When the eyes move fast and jerky, the brain reactions are quite different from when they fixate at a point or move slowly while reading. Visual stability is easily maintained in fixation and smooth pursuit movements, i.e. the images are placed in the receptive field of eyes without blurring. It is expected that the stability of visual perception is not to be maintained during saccades, i.e. rapid movement of the eye between fixation points. However, the brain maintains the stability of perception across saccades and there is no feeling of blur. Studies have shown that in order not to lose the continuity of the images, the brain modulates visual neuron responses by several sources, i.e. saccadic suppression, future field remapping, and saccade target remapping. This study aims to analyze and model the neural coding during saccadic eye movement. The data used in this study is the neural response of 41 neurons in the Middle Temporal(MT) cortex of two macaque monkeys, recorded while observing 81 flashing probes. At the beginning of each task, a red fixation point appeared at the center of the screen. After a randomized interval, the initial red fixation point disappeared as a cue for the monkey to make a saccade to the second red fixation point. 81 probes were also flashing during the task randomly. Due to the small number of trials, the amount of available data is limited. On the other hand, the stimulus-response relationship is time-variant during saccades. Computational models such as the adaptive filters and Generalized Linear Models(GLM) are widely used to model neural responses at fixation. But the adaptive filters require a large amount of data for proper performance, and GLM is a time-invariant system. Therefore, these two models are not suitable for modeling neural responses around saccades. A modification of GLM, named NSGLM, adds a new multiplicative block to adopt it to non-stationary dynamics of neural responses around saccades. NSGLM has been used as the basic model in this research. The firing rate of 28 neurons was estimated accurately by NSGLM. The KS-test was used as the criterion to eva‎luate the efficiency of the model. This criterion measures the agreement between the estimated firing rate and the neural responses. The modeling results showed that before and after the saccade, each neuron responded to only a limited number of 81 probes. In the 2nd simulation, the firing rate of 28 neurons was re-estimated with 50% of the parameters omitted. The results were quite acceptable. This reduction in the number of parameters reduced the execution time of the algorithm by 3.72 times. Moreover, the possibility of synapses between simultaneously recorded neurons was examined by the JPSTH method. The JPSTH results indicate that one of the neurons in this study is probably connected to two other neurons via synapses. It is important to note that the previous firing rate of this neuron was not estimated properly by NSGLM. To consider the synaptic correlation between neurons, the spike responses of those two neurons are convoluted with the new hc block and entered into the NSGLM as a new input. This innovation has improved the accuracy of the estimated firing rate of the neuron. Thus, the firing rate of 29 neurons was predicted appropriately in total. To improve the mathematical analysis of NSGLM, the nonlinear function and the basis functions were changed from exponential to sigmoid and from raised cosine to B-spline respectively.

بهزاد نظري، فرزانه شايق

نغمه سادات مويديان، سيد جلال ذهبي