Multi-domain feature selection in auditory MisMatch Negativity via PARAFAC-based template matching approach

Area of Science:

• Cognitive Neuroscience
• Neuroimaging Analysis
• Signal Processing

Background:

• MisMatch Negativity (MMN) is a crucial event-related potential (ERP) for assessing sensory learning and perceptual accuracy in cognitive research.
• Group-level analysis of ERPs is vital for identifying subtle differences between conditions, especially with low signal-to-noise ratios.
• Tensor factorization offers a powerful framework for comprehensive ERP analysis by integrating multiple data domains.

Purpose of the Study:

• To develop an automated technique for selecting relevant multi-domain features from decomposed ERP tensors.
• To address the manual and time-consuming nature of component selection in group-level ERP analysis.
• To enhance the identification of group or condition-specific differences in MMN responses.

Main Methods:

• Generated a 4-way ERP tensor (time × frequency × channel × subjects/condition).
• Applied Parallel Factor Analysis (PARAFAC) for tensor decomposition.
• Implemented a template matching approach using MMN temporal and spectral signatures for feature selection, followed by statistical testing.

Main Results:

• Successfully identified a multi-domain ERP feature discriminating between experimental conditions.
• The selected feature, based on MMN signatures, significantly differentiated subjects across conditions.
• Demonstrated the efficacy of the automated template matching approach in ERP analysis.

Conclusions:

• The proposed template matching technique automates the selection of significant multi-domain ERP features.
• This method enhances group-level analysis of MMN, improving efficiency and accuracy.
• Facilitates the identification of neural correlates of cognitive processes in group comparisons.