Constrained temporal parallel decomposition for EEG-fMRI fusion.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Multimodal neuroimaging, particularly simultaneous electroencephalography (EEG) and functional magnetic resonance imaging (fMRI), is crucial for understanding brain activity due to the complementary spatiotemporal resolutions of the modalities.
• Existing EEG-fMRI fusion techniques, such as joint-ICA and parallel-ICA, often rely on sensor-level EEG features.
• Source-space independent component analysis (ICA) has recently emerged as a superior alternative to sensor-space ICA for EEG-MEG source reconstruction and component identification.

Purpose of the Study:

• To extend source-space ICA for the fusion of EEG-fMRI data, incorporating a paradigm signal constraint and a lag-based decomposition approach.
• To address the potentially variable lag structure between electrophysiological (EEG) and blood-oxygen-level-dependent (BOLD) signals.
• To evaluate the proposed method on both simulated and real concurrent EEG-fMRI data during cognitive tasks.

Main Methods:

• Developed and applied an extended source-space ICA framework for EEG-fMRI data fusion.
• Incorporated a paradigm signal constraint and a lag-based signal decomposition to model variable time lags between EEG and fMRI signals.
• Validated the method using simulated EEG-fMRI data during a boxcar task and real EEG-fMRI data from participants performing an N-Back working memory task.

Main Results:

• The proposed method successfully estimated brain maps for fused EEG and fMRI components.
• For real N-Back task data, frontal theta and right posterior alpha EEG components showed high contributions (~0.5) to the paradigm-related fused component.
• Key fMRI components identified were related to the attention network (~0.8) and the default mode network (~0.3).

Conclusions:

• The novel source-space ICA approach effectively fuses EEG and fMRI data, providing insights into neural activity.
• The method accounts for the temporal dynamics and variable lag structure between electrophysiological and BOLD signals.
• This technique offers a robust framework for multimodal neuroimaging analysis, enhancing the understanding of brain function during cognitive tasks.