Comparing the efficacy of data-driven denoising methods for a multi-echo fMRI acquisition at 7T

Area of Science:

• Neuroimaging
• Magnetic Resonance Imaging
• Brainstem Neuroimaging

Background:

• Functional magnetic resonance imaging (fMRI) signal is susceptible to noise from physiological, motion, and thermal sources.
• Ultrahigh field strength (e.g., 7T) fMRI offers improved spatial accuracy, particularly beneficial for brainstem imaging.
• Physiological noise increases with field strength, necessitating effective denoising strategies for brainstem fMRI.

Purpose of the Study:

• To compare the efficacy of various denoising methods for multi-echo fMRI data at ultrahigh field strength (7T).
• To evaluate Multi-Echo Independent Component Analysis (ME-ICA) against other data-driven approaches like aCompCor and ICA-AROMA.
• To assess the impact of denoising on data quality metrics and the ability to remove physiological noise.

Main Methods:

• Comparison of denoising methods: ME-ICA, anatomical component based correction (aCompCor), and Automatic Removal of Motion Artifacts (ICA-AROMA) (aggressive and non-aggressive).
• Assessment of data quality using metrics such as temporal signal-to-noise ratio (tSNR), DVARS, and spectral properties.
• Evaluation of methods' ability to decouple global signal and respiration.

Main Results:

• ME-ICA followed by aCompCor demonstrated superior post-processing for removing spatially diffuse noise.
• ME-ICA combined with aCompCor and aggressive ICA-AROMA were highly effective for denoising 7T multi-echo fMRI data.
• ME-ICA with aCompCor potentially retained more signal of interest than aggressive ICA-AROMA.

Conclusions:

• ME-ICA combined with aCompCor is a highly effective denoising strategy for ultrahigh field, multi-echo fMRI.
• This combined approach offers advantages in preserving neural signal while mitigating noise at 7T.
• The findings support the use of ME-ICA and aCompCor for advanced brainstem fMRI studies.