Spike train discrimination and analysis in neural and surface electromyography (sEMG) applications

The term “spike” is used to describe a short-time event that is the result of the activity of its source. Spikes can be seen in di?erent signal modalities. In these modalities, often more than one source generates spikes. Classi?cation algorithms can be used to group similar spikes, ideally spikes from the same source. This work examines the classi?cation of spikes generated from neurons and muscles. When each detected spike is assigned to its source, the spike trains of these sources can provide information on complex brain network functioning, muscle disorders, and other applications. During the past several decades, there were many attempts to create and improve spike classi?cation algorithms. No matter how advanced these methods are today, errors in classi?cation cannot be avoided. Therefore, methods that would determine and improve reliability of classi?cation are very desirable. In this work, it is attempted to ?nd e?cient solutions for post-processing of spike trains extracted after the classi?cation to increase their accuracy and reliability. This is done for two di?erent signal types: neural and muscle signals. The aim was to create a reliable and automated signal processing framework. In modern neural microarray probes, many recording channels provide large quantities of data, which require an enormous amount of time to be analyzed. A fully automated system is therefore needed but is still out of reach. As for the analysis of muscle signals, treated by surface electromyography (sEMG many recording channels observe the same spikes simultaneously. Due to the frequently overlapping spikes, only automated sophisticated methods are able to perform classi?cation. Starting from the problem of spike classi?cation errors as such, independent from the used algorithm, the possibility of classi?cation in low signal-to-noise ratio (SNR) cases is examined. This part of the work indicated the under limits in terms of SNR for trustworthy classi?cation. Then, parameters to describe ?ring pattern of spikes are described and new parameters are proposed. These parameters are tested for reliability and the amount of information they are able to retrieve from the ?ring pattern. This logic and knowledge is then exploited in a obsessive-compulsive disorder (OCD) study where the analysis of extracted spike trains revealed the connection of a speci?c brain area (bed nucleus of striaterminalis – BST) to this disorder. In the case of the analysis of muscle signals using surface multichannel recording grids, classi?cation of spikes is called decomposition. Due to more frequent spiking of di?erent sources, spikes of di?erent sources are often overlapping and the goal is to extract time instances of individual sources as well as the respective spike shapes. Achieving the decomposition of surface electromyography (sEMG) signals is needed to better understand certain muscle disorders and increase knowledge of the neuromuscular system. A new method for decomposition is proposed and successfully applied as demonstrated on real and simulated data. Motor units are the lowest functional units in muscles. Each motor unit produces characteristic spikes. The goal of motor unit tracking is to identify the same motor units in subsequent measurements. Using the developed decomposition method to extract spikes from voluntary muscle contractions, a solution is devised that assists the motor unit tracking. The method is able to compensate for the e?ects of recording grid displacements in an automated manner. This results in substantial reduction of previously needed man-hours to achieve very precise placement of the measurement grid. The feasibility of this method is demonstrated on a recording mimicking accidental grid displacements. This tool provides the possibility to do non-invasive follow-up studies investigating the e?ects of various muscle disorders.