Electroencephalography (EEG) and magnetoencephalography (MEG) measure signals generated by activity across the whole brain with high temporal resolution. But what neural information is contained in these signals, how do we relate it to stimuli, and how can we separate neural sources from noise and artifacts?

This course provides a comprehensive overview of EEG and MEG analysis methods, from ERP-based approaches to time-frequency and connectivity analyses and statistical modeling. Using both generative simulations and real-world data, participants will gain extensive experience with MNE-Python alongside NumPy, Matplotlib, Scikit-learn, MNE-Connectivity, and mTRFpy.

Course Outline

The course meets weekly. Each course day consists of a 3.5-hour live online session and a homework assignment for a total workload of 20 hours.

Work with raw EEG data, epoch continuous recordings to create event-related potentials, visualize ERPs in butterfly plots and topographic maps, learn how re-referencing affects their shape, and statistically compare ERPs across experimental conditions using t-tests and permutation-based cluster analyses.

Session 2: Spatial and Temporal Filters

Explore the relationship between EEG and MEG recordings and the underlying neural activity by simulating data from a source model, separate neural and artifactual sources with PCA and ICA, and improve the signal-to-noise ratio by applying FIR filters while examining their advantages and potential pitfalls.

Session 3: Time-Frequency and Connectivity Analysis

Understand complex wavelet convolution and its outputs, use Morlet wavelets to compute time-frequency representations and characterize task-related changes in spectral power, and estimate functional connectivity using the phase-locking value (PLV) and phase lag index (PLI).

Session 4: Modeling Continuous Neural Responses

Apply ordinary least squares and ridge regression, compare model accuracy using correlation, mean squared error, and the coefficient of determination, construct time-lagged design matrices, build linear encoding models for continuous neural recordings, and model EEG responses to naturalistic speech.

Learning and participation

Topics

  • Analyzing Event-Related Potentials: computing ERPs, creating butterfly and scalp plots, and applying permutation-based cluster tests with MNE.
  • Spatial and Temporal Filters: simulating data from a source model, separating sources with PCA and ICA, and preprocessing data with FIR filters using MNE and Scikit-learn.
  • Time-Frequency and Connectivity Analysis: computing wavelet convolution, time-frequency representations, and phase-based connectivity with MNE and MNE-Connectivity.
  • Linear Modeling of Time-Series Data: regression, regularization, model fitting, and evaluation with NumPy and mTRFpy.

Intended participants

  • Researchers and students from all universities are welcome.
  • Participants of all educational backgrounds are welcome.
  • Comfort with Python data analysis and basic familiarity with neuroscience or signal-processing concepts are helpful.

Useful resources for preparation

Certification

Participants who attend at least 75% of the course and complete the assignments will receive a certificate by email at the end of the course.