MVPAlab

MVPAlab implements multivariate pattern analysis and related machine-learning methods to decode and characterize spatiotemporal patterns in electroencephalography (EEG) and magnetoencephalography (MEG) data.

Key Features:

• Multivariate Pattern Analyses: Implements a range of machine-learning algorithms for decoding multivariate patterns in EEG and MEG data.
• Cross-Classification and Temporal Generalization Matrices: Supports cross-classification and temporal generalization matrices to assess pattern generalization across experimental conditions and time.
• Feature and Frequency Contribution Analyses: Provides analyses of feature and frequency contributions to identify informative channels, timepoints, and frequency bands.
• Preprocessing Routines: Includes preprocessing routines such as data normalization, smoothing, dimensionality reduction, and supertrial generation.
• Statistical Inference with Non-Parametric Cluster-Based Permutation: Implements non-parametric cluster-based permutation testing for group-level statistical inference and multiple-comparisons control.

Scientific Applications:

• Cognitive neuroscience: Decoding neural representations underlying cognitive processes using multivariate EEG/MEG analyses.
• Sensory perception: Characterizing spatiotemporal dynamics of sensory perception from EEG/MEG signals.
• Neurological disorder research: Investigating neural signatures associated with neurological disorders using multivariate EEG/MEG analyses.
• Brain dynamics studies: Studying complex spatiotemporal brain dynamics with temporal-generalization and cross-classification approaches.

Methodology:

Applies machine-learning-based multivariate pattern analyses, cross-classification, temporal generalization matrices, feature and frequency contribution analyses, preprocessing routines (normalization, smoothing, dimensionality reduction, supertrial generation), and non-parametric cluster-based permutation testing; validated on a sample EEG dataset with significant clusters observed (p < 0.01) across decoding analyses.