Neural network classification of nerve activity recorded in a mixed nerve

Whole-nerve cuff electrodes can be used to record electrical nerve activity in peripheral nerves and are suitable for chronic implantation in animals or humans. If the whole nerve innervates multiple target organs or muscles then the recorded activity will be the superposition of the activity of different nerve fibers innervating these organs. In certain cases it is desirable to monitor mixed nerve activity and to determine the origin (modality) of the recorded activity. A method using the autocorrelation function of recorded nerve activity and an artificial neural network was developed to classify the modality of nerve signals. The method works in cases where different end organs are innervated by nerve fibers having different diameter distributions. The electrical activity in the cat S1 sacral spinal root was recorded using a cuff electrode during the activation of cutaneous, bladder, and rectal mechanoreceptors. Using the classification method, 87.5% of nerve signals were correctly classified. This result demonstrates the effectiveness of the neural network classification method to determine the modality of the nerve activity arising from activation of different receptors.