Robust spinal cord resting-state fMRI using independent component analysis-based nuisance regression noise reduction

Area of Science:

• Neuroimaging
• Biomedical Engineering

Background:

• Physiological noise significantly impacts spinal cord (SC) resting-state fMRI (rsfMRI) data quality.
• Effective noise reduction is crucial for reliable SC rsfMRI analysis.

Purpose of the Study:

• To evaluate the efficacy of an independent component analysis (ICA)-based nuisance regression (ICANR) method for reducing physiological noise in SC rsfMRI.
• To compare ICANR with conventional noise reduction techniques in terms of noise influence and data reproducibility.

Main Methods:

• A retrospective study involving ten healthy subjects was conducted.
• rsfMRI data were acquired using 3T/gradient-echo echo planar imaging (EPI).
• ICANR was compared against three methods: no regression (Nil), ROI-based noise reduction, and CORSICA. Physiological noise influence was assessed using correlation coefficients (CC), and reproducibility was measured by intraclass correlation (ICC).

Main Results:

• ICANR significantly reduced cerebrospinal fluid (CSF) pulsation and tissue motion influence (P < 0.001 for CSF, P < 0.02 for motion) compared to Nil.
• CORSICA showed increased influence of CSF pulsation and tissue motion (P = 0.048 for both).
• ICANR achieved the highest reproducibility (ICC = 0.766), outperforming Nil (0.669), ROI-based (0.645), and CORSICA (0.561).

Conclusions:

• ICANR is a highly effective method for reducing physiological noise in SC rsfMRI.
• ICANR significantly improves the robustness and reproducibility of SC rsfMRI analysis compared to existing methods.