Note
Go to the end to download the full example code.
Time-Resolved Decoding with SlidingEstimator#
This example shows how to perform time-resolved decoding of EEG signals using
mne.decoding.SlidingEstimator. Instead of reducing the entire trial to
a single score, a SlidingEstimator fits an independent classifier at each time
point, revealing when during a trial the neural signal carries information
about the mental state.
This approach is a natural alternative to pseudo-online evaluation (using overlapping windows): rather than simulating an online scenario by slicing the raw signal with a sliding window, we directly assess decoding accuracy at each sample of the already-epoched trial.
We use the BNCI2014-001 motor-imagery dataset (left- vs right-hand) and apply
a logistic-regression classifier wrapped in a SlidingEstimator. For each
subject the score is evaluated via stratified 5-fold cross-validation using
mne.decoding.cross_val_multiscore(), and the results are averaged across
subjects and visualised as a time course.
# Authors: MOABB contributors
#
# License: BSD (3-clause)
# sphinx_gallery_thumbnail_number = 2
import warnings
import matplotlib.pyplot as plt
import numpy as np
from mne.decoding import SlidingEstimator, cross_val_multiscore
from scipy.stats import ttest_1samp
from sklearn.linear_model import LogisticRegression
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import StandardScaler
import moabb
from moabb.datasets import BNCI2014_001
from moabb.paradigms import LeftRightImagery
moabb.set_log_level("info")
warnings.filterwarnings("ignore")
Loading the Dataset#
We instantiate the BNCI2014-001 dataset and use all 9 subjects.
dataset = BNCI2014_001()
Choosing a Paradigm#
The LeftRightImagery paradigm extracts
left-hand and right-hand motor-imagery epochs, applies a band-pass filter
(8–32 Hz by default), and returns the data as a 3-D NumPy array of shape
(n_trials, n_channels, n_times).
Building a Time-Resolved Pipeline#
A SlidingEstimator wraps any scikit-learn compatible
estimator and fits/scores it independently at every time point.
Here we use a simple logistic-regression classifier with Z-score
normalisation. The scoring='roc_auc' argument tells the estimator to
use AUC as the evaluation metric.
clf = make_pipeline(StandardScaler(), LogisticRegression(max_iter=1000))
sliding = SlidingEstimator(clf, scoring="roc_auc", n_jobs=1)
Evaluating Each Subject#
For each subject we:
Retrieve the preprocessed epochs via the paradigm using
return_epochs=Trueso we can extract the correct time vector and sampling frequency from themne.Epochsmetadata.Run stratified 5-fold cross-validation with
cross_val_multiscore(), which returns an array of shape(n_folds, n_times).Average over folds to obtain a single time course per subject.
All per-subject time courses are collected for later aggregation.
all_scores = []
for subject in dataset.subject_list:
epochs, y, meta = paradigm.get_data(
dataset=dataset, subjects=[subject], return_epochs=True
)
X = epochs.get_data()
# cross_val_multiscore returns (n_folds, n_times)
scores = cross_val_multiscore(sliding, X, y, cv=5, n_jobs=1)
all_scores.append(scores.mean(axis=0)) # average over folds
# Stack into (n_subjects, n_times)
all_scores = np.array(all_scores)
0%| | 0.00/44.8M [00:00<?, ?B/s]
0%| | 16.4k/44.8M [00:00<05:36, 133kB/s]
0%| | 56.3k/44.8M [00:00<03:03, 244kB/s]
0%| | 96.3k/44.8M [00:00<02:39, 279kB/s]
0%|▏ | 161k/44.8M [00:00<01:59, 374kB/s]
1%|▏ | 224k/44.8M [00:00<01:45, 423kB/s]
1%|▎ | 329k/44.8M [00:00<01:18, 565kB/s]
1%|▍ | 440k/44.8M [00:00<01:05, 673kB/s]
1%|▍ | 537k/44.8M [00:00<01:02, 705kB/s]
1%|▌ | 648k/44.8M [00:01<00:57, 765kB/s]
2%|▋ | 801k/44.8M [00:01<00:48, 908kB/s]
2%|▊ | 936k/44.8M [00:01<00:45, 963kB/s]
3%|▉ | 1.18M/44.8M [00:01<00:34, 1.26MB/s]
3%|█▏ | 1.43M/44.8M [00:01<00:28, 1.50MB/s]
4%|█▍ | 1.79M/44.8M [00:01<00:22, 1.92MB/s]
5%|█▊ | 2.21M/44.8M [00:01<00:18, 2.35MB/s]
6%|██▎ | 2.82M/44.8M [00:01<00:13, 3.13MB/s]
8%|██▉ | 3.50M/44.8M [00:02<00:10, 3.83MB/s]
10%|███▊ | 4.58M/44.8M [00:02<00:07, 5.29MB/s]
13%|████▋ | 5.74M/44.8M [00:02<00:06, 6.48MB/s]
17%|██████▎ | 7.68M/44.8M [00:02<00:04, 9.23MB/s]
22%|███████▉ | 9.64M/44.8M [00:02<00:03, 11.2MB/s]
29%|██████████▉ | 13.2M/44.8M [00:02<00:01, 16.4MB/s]
37%|█████████████▋ | 16.6M/44.8M [00:02<00:01, 19.7MB/s]
51%|██████████████████▋ | 22.7M/44.8M [00:02<00:00, 30.2MB/s]
60%|██████████████████████▎ | 27.0M/44.8M [00:03<00:00, 32.6MB/s]
71%|██████████████████████████▎ | 31.8M/44.8M [00:03<00:00, 34.5MB/s]
86%|███████████████████████████████▋ | 38.4M/44.8M [00:03<00:00, 39.4MB/s]
0%| | 0.00/44.8M [00:00<?, ?B/s]
100%|██████████████████████████████████████| 44.8M/44.8M [00:00<00:00, 186GB/s]
0%| | 0.00/44.8M [00:00<?, ?B/s]
0%| | 8.19k/44.8M [00:00<11:49, 63.1kB/s]
0%| | 56.3k/44.8M [00:00<03:03, 244kB/s]
0%| | 104k/44.8M [00:00<02:28, 302kB/s]
0%|▏ | 153k/44.8M [00:00<02:16, 328kB/s]
1%|▏ | 232k/44.8M [00:00<01:43, 431kB/s]
1%|▏ | 281k/44.8M [00:00<01:48, 410kB/s]
1%|▎ | 384k/44.8M [00:00<01:22, 536kB/s]
1%|▍ | 480k/44.8M [00:01<01:13, 601kB/s]
1%|▍ | 568k/44.8M [00:01<01:10, 625kB/s]
2%|▌ | 705k/44.8M [00:01<00:58, 755kB/s]
2%|▋ | 784k/44.8M [00:01<01:01, 712kB/s]
2%|▊ | 977k/44.8M [00:01<00:46, 947kB/s]
2%|▉ | 1.09M/44.8M [00:01<00:47, 920kB/s]
3%|█▏ | 1.37M/44.8M [00:01<00:33, 1.29MB/s]
4%|█▍ | 1.67M/44.8M [00:01<00:26, 1.61MB/s]
5%|█▊ | 2.15M/44.8M [00:02<00:19, 2.23MB/s]
6%|██▏ | 2.70M/44.8M [00:02<00:14, 2.82MB/s]
8%|██▉ | 3.49M/44.8M [00:02<00:10, 3.80MB/s]
10%|███▌ | 4.36M/44.8M [00:02<00:08, 4.67MB/s]
13%|████▊ | 5.86M/44.8M [00:02<00:05, 6.72MB/s]
16%|█████▉ | 7.12M/44.8M [00:02<00:04, 7.73MB/s]
22%|████████ | 9.70M/44.8M [00:02<00:03, 11.4MB/s]
28%|██████████▎ | 12.4M/44.8M [00:02<00:02, 14.3MB/s]
38%|█████████████▉ | 16.8M/44.8M [00:03<00:01, 20.1MB/s]
49%|█████████████████▉ | 21.7M/44.8M [00:03<00:00, 25.4MB/s]
62%|███████████████████████ | 27.9M/44.8M [00:03<00:00, 31.6MB/s]
76%|████████████████████████████▏ | 34.2M/44.8M [00:03<00:00, 36.1MB/s]
89%|█████████████████████████████████ | 40.0M/44.8M [00:03<00:00, 38.4MB/s]
0%| | 0.00/44.8M [00:00<?, ?B/s]
100%|██████████████████████████████████████| 44.8M/44.8M [00:00<00:00, 199GB/s]
Extracting the Time Vector#
Because we used return_epochs=True, we can read the time axis and
sampling frequency directly from the last Epochs object rather than
hard-coding dataset-specific values.
times = epochs.times
sfreq = epochs.info["sfreq"]
print(f"Sampling frequency: {sfreq} Hz, {len(times)} time points")
Sampling frequency: 250.0 Hz, 1001 time points
Statistical Significance#
We run a one-sample t-test against chance level (AUC = 0.5) at each time point. Time points with p < 0.05 (uncorrected) are flagged as significant.
_, p_values = ttest_1samp(all_scores, 0.5, axis=0)
sig_mask = p_values < 0.05
Plot 1 – Mean AUC Time Course with Significance#
We plot the group-average AUC score together with the standard error of the mean (SEM) across subjects. A horizontal dashed line at 0.5 indicates chance level. Time points that are significantly above chance are highlighted with an orange bar along the x-axis.
mean_scores = all_scores.mean(axis=0)
sem_scores = all_scores.std(axis=0) / np.sqrt(len(dataset.subject_list))
fig, ax = plt.subplots(figsize=(8, 4))
ax.plot(times, mean_scores, label="Mean AUC across subjects", color="steelblue")
ax.fill_between(
times,
mean_scores - sem_scores,
mean_scores + sem_scores,
alpha=0.3,
color="steelblue",
label="\u00b1SEM",
)
ax.axhline(0.5, linestyle="--", color="k", label="Chance level (AUC = 0.5)")
ax.axvline(times[0], linestyle=":", color="gray", label="MI onset")
# Mark significant time points with a bar at the bottom of the axes
ax.fill_between(
times,
0.0,
0.03,
where=sig_mask,
color="tab:orange",
alpha=0.7,
label="p < 0.05 (uncorrected)",
transform=ax.get_xaxis_transform(),
)
ax.set_xlabel("Time (s)")
ax.set_ylabel("AUC")
ax.set_title("Time-Resolved Decoding \u2013 Left vs. Right Motor Imagery\n(BNCI2014-001)")
ax.legend(loc="upper left", fontsize="small")
ax.set_xlim(times[0], times[-1])
ax.set_ylim(0.4, 1.0)
plt.tight_layout()
plt.show()
Plot 2 – Per-Subject Heatmap#
A heatmap of AUC scores (subjects x time) gives a richer picture than the mean curve alone, revealing inter-subject variability and the temporal structure of discriminability for each participant.
fig, ax = plt.subplots(figsize=(8, 4))
im = ax.imshow(
all_scores,
aspect="auto",
origin="lower",
extent=[times[0], times[-1], 0.5, len(dataset.subject_list) + 0.5],
cmap="RdBu_r",
vmin=0.3,
vmax=0.7,
)
ax.set_xlabel("Time (s)")
ax.set_ylabel("Subject")
ax.set_yticks(range(1, len(dataset.subject_list) + 1))
ax.set_title("Per-Subject Time-Resolved AUC\n(BNCI2014-001)")
ax.axvline(times[0], linestyle=":", color="k", linewidth=0.8)
ax.set_xlim(times[0], times[-1])
fig.colorbar(im, ax=ax, label="AUC")
plt.tight_layout()
plt.show()
Total running time of the script: (6 minutes 29.910 seconds)
Estimated memory usage: 933 MB
Run this example
