Evaluating the efficacy of multi-echo ICA denoising on model-based fMRI

Area of Science:

• Neuroimaging
• Cognitive Neuroscience
• Data Analysis

Background:

• Functional magnetic resonance imaging (fMRI) is crucial for neuroscience but suffers from noise, limiting analysis techniques like voxel-wise modeling.
• Multi-echo ICA denoising (ME-ICA) shows promise for reducing noise in resting-state fMRI, but its impact on task-based paradigms is less understood.

Purpose of the Study:

• To investigate the effectiveness of ME-ICA in improving data quality and model fitting performance for task-based fMRI, specifically during visual population receptive field (pRF) mapping.
• To compare ME-ICA with optimally combined multi-echo data and standard single-echo processing.

Main Methods:

• Acquired multi-echo fMRI data during a visual pRF mapping experiment (N=13).
• Applied three preprocessing pipelines: ME-ICA, optimal multi-echo combination (no ICA), and single-echo processing.
• Compared temporal signal-to-noise ratio (SNR), model fitting performance (variance explained), and parameter estimate reliability.

Main Results:

• Multi-echo fMRI enhanced temporal SNR compared to single-echo, with ME-ICA providing further improvement over optimal combination alone.
• Unexpectedly, improved SNR did not directly translate to better model fitting; only ME-ICA significantly improved model fitting performance compared to single-echo acquisition.
• ME-ICA enhanced variance explained by the pRF model across the visual system, including low-SNR anterior regions, and improved parameter reliability.

Conclusions:

• ME-ICA effectively denoises task-based fMRI data for modeling analyses, preserving data integrity.
• ME-ICA offers significant benefits for complex fMRI experiments, including voxel-wise modeling and naturalistic paradigms, by improving model fitting and parameter reliability.