Dissertation
Dissertation > Medicine, health > Basic Medical > Human Physiology > Neurophysiology

Novel Method of Neuronal Spike Detection and Sorting Based on Broad Band Signal and Its Application

Author ChenBaiZuo
Tutor FengZhouYan
School Zhejiang University
Course Biomedical Engineering
Keywords Broadband Recording Signal High Frequency Electrical Stimulus Detection Classification Multi-Channel Window Detection Method Template Matching
CLC R338
Type Master's thesis
Year 2014
Downloads 1
Quotes 0
Download Dissertation

Research about encoding mechanism of brain systems and signal processing mechanisms is based on extracellular neuronal action potential (spike). Microelectrode array provides a good way to record action spikes and field potentials. So how to detect spikes and cluster spikes more effectively, especially the types of neurons which create spikes, is one of the key problems in brain research.Here we solved two problems:how to use the broad band signals for spikes detection and sorting directly; how to detect spikes from high frequency electrical stimulations signals. These problems solved is meaningful to distinguish the types of neurons correctly, and helpful to the research of deep brain stimulation mechanism.To analyze the recorded experimental signals, the first thing is filtering signals to eliminate high and low frequency interference, and then we continue the detection and classification of spikes. But filter make waveforms of spikes deform, that lead to the error of spike classification, and the types of neurons is hard to distinguish. For example, the intermediate neurons and pyramidal neurons in the hippocampus, the width of pyramidal neurons waveforms is larger than interneurons. High-pass filter will make the width of pyramidal neurons waveforms narrower, wlieh the amplitudes of pyramidal neurons’waveforms will be similar with interneurons’on the same channel, that makes identification of spikes harder. To solve this problem, we designed an algorithm of spikes detection and classification based broadband recording signals, without filtering. The method include a multi-channel broadband signal window method for spike detection and adaptive feature extraction method to extract16-25features for distinction of different types of neurons and spike sorting. The results of analyzing experimental signals and simulation data showed that:multi-channel window detection method could detect spikes in the low-frequency field potential of broadband signals and performed accurately; and adaptation feature extraction of spikes from broadband signals could effectively distinguish spikes of pyramidal neurons and interneurons, whose amplitudes are similar on the same channel, therefore it provides a new approach for correct classification and distinction of different types of neurons. We also extend the method based on broad band signals. come up with a new algorithm of spike detection and classification based on template matching to analyze the high frequency electrical stimulation signals. In the high frequency electrical stimulation signals, there are stimulus artifacts and PS (population spike. PS) recorded in the signals. If PS waves through the high pass filter will lead to an "oscillation" wave that block the detection and classification of spikes. Therefore, we design the new algorithm to detect and cluster spikes from the high frequency electrical stimulation signals. First of all. we extract the template from the signals, and then calculate the difference of the matching template and the signal segment to detect particular spikes. For improving the calculating speed of algorithm, we take two measures:1) calculating the difference between the maximum and minimum values of the signal segment in the moving window, if this difference is between the upper and lower thresholds, then we calculate the difference of the matching template and the signal segment.2) using the C language template matching algorithm in MATLAB. the running time of C program is shorter than MATLAB. just2.5percent of the latter. Experimental results showed that:when there is a broadband signal and the stimulus artifact and PS are also in the signal, we can detect and classify the spikes with high accuracy through template matching method, without eliminating the stimulus artifact and PS. we could complete the detection and classification of original signal directly.In a word, this paper presents the new methods of spikes detection and classification based on the broadband recording signals, the signal don’t need to be processed, they can analyze the original signals directly:for the signal including electrical stimulus, current source localization, where the signals cannot be filtered for analysis, here proposed new methods of spikes detection and classification.

Related Dissertations
More Dissertations