Correlation Analysis Of Firing Activities Of Retinal Ganglion Cells

Retina, as the first stage of visual pathway, converts the optical visual signal to electrical signals and arouses action potentials in ganglion cells, which then transfer the visual information to higher perceptive cortex. It is of great significance to study the response properties and coding methods of ganglion cells, which helps to explain the working mechanism of visual system. In this thesis, the firing activities of multiple ganglion cells on chicken and bullfrog retina recorded by multi-electrode extracellular recording system are analyzed, and the correlation activities of neuron population under different stimuli are discussed.First, based on the data that was recorded from chicken retinal ganglion cells during response to natural movies and checkerboard, cross-correlation and Lempel-Ziv distance, which measure the timing and pattern correlation between spike trains respectively, are calculated for both single neurons and neuron population. The result shows different firing patterns of single neurons in response to natural movies and checkerboard, and higher timing and pattern correlation between neurons in response to natural movies. The synchronized degree between neurons revealed by cross-correlation is closely related to the physiological connections, while the similarity of firing patterns reflected by Lempel-Ziv distance shows the difference of stimulus properties more obviously. The result indicates that neurons can change their firing patterns and correlation strength dynamically to code stimuli in optimal ways.Second, the multi-electrode extracellular recording experiments were designed and implemented on bullfrog retinal ganglion cells to measure the coding features of different stimulus statistics. Low order stimulus statistics such as light intensity and contrast are clearly reflected in single neuron's firing rates and temporal response properties. However, neurons are capable of handling tougher tasks in coding complex stimulus. So the neural responses to higher order statistics as skewness and kurtosis are investigated, and the role of correlation activity in coding stimulus statistics are discussed. The result indicates that single neurons'firing properties fail to reflect the difference of skewness and kurtosis as obviously as intensity and contrast. But correlation in neuron population tells the difference of skewness, and the more neurons are involved in correlated activity, the more obviously they show the stimulus difference. As a result, neural population activities may be more reliable and powerful in coding stimulus statistics.Compared to single neurons, neurons'correlation activities reveal the stimulus property more obviously, which may be an important coding mechanism for neurons to process and transfer the complex stimulus in more effective and reliable way.