Magnetometers vs Gradiometers for Neural Speech Decoding.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Neural speech decoding aims to restore communication for patients with severe motor impairments, like those with amyotrophic lateral sclerosis (ALS).
• Magnetoencephalography (MEG) offers high temporal resolution crucial for decoding rapid speech dynamics.
• Optically Pumped Magnetometers (OPMs) in wearable MEG devices present future BCI possibilities, but their use in speech decoding is unevaluated.

Purpose of the Study:

• To evaluate the performance of magnetometers versus gradiometers for neural speech decoding using MEG.
• To assess the impact of wavelet-based denoising on speech decoding accuracy for both sensor types.

Main Methods:

• Decoded imagined and spoken speech from MEG signals of seven healthy participants.
• Compared decoding performance between magnetometers and gradiometers.
• Applied a wavelet-based denoising strategy to MEG data.

Main Results:

• Magnetometers demonstrated potential for neural speech decoding, but with significantly lower performance than gradiometers.
• Wavelet denoising substantially enhanced decoding accuracy for both magnetometer and gradiometer data.
• Gradiometers remain preferable for current MEG-based speech decoding analysis.

Conclusions:

• While gradiometers are currently superior, magnetometers (and OPMs) show promise for future speech brain-computer interface (BCI) development.
• Wavelet denoising is an effective strategy to improve neural speech decoding performance with MEG.
• This study opens avenues for next-generation, potentially wearable, speech BCIs.