Toward Completely Sampled Extracellular Neural Recording During fMRI

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• fMRI scanning introduces severe artifacts in neural recordings.
• Previous methods using PCA for EEG denoising during fMRI are limited in frequency range.
• Extracellular neural recordings capture a broad frequency spectrum (1-6000 Hz).

Purpose of the Study:

• To estimate and remove fMRI artifacts from extracellular neural recordings at ultrahigh magnetic fields.
• To adapt PCA-based denoising for the wide frequency range of extracellular field potentials (EFPs).
• To enable simultaneous fMRI and complete neural recording.

Main Methods:

• Adaptation of PCA for EFP denoising, covering 1-6000 Hz.
• Comparison of Singular Value Decomposition (SVD)-PCA Singular Value Shrinkage (SVS) with two shrinkage rules and sliding template subtraction.
• Development of a novel technique using temporal first difference for estimating singular value upper bounds.

Main Results:

• SVS methods proved advantageous over sliding template subtraction, particularly for high-frequency extracellular action potentials (EAPs).
• The developed techniques successfully uncovered EAPs during fMRI gradient interferences in artificial datasets.
• Analysis of natural datasets from rat cortex confirmed the efficacy of SVS for both local field potentials (LFPs) and EAPs.

Conclusions:

• The proposed SVS methods effectively denoise extracellular neural recordings corrupted by fMRI artifacts.
• This work facilitates simultaneous fMRI and high-fidelity neural recording (1-6000 Hz).
• Enables further investigation of brain function and neurovascular coupling at ultrahigh fields.