MOABB Dataset Example

In this example, we show how to fetch and prepare a MOABB dataset for usage with Braindecode.

# Authors: Lukas Gemein <l.gemein@gmail.com>
#          Hubert Banville <hubert.jbanville@gmail.com>
#          Simon Brandt <simonbrandt@protonmail.com>
#
# License: BSD (3-clause)

from collections import OrderedDict

import matplotlib.pyplot as plt
from IPython.display import display

from braindecode.datasets import MOABBDataset
from braindecode.datautil.windowers import \
    create_windows_from_events, create_fixed_length_windows
from braindecode.datautil.preprocess import preprocess, MNEPreproc

First, we create a dataset based on BCIC IV 2a fetched with MOABB,

ds = MOABBDataset(dataset_name="BNCI2014001", subject_ids=[1])

Out:

48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]

ds has a pandas DataFrame with additional description of its internal datasets

display(ds.description)

Out:

    subject  ...    run
0         1  ...  run_0
1         1  ...  run_1
2         1  ...  run_2
3         1  ...  run_3
4         1  ...  run_4
5         1  ...  run_5
6         1  ...  run_0
7         1  ...  run_1
8         1  ...  run_2
9         1  ...  run_3
10        1  ...  run_4
11        1  ...  run_5

[12 rows x 3 columns]

We can iterate through ds which yields one time point of a continuous signal x, and a target y (which can be None if targets are not defined for the entire continuous signal).

for x, y in ds:
    print(x.shape, y)
    break

Out:

(26, 1) None

We can apply preprocessing transforms that are defined in mne and work in-place, such as resampling, bandpass filtering, or electrode selection.

transforms = [
    MNEPreproc("pick_types", eeg=True, meg=False, stim=True),
    MNEPreproc("resample", sfreq=100),
]
print(ds.datasets[0].raw.info["sfreq"])
preprocess(ds, transforms)
print(ds.datasets[0].raw.info["sfreq"])

Out:

250.0
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
48 events found
Event IDs: [1 2 3 4]
100.0

We can easily split ds based on a criteria applied to the description DataFrame:

subsets = ds.split("session")
print({subset_name: len(subset) for subset_name, subset in subsets.items()})

Out:

{'session_E': 232164, 'session_T': 232164}

Next, we use a windower to extract events from the dataset based on events:

windows_ds = create_windows_from_events(
    ds, trial_start_offset_samples=0, trial_stop_offset_samples=100,
    window_size_samples=400, window_stride_samples=100,
    drop_last_window=False)

Out:

Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped
Used Annotations descriptions: ['feet', 'left_hand', 'right_hand', 'tongue']
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
96 matching events found
No baseline correction applied
0 projection items activated
Loading data for 96 events and 400 original time points ...
0 bad epochs dropped

We can iterate through the windows_ds which yields a window x, a target y, and window_ind (which itself contains i_window_in_trial, i_start_in_trial, and i_stop_in_trial, which are required for combining window predictions in the scorer).

for x, y, window_ind in windows_ds:
    print(x.shape, y, window_ind)
    break

Out:

Loading data for 1 events and 400 original time points ...
(23, 400) 3 [0, 300, 700]

We visually inspect the windows:

max_i = 2
fig, ax_arr = plt.subplots(1, max_i + 1, figsize=((max_i + 1) * 7, 5),
                           sharex=True, sharey=True)
for i, (x, y, window_ind) in enumerate(windows_ds):
    ax_arr[i].plot(x.T)
    ax_arr[i].set_ylim(-0.0002, 0.0002)
    ax_arr[i].set_title(f"label={y}")
    if i == max_i:
        break

Out:

Loading data for 1 events and 400 original time points ...
Loading data for 1 events and 400 original time points ...
Loading data for 1 events and 400 original time points ...

Alternatively, we can create evenly spaced (“sliding”) windows using a different windower.

sliding_windows_ds = create_fixed_length_windows(
    ds, start_offset_samples=0, stop_offset_samples=0,
    window_size_samples=1200, window_stride_samples=1000,
    drop_last_window=False)

print(len(sliding_windows_ds))
for x, y, window_ind in sliding_windows_ds:
    print(x.shape, y, window_ind)
    break

Out:

/home/circleci/project/braindecode/datautil/windowers.py:245: UserWarning: Meaning of `trial_stop_offset_samples`=0 has changed, use `None` to indicate end of trial/recording. Using `None`.
  'Meaning of `trial_stop_offset_samples`=0 has changed, use `None` '
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
Adding metadata with 4 columns
Replacing existing metadata with 4 columns
39 matching events found
No baseline correction applied
0 projection items activated
Loading data for 39 events and 1200 original time points ...
0 bad epochs dropped
468
Loading data for 1 events and 1200 original time points ...
(23, 1200) -1 [0, 0, 1200]

Transforms can also be applied on windows in the same way as shown above on continuous data:

def crop_windows(windows, start_offset_samples, stop_offset_samples):
    fs = windows.info["sfreq"]
    windows.crop(tmin=start_offset_samples / fs, tmax=stop_offset_samples / fs,
                 include_tmax=False)

epochs_transform_list = [
    MNEPreproc("pick_types", eeg=True, meg=False, stim=False),
    MNEPreproc(crop_windows, start_offset_samples=100, stop_offset_samples=900),
]

print(windows_ds.datasets[0].windows.info["ch_names"],
      len(windows_ds.datasets[0].windows.times))
preprocess(windows_ds, epochs_transform_list)
print(windows_ds.datasets[0].windows.info["ch_names"],
      len(windows_ds.datasets[0].windows.times))

max_i = 2
fig, ax_arr = plt.subplots(1, max_i+1, figsize=((max_i+1)*7, 5),
                           sharex=True, sharey=True)
for i, (x, y, window_ind) in enumerate(windows_ds):
    ax_arr[i].plot(x.T)
    ax_arr[i].set_ylim(-0.0002, 0.0002)
    ax_arr[i].set_title(f"label={y}")
    if i == max_i:
        break

Out:

['Fz', 'FC3', 'FC1', 'FCz', 'FC2', 'FC4', 'C5', 'C3', 'C1', 'Cz', 'C2', 'C4', 'C6', 'CP3', 'CP1', 'CPz', 'CP2', 'CP4', 'P1', 'Pz', 'P2', 'POz', 'stim'] 400
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
Loading data for 96 events and 400 original time points ...
/home/circleci/project/examples/plot_dataset_example.py:105: RuntimeWarning: tmax is not in epochs time interval. tmax is set to epochs.tmax
  include_tmax=False)
['Fz', 'FC3', 'FC1', 'FCz', 'FC2', 'FC4', 'C5', 'C3', 'C1', 'Cz', 'C2', 'C4', 'C6', 'CP3', 'CP1', 'CPz', 'CP2', 'CP4', 'P1', 'Pz', 'P2', 'POz'] 299

Again, we can easily split windows_ds based on some criteria in the description DataFrame:

subsets = windows_ds.split("session")
print({subset_name: len(subset) for subset_name, subset in subsets.items()})

Out:

{'session_E': 576, 'session_T': 576}

Estimated memory usage: 402 MB