Extraction of motor unit action potentials from electromyographic signals through generative topographic mapping

The extraction of motor unit action potentials (MUAPs) from electromyographic (EMG) signals (also known as EMG decomposition) is an important step in investigations aiming to obtain information on control strategies of the neuromuscular system and its state. For instance, the analysis of the shape of MUAPs and their frequency of occurrence may be used as an additional tool in the detection of some neuromuscular disorders. Although MUAPs can be manually extracted from the EMG, such a procedure is often time consuming and prone to error. In this context, systems which aim to automate the extraction of MUAPs play an important role. First, they allow for the reduction in the processing time of signals, and secondly, they introduce consistency across analyses. In this work, we present an automatic system for the extraction of MUAPs based on generative topographic mapping (GTM), which is a recently developed technique for data clustering and visualization. The system is composed of several signal processing units, including signal filtering and detection, feature selection, data clustering and visualization. Its input is a time-series, representing EMG activity, and its output is the visualization of MUAPs obtained through GTM. The performance of the system was assessed via the analysis of synthetic and experimental EMG signals, detected by means of concentric needle and surface electrodes, collected from healthy subjects executing muscle contractions with distinct levels of force. Our results show that the system is capable of accurately extracting MUAPs from the EMG.