Automatic threshold optimization in nonlinear energy operator based spike detection.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• The performance of neural spike sorting systems heavily relies on accurate spike detection.
• The Non-linear energy operator (NEO) is a widely used spike detector known for its accuracy and hardware efficiency.
• Current NEO methods use approximated thresholds, leading to suboptimal performance, especially in low signal-to-noise ratio (SNR) environments.

Purpose of the Study:

• To develop an automatic and robust threshold calculation method for the Non-linear energy operator (NEO) spike detector.
• To improve spike detection accuracy in neural signal processing, particularly under varying SNR conditions.
• To address the limitations of approximated thresholds in real-time spike sorting systems.

Main Methods:

• An empirical gradient technique was developed for automatic and robust threshold calculation.
• The proposed method was evaluated on two distinct neural signal datasets.
• Performance was assessed by comparing detection accuracy against traditional NEO thresholding methods.

Main Results:

• The optimized NEO threshold significantly improved detection accuracy across both high and low SNR signals.
• Statistical analysis demonstrated a notable increase in accuracy, with the 75th percentile accuracy reaching 98.7% for the optimized method versus 93.5% for the traditional method.
• The proposed method proved effective in real-time scenarios where SNR estimation is difficult.

Conclusions:

• The developed empirical gradient technique offers a superior approach to thresholding for NEO spike detectors.
• This optimization enhances spike detection performance, especially in challenging low SNR conditions common in neural recordings.
• The findings contribute to more reliable and accurate neural spike sorting systems.