From mixed signals to neural code: compensating volume conduction for noninvasive generic neural interfaces.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Noninvasive neural interfaces offer scalable access to neural data without surgical risks.
• A key challenge is signal distortion caused by biological tissues acting as a volume conductor.
• Recorded signals (EEG, ENG, surface EMG) are convolved with tissue properties, obscuring true neural information.

Purpose of the Study:

• To address the volume conductor effect in noninvasive neural recording.
• To propose and discuss strategies for compensating tissue-induced signal distortions.
• To enable more accurate and scalable neural information extraction.

Main Methods:

• Framing the challenge using a generic convolutive model.
• Discussing direct compensation via deconvolution methods to recover neural sources.
• Exploring indirect compensation through learning invariant representations from diverse datasets.

Main Results:

• The volume conductor effect fundamentally limits noninvasive neural interfaces.
• Deconvolution can potentially separate and recover neural sources.
• Learning invariant representations may overcome variability in tissue properties.

Conclusions:

• Explicitly recognizing and compensating for volume conductor effects is crucial for advancing noninvasive neural interfaces.
• Both direct and indirect compensation strategies are vital.
• These approaches can lead to scalable interfaces that isolate neural information from tissue distortions, reducing reliance on subject-specific calibration.