Performance of real time separation of multi-neuron recordings with a DSP32C microprocessor

Area of Science:

• Neuroscience
• Signal Processing
• Computational Biology

Background:

• Accurate sorting of neuronal spike waveforms is crucial for understanding neural activity.
• Existing methods require robust evaluation under realistic noise conditions.

Purpose of the Study:

• To evaluate the performance of a specific waveform sorting method (Gädicke and Albus, 1995) for multi-neuron data.
• To assess the system's capabilities in real-time processing and classification of artificial spike patterns with added noise and simulated EEG waves.

Main Methods:

• Utilized computer-generated artificial spike patterns.
• Introduced noise and free-running sine waves (simulating EEG) to spike data.
• Employed a DSP32C for continuous spike processing and real-time sorting.
• Analyzed system performance with varying signal-to-noise ratios and sine wave amplitudes.

Main Results:

• The DSP32C processed spikes at 183.106 Hz, performing real-time sorting and running averages.
• Discrimination against a 50 Hz sine wave was accurate for amplitudes up to 2.5 times the smallest spike.
• Classification errors were <0.1% when model spikes were derived from noiseless data.
• Errors increased to ~4% for low-amplitude spikes (SNR 3.3) when using noisy spikes for model definition, but were <1% for higher amplitude spikes (SNR ≥5).

Conclusions:

• The evaluated waveform sorting method demonstrates high accuracy and efficiency for multi-neuron data processing.
• The system is capable of real-time analysis, handling noise and simulated EEG interference effectively.
• Standardized artificial patterns facilitate objective comparison of different spike-sorting techniques.