Spike detection in human muscle sympathetic nerve activity using a matched wavelet approach

Aryan Salmanpour, Lyndon J. Brown, J. Kevin Shoemaker
Journal of Neuroscience Methods (2010)

Microneurography is often used to visualize nerve activity. Although these recordings can show changes in nerve activity, it is difficult to differentiate between background noise and actual nerve activity. This study discusses a new method for detecting action potentials in nerve activity from human muscle. The method attempts to find the location of action potentials in high levels of noise by improving wavelet techniques.

Muscle sympathetic nerve activity (MSNA) data was recorded from seven healthy individuals between the ages of 23 and 30. There were two males and five females. The peroneal nerve was used to retrieve the recordings because it contains nerves with skin and muscle blood vessels as their destination. The main purpose of wavelet analysis is to take the signal that has all the action potentials that are buried in noise and decompose it using continuous wavelet transformation (CWT) to filter the MSNA signal into a matched wavelet called a mother wavelet. To create the mean action potential template, the ten largest sympathetic bursts from each patient were used. The non-overlapping action potentials in ten bursts for each patients were then averaged to create the mean action potentials. The participant’s mean action potentials were then averaged to create one mean action potential template called the mother wavelet template. The mother wavelet helps create a resemblance index which can give you wavelet coefficients between the signal of interest and the mother wavelet. The wavelet coefficients are large in the presence of action potentials, which means there is a strong resemblance, and very low in segments with only noise. Thresholds were then created to determine the location of individual action potentials. The action potentials are then separated from the original filtered MSNA.

This new method has advanced the ability to look at spikes in sympathetic nerve activity in humans and helps to further understand just how the nervous system sends signals. Although this process was designed for human sympathetic action potentials, it could possibly be used for spike detection in other nerve recordings. The algorithm should work fine, just as long as there is an action potential template for recordings being studied. The optimized threshold and mother wavelet used in this study does not apply to all other tests. It was created specifically for this study. Future experiments could attempt to find a threshold that can become standard for all tests. Also, this method finds the locations of action potentials in high levels of noise, but cannot distinguish action potentials from different axons. 

Paul M