Mapeamento topológico virtual de neurônios proporcional às atividades eletrofisiológicas em matrizes de microeletrodos

Abstract

This thesis combines image and signal processing to obtain virtual neuron distribution maps in a Microelectrode Array (MEA), which are devices designed for non-invasive electrophysiological signal recording for in vitro cultures of neuron cells. In the electrophysiological signal analysis, it is of interest the knowledge of the topological distribution of the cells along the MEA microelectrodes, but, usually the photographic images of the cell culture are not available. This doctoral work presents an approach to obtain the statistical topologic distribution of the neurons of an in vitro cell culture, denoted virtual distribution of neurons, from the electrophysiological signals. To certify that the statistical computation of the neuron counting is associated to each MEA microelectrode, it is used the ICA (Independent component Analysis) technique, for the separation of the neuron signals distributed throughout the MEA area, to obtain for each microelectrode, only the signals from its adjacent neurons. Assuming the hypothesis that the spontaneous neuron activities, spikes and bursts, are directly proportional to the neuron counting, it is realized the spike counting and burst counting, and it is assigned for each microelectrode, a number of neurons proportional to that numbers of activities. For the validation of the proposal, as well as for calibration of the system, to obtain the estimated number of neurons, it was used an experiment denoted 371, realized in Genoa University, Italy, in which it was recorded electrophysiological signals in 46 DIVs (Days In- Vitro), obtaining 20 minutes of recording in 25, 29, 32, 36, 39, 43, and 46 DIVs, and a set of photographic images in 38 DIV. Assuming that microelectrode neuron counting in the 38 DIV photographic image is proportional to the 39 DIV spontaneous electrophysiological activity signal recording, one day after the imaging, if was determined the neuron counting as function of the spontaneous electrophysiological activities recording, in a process denoted as calibration of the virtual number of neurons. The distance error from the neuron activities as function of the neuron counting in photographic image and in function of the recorded electrophysiological signals was calculated and compared for validation. In this way, it was possible to construct virtual topologic maps of neurons, proportional to the electrophysiological activities measured in 39 DIV, as a function of the spike and the burst countings. Comparing these two virtual maps, the spike counting virtual map was more close to the real neuron distributions viewed at the photographic image of 38 DIV. Also, the variance of the spike and burst counting along the 20 min of electrophysiological recording in a DIV, was calculated, and noted that the spike counting is more stable than burst counting.