Decoding Luminance Information From A Population Of The Optic Tectum In Pigeons

Brightness is the most basic information received by animals’ visual system and the detection of luminance information is the basis of dealing with other information for visual system. Therefore, the studies of coding and decoding luminance information in the visual system have a heuristic significance for us to understand the working mechanism in the visual system. The tectofugal pathway is the most important way for birds to process visual information in which the optic tectum(OT)is a key structure for processing visual information. OT also contributes to modulate,code and integrate visual information. The study of OT information processing mechanism plays a vital role for clearly understanding the birds’ ability to process visual information.In this thesis, pigeons were chosen as model animals. The responses signal of neural population in OT were recorded simultaneously using a multielectrode array,the receptive field of OT neurons was determined, the spike firing rates of neural population were extracted under time-varying brightness stimulus, two decoders including multivariate inverse linear filtering and neural network were designed, the performance of decoders was evaluated from the perspective of information theory and the neural mechanism that how visual luminance information is encoded by a population of tectal tectum(OT) neurons in pigeons was analyzed. The main results are listed as follows:(1) The receptive field of neurons in OT was tested.The checkerboard stimulation pattern was combined with spike-triggered average to determine the receptive field and positions were verified. The experimental results showed that: the receptive field of neurons measured in this test focus on the top left of screen so that the effective neurons are determined.(2) The visual luminance information encoded by neural population in OT was decoded.First, the visual luminance stimulus consisted of spatially uniform and transientcharacteristics of flicker was designed. Second, spike firing rates of neural population under different brightness stimulus were extracted. Multivariate linear inverse filters and artificial neural network were constructed to decode visual stimulus intensity. The actual stimulus was compared with the stimulus constructed and decoding accuracy was measured using normalized cross-correlation coefficient,realizing that these two methods effectively decode luminance information and decoding results are similar. Then, the influence number of neural population, the time bin, the duration of response from stimulus onset and the time from stimulus onset on decoding results was respectively analyzed under two kinds of decoding methods. Some results were got:(1) Similarity between reconstruction and actual stimulus increases with the increase of number of neurons. When the number of neurons increases to about 13, reconstruction results are best. But the cross-correlation coefficient between reconstruction and actual stimulus is not significantly improved even though continuing to increase the number of neurons.(2)When the time bin is 5ms, the cross-correlation coefficient reaches the highest.(3)The cross-correlation coefficient increases gradually when the duration of response from stimulus onset is within 35 ms and it begins to stabilize after 35 ms.(4)When the time from stimulus onset is after 15 ms, the cross-correlation coefficient fall sharply.Finally, the influence parameters varying on decoding results was analyzed when stimulus appeared(on) and stimulus removed(off). The results showed that there is no significant difference when the number of neurons and the time bin are considered; neurons fires longer time and quicker starting time when stimulus appears, but neurons fires shorter time and slower starting time when stimulus removes.(3) The decoding results were evaluated by information theory from multiple views and the visual stimulus characteristics encoded effectively by spike trains from neural population in OT were determined.First, power spectrum of the reconstruction was calculated and it was found that power of the reconstruction drops off both at low frequency and at high frequency;the reconstruction is of higher power between 20 Hz and 60Hz; when frequency is greater than 70 Hz, the reconstruction power essentially vanishes and thus decodersextract no information about stimulus. Then, the information spectral density curve of the reconstruction was calculated and it was served as a judging standard to analyze information redundancy and saturation of information transfer in the process of OT encoding brightness information for neural population. Results showed that certain amount of redundant information is passed when neurons in OT code visual brightness stimulus. When the number of neurons increases to a certain number,information transferred in the process of encoding reaches saturation; similar luminance information is received by two decoding methods, suggesting that most of the information about this stimulus can be extracted by linear operations on the spike trains. Finally, the shape of single-neuron filter in population with different numbers was calculated, finding that the synergy of neural population has a strong influence on the shape of filters which indicates that the meaning of an action potential in coding depends strongly on the messages received from other neurons.