Low-latency single channel real-time neural spike sorting system based on template matching

Area of Science:

• Neural Engineering
• Computational Neuroscience
• Biomedical Engineering

Background:

• Advancements in neural engineering enable simultaneous recording and control of neural circuits.
• Closed-loop neural control systems require rapid processing of neural data.

Purpose of the Study:

• To develop a rapid spike sorting system for real-time calculation of instantaneous firing rates.
• To compare Euclidean distance and correlational matching for spike sorting accuracy and firing rate calculation performance.

Main Methods:

• A hybrid hardware/software system was developed using Super-paramagnetic clustering for template generation and field-programmable gate array for template matching.
• Two matching techniques, Euclidean distance (ED) and correlational matching (CM), were evaluated.
• System performance was validated using artificial data, pre-recorded neural spikes, and real-time recordings from awake mice.

Main Results:

• Both ED and CM achieved high sorting accuracies.
• CM demonstrated greater robustness for in vivo recordings, handling non-Gaussian spike distributions effectively.
• The system achieved a sorting latency under 2 ms and a maximum sorting rate of 941 spikes/second, outperforming other off-line algorithms in speed.

Conclusions:

• The developed rapid spike sorting system offers high accuracy and significantly reduced sorting time compared to existing methods.
• The system's low latency and high sorting rate are crucial for real-time neural activity analysis.
• This technology facilitates future developments in neural circuit modulation via real-time neural activity analysis.