A spatiotemporal framework for MEG/EEG evoked response amplitude and latency variability estimation

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Evoked response analysis in magnetoencephalography (MEG) and electroencephalography (EEG) is crucial for understanding neural dynamics.
• Estimating single-trial parameters like latency and amplitude is challenging due to inherent variability and noise.

Purpose of the Study:

• To develop a spatiotemporal framework for accurate estimation of single-trial response latencies and amplitudes from MEG/EEG data.
• To introduce a generalized expectation-maximization algorithm, TriViAL, for maximum likelihood estimation of neural response parameters and noise characteristics.

Main Methods:

• Utilized spatial and temporal bases to model consistent evoked response features across trials.
• Developed the Trial Variability in Amplitude and Latency (TriViAL) algorithm, a generalized expectation-maximization method.
• Integrated source localization by scanning the TriViAL algorithm across cortical surface locations.

Main Results:

• TriViAL successfully computed maximum likelihood estimates for amplitudes, latencies, basis coefficients, and noise covariance.
• Demonstrated effective source localization using simulated and human evoked response data.
• Validated latency variability estimation using auditory M100 responses and showed improved amplitude estimation when latency was modeled.

Conclusions:

• The TriViAL framework provides a robust method for analyzing single-trial evoked responses in MEG/EEG.
• Accounting for trial-to-trial variability in amplitude and latency enhances the accuracy of neural signal analysis and source localization.
• This approach offers improved insights into neural processing dynamics and timing.