| Summary: | 碩士 === 國立臺灣大學 === 醫學工程學研究所 === 95 === The advance of ensemble recording makes it possible to simultaneously observe the neuronal activities of a specific brain area and has been a trend in neuroscience. The data from ensemble recording are multi-unit action potentials (APs), which need computer software to analyze before giving useful information. In our laboratory, we have focused on studying evoked responses. This thesis is mainly about developing software for the analysis of multi-unit APs, i.e. spike sorting. A data acquisition (DAQ) interface is also developed in this thesis.
We found that evoked multi-unit analysis is very different from analyzing high SNR spontaneous neural signals. First, we finished the ordinary sorting methods and applied to our data. Principal component analysis (PCA) and k-means were used to identify units from the detected spikes. Then, we improved the spikes detection, which is the first step of spike sorting. By using modified cumulative energy difference (CED), the detection of multi-unit action potentials was refined to have better discrimination. Further, to see if this detection has significant differences to the original CED, we compared the numbers and the flips of spikes collected by these two detections.
The DAQ interfaces we developed have functions of counters for firing rate counting, tape play back, and multi-channel acquisitions of action potentials and/or field potentials. All of these functions support triggers from the stimulators for synchronized responses.
We compared our program to two commercial products, which records and analyzes spontaneous APs. They are less suitable for our experiments because they don’t support triggered recording. Our spike sorting was developed for evoked APs, which are usually overlapped with each other. Further, we are still improving our spike sorting program and trying to include new appropriate techniques.
|
|---|
