Tracking changes in action potential shapes in chronic multi-unit intrafascicular recordings using neural network pattern recognition techniques

Abstract

A novel scheme is proposed to train an Artificial Neural Network (ANN) classifier, on a repeated basis, in order to track temporal changes in the shapes of the action potentials recorded through chronically implanted intrafascicular electrodes. This scheme uses classification results of the ANN classifier on the most recent neural recordings to label the new action potentials. The ANN classifier is retrained using the new samples so that it recognizes any changes in the shapes of the action potentials. The procedure is repeated continuously using the most recently trained ANN classifier. This scheme was tested on different simulated situations that may arise in a two unit neural recording. The results indicate that proposed method allows us to track the changes in the shapes of the action potentials in most plausible scenarios that might arise in chronic intrafascicular recordings.