Academy

Brain–Computer Interfaces

Definition: A brain–computer interface (BCI) is a system that translates neural signals into commands for external devices, enabling direct communication between the brain and machines. BCIs can restore function for patients with paralysis or communication disorders by decoding intended movements or speech from measured brain activity.

Key Components:

• Signal acquisition: Recording neural activity through noninvasive (EEG), partially invasive (ECoG), or invasive (SEEG, microelectrodes) methods.
• Signal processing: Filtering, artifact removal, and feature extraction (e.g., frequency bands such as high-gamma).
• Decoding algorithms: Machine learning or deep learning models (e.g., LSTM networks) that map neural features to intended actions.
• Application interface: Output device such as a robotic arm, cursor, or communication aid.

How BCIs Work:

1. Neural signal capture: Electrodes measure electrical activity in specific brain regions involved in motor planning or sensory processes.
2. Preprocessing: Remove noise (e.g., line noise, muscle artifacts) and re-reference signals to improve signal-to-noise ratio.
3. Feature extraction: Compute time-frequency representations and isolate informative bands like high-gamma (70–200 Hz).
4. Dimensionality reduction: Techniques such as demixed PCA separate task-relevant variance, yielding compact features that reflect movement direction or state.
5. Decoding: Deep learning models (e.g., bidirectional LSTM) learn temporal patterns to predict user intent (e.g., left vs. right movement).
6. Command execution: Predictions drive external devices, providing real-time feedback and enabling closed-loop control.

Relevance to Longevity Science: As populations age, neurodegenerative conditions (e.g., ALS, stroke) impair motor function. BCIs offer potential to maintain independence and quality of life by bypassing damaged pathways and restoring communication or limb control.

High-Gamma Neural Activity

Definition: High-gamma refers to neural oscillations between 70 and 200 Hz. These fast rhythms closely correlate with local neuronal firing rates and encode specific task information, such as motor intent or sensory processing.

Characteristics:

• Spatial specificity: High-gamma power localizes sharply to active cortical or subcortical sites.
• Temporal precision: Rapid fluctuations track fine timescales of movement planning and execution.
• Task modulation: Bandpower increases during active tasks, such as reaching or grasping, and decreases at rest.

Measurement Methods:

• SEEG: Depth electrodes sample from deep structures (e.g., insula, hippocampus) for high spatial resolution.
• ECoG: Surface grids record cortical high-gamma signals with minimal invasiveness.

Role in Decoding: Because high-gamma power carries movement-specific information, isolating this band enhances decoding accuracy. Combining multitaper spectral analysis with re-referencing maximizes the clarity of high-gamma features for machine learning models.