Intrinsic mode entropy based on multivariate empirical mode decomposition and its application to neural data analysis

Area of Science:

• Neuroscience
• Complexity Science
• Signal Processing

Background:

• Traditional entropy analysis of brain dynamics is limited by single-scale approaches.
• Multiscale entropy measures like intrinsic mode entropy (IMEn) offer improvements but face challenges with multivariate data.
• Existing methods struggle with mode-misalignment and mode-mixing in complex neural signals.

Purpose of the Study:

• To address limitations of existing multiscale entropy methods for neural data.
• To develop an improved intrinsic mode entropy (IMEn) analysis using multivariate empirical mode decomposition (MEMD).
• To enhance the discriminative power of entropy analysis by incorporating variance for analyzing neural dynamics.

Main Methods:

• Extended multivariate empirical mode decomposition (MEMD) for multi-channel, multi-trial neural data.
• Computed intrinsic mode entropy (IMEn) across multiple scales derived from MEMD.
• Incorporated variance into IMEn analysis for improved discriminant analysis.
• Applied the novel approach to local field potentials (LFPs) from macaque monkeys during a visual task.

Main Results:

• Neural entropy is scale-dependent and correlates with perceptual conditions.
• The proposed IMEn with variance achieved 83.05% accuracy in discriminating perceptual conditions.
• This accuracy is significantly higher than using IMEn alone (76.27%).

Conclusions:

• The MEMD-based IMEn with variance effectively captures complex neural dynamics.
• This enhanced method offers improved sensitivity for analyzing neural data and perceptual states.
• The findings suggest a more robust approach to understanding brain complexity and function.