High-resolution neural recordings improve the accuracy of speech decoding

Area of Science:

• Neuroscience
• Biomedical Engineering
• Speech Communication

Background:

• Neurodegenerative diseases impair communication, impacting patients' quality of life.
• Restoring communication necessitates decoding brain signals for neural speech prostheses.
• Current decoding methods are limited by inadequate neural recording resolution.

Purpose of the Study:

• To investigate the impact of high-resolution neural recordings on speech decoding accuracy.
• To evaluate micro-electrocorticography (µECoG) for capturing spatio-temporal brain signal structure.
• To enhance the capabilities of future neural speech prostheses.

Main Methods:

• Performed high-resolution micro-electrocorticography (µECoG) neural recordings during intra-operative speech production.
• Compared µECoG signals to macro-ECoG and SEEG for spatial resolution and signal-to-noise ratio.
• Utilized non-linear decoding models to leverage enhanced spatio-temporal neural information.

Main Results:

• Achieved 57x higher spatial resolution and 48% higher signal-to-noise ratio with µECoG.
• Demonstrated a 35% improvement in speech decoding accuracy compared to standard intracranial signals.
• Confirmed that accurate decoding depends on high-spatial resolution neural interfaces and non-linear models.

Conclusions:

• High-density µECoG provides superior neural signal quality for speech decoding.
• Advanced decoding models effectively utilize the enhanced spatio-temporal information from µECoG.
• This technology holds promise for developing high-quality neural speech prostheses.