset",
    kind="bar",
    palette="viridis",
)
plt.show()