Multivariate Analysis on Preprocessed Time-Frequency Representations of Electrode Voltage Signals from Microelectrode Array Experiments on an in-vitro Dopaminergic Neuronal Culture
MetadataVis full innførsel
The Master's thesis presents custom developed Fourier based preprocessing methods as well as results from principal component analysis (PCA) and partial least squares projection to latent structures (PLS) regression on biological neural data from a microelectrode array (MEA) in-vitro culture. The mixed neuronal cell culture was grown by PhD students at Department of Neuromedicine and Movement Science (INB), Faculty of Medicine and Health Sciences, NTNU and contained a specific type of dissociated human midbrain dopamine neurons known to have a selective vulnerability in Parkinson s disease (PD). The differentiation of the culture was part of PD related research. The goal of the analysis has been to look for neuronal signaling properties preferably related to motor tasks that can be used to control a robot in the NTNU Cyborg project. 13 MEA experiments from the time span 2017-03-20 to 2018-01-22 were analyzed in the thesis. Each MEA experiment contained approximately 10 minute electrical pV recordings from 60 electrodes. The combination of the preprocessing and PCA and PLS regression resulted in a study of spatiotemporal variation of detected action potentials (APs) across frequency components in multiple combinations of electrode signals. PCA was used as exploratory analysis of spatiotemporal variation of detected action potentials in individual MEA experiments, while PLS together with a variable influence on projection (VIP) method was used to compare sets of two MEA experiments based on spatiotemporal variation of detected action potentials. It was discovered that synchronized (coherent) oscillations occurring in bursts on multiple electrodes gradually develop into shorter synchronized bursts with shorter pauses as the age increases, until synchronization of APs is not apparent in the most adult culture. Age increase also leads to power increase especially in larger frequencies in the investigated frequency range 300-3000 Hz. In properly preprocessed MEA experiments of the adult culture, certain frequency components show distinct variational patterns (for example the component 2540-2550 Hz (a frequency resolution of 10 Hz was used). In such a MEA experiment, the patterns are observable on most electrodes, so the observed patterns are highly independent of physical electrode location. Which components that show distinct variational patterns differ across the relevant MEA experiments. Moreover, the lower the frequency component lays in the range $300-3000$ Hz, the more important it is to describe the difference between two experiments when each experiment represents a different age of the culture. The age discrimination is shown to be based on difference in electrical power in these frequency components, and certain electrodes (31, 32, 33, etc.) are more influential than others in the age discrimination. Conversely, the higher the frequency component lays in the range 300-3000 Hz, the more important it is to describe the difference between two experiments when each experiment represents the same age of the culture. Two adult culture MEA experiments were selected for this analysis. All electrodes seem to be equally influential in discrimination using these frequency components. Only frequency components of electrode 35 show some unique features in the discrimination. This is the same electrode to have picked up the most spiking variation in many of the analyses of younger culture.