FMRI signal analysis using empirical mean curve decomposition

Area of Science:

• Neuroimaging
• Signal Processing
• Computational Neuroscience

Background:

• Functional magnetic resonance imaging (fMRI) data presents complex nonlinear time series with components across multiple temporal scales.
• Existing fMRI analysis predominantly relies on model-based approaches, with limited focus on data-driven methods for signal decomposition.
• The nonlinear nature and multiscale temporal components of fMRI signals pose significant analytical challenges.

Purpose of the Study:

• To present a novel data-driven framework for analyzing fMRI signals, named Empirical Mean Curve Decomposition (EMCD).
• To develop a method capable of inferring meaningful low-frequency information from complex fMRI data.
• To enhance functional brain mapping through advanced signal decomposition techniques.

Main Methods:

• The proposed Empirical Mean Curve Decomposition (EMCD) framework optimizes mean envelopes from fMRI signals.
• EMCD iteratively extracts signal components from coarser to finer temporal scales.
• The method is data-driven, requiring fewer assumptions compared to model-based approaches.

Main Results:

• EMCD was successfully applied to resting-state fMRI, task-based fMRI, and natural stimulus fMRI datasets.
• The framework effectively inferred meaningful low-frequency information from blood oxygenation level-dependent (BOLD) signals.
• Promising results were obtained, demonstrating the utility of EMCD for functional brain mapping.

Conclusions:

• Empirical Mean Curve Decomposition (EMCD) offers a novel and effective data-driven approach for fMRI signal analysis.
• The multiscale decomposition capability of EMCD is particularly valuable for understanding complex brain dynamics.
• EMCD shows significant potential for advancing functional brain mapping and analyzing diverse fMRI data types.