Unsupervised automatic online spike sorting using reward-based online clustering

Abstract

Brain-machine interfaces (BMIs) can enable paralyzed people to regain mobility. In these interfaces, some different type of signals can be obtained from the brain, one of which is the action potential waveform (spike). In the case of using spikes, sorting the recorded signals and isolating the effects of the individual neurons can lead to a greater efficiency. Also, because of the nature of BMIs, real-time spike sorting is necessary. In many spike sorting approaches, the main outline consists of the following steps: spike detection, feature extraction, and clustering. In this study, a novel method for clustering is presented. This method is referred to as Reward-Based Online Clustering (RBOC) which is formed based on the reinforcement learning algorithm. The significant property of this proposed technique is its capability for real-time implementation that is required by BMIs. This method can automatically detect the clusters while there is no knowledge about the number of clusters. The performance of the proposed method is demonstrated through both simulation and experimental study. Evaluation with artificially simulated (ground truth) data shows that, on average, the accuracy of categorizing the spikes from the same origins is above 94 percent. Moreover, implementation of the method on the experimental data obtained from the rat brain represents convincing sorting results. It is noteworthy to say that, in most cases, this new method outperforms the results of similar previous works.