Optimizing Neural Recording Front-Ends Toward Enhanced Spike Sorting Accuracy in High-Channel-Count Systems.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Spike sorting is crucial for analyzing extracellular neural recordings.
• Neural recording front-end design significantly impacts spike sorting performance.
• Efficient front-ends are needed for power- and area-constrained multi-channel applications.

Purpose of the Study:

• To determine minimal analog front-end requirements for accurate spike sorting.
• To optimize designs for power and area efficiency in neural interfaces.
• To guide the development of high-channel-count recording systems.

Main Methods:

• Simulated neural recording front-end parameters using a MATLAB model.
• Evaluated filter order, cutoff frequency, ADC resolution, and sampling frequency.
• Utilized synthetic and real neural datasets with ground truth.

Main Results:

• Optimal spike sorting achieved with a 1st-order Butterworth bandpass filter (700 Hz - 7.5 kHz).
• Recommended analog-to-digital converter (ADC) specifications: 15 kHz sampling, 8-bit resolution, no missing codes.
• Identified key electrical parameters affecting signal integrity and sorting accuracy.

Conclusions:

• Specific filter and ADC parameters are critical for high spike sorting accuracy.
• Minimal design requirements enable power- and area-efficient neural recording front-ends.
• Findings are vital for optimizing CMOS circuits in high-density neural interfaces.