Automatic noise-level detection for extra-cellular micro-electrode recordings.

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Extracellular neuro-recording signals are crucial for functional mapping in deep brain stimulation (DBS) surgery and brain-computer interfaces (BCIs).
• Poor signal-to-noise ratio (SNR) in these recordings hinders accurate spike extraction.
• Current noise estimation methods often overestimate noise levels, especially with increased neuronal activity or artifacts.

Purpose of the Study:

• To develop a novel and reliable automatic noise estimation method for neural recordings.
• To address the overestimation bias of existing noise estimation techniques.
• To improve the accuracy of spike extraction in neurophysiological data.

Main Methods:

• Proposed a new method based on estimating the mode of the signal envelope distribution.
• Utilized the Rayleigh distribution, characteristic of the envelope of band-limited Gaussian noise.
• The peak of the Rayleigh distribution provides a reliable noise-level estimate.

Main Results:

• The novel 'envelope' method demonstrated superior performance compared to existing methods.
• Validation was performed on both simulated neuro-recording data and actual micro-electrode recordings.
• The method proved effective even in the presence of high neuronal activity and artifacts.

Conclusions:

• The proposed envelope-based noise estimation method offers a more accurate and reliable approach.
• This technique enhances spike extraction from neural signals, crucial for DBS and BCIs.
• The method shows significant potential for improving the analysis of neurophysiological data in clinical and research settings.