Research On The Encoding Model Of Continuous Motion-sensitive Neurons In The Intermediate And Deep Layers Of The Optic Tectum

There are a kind of continuous motion-sensitive neurons in the intermediate and deep layers of the Optic Tectum(OT)of birds and the superficial Superior Colliculus(s SC)of mice,and the lobula complex of the insects.This kind of neurons are usually sensitive to small targets with continuous motion,but not sensitive to stationary objects,sudden appearances or full-field stimulation.The local inhibition produced by suppressor cells in the same tissue area may shape their target size selectivity,for example,horizontal cells.In addition,compared with the classical information integration mechanism of neuron,that is,the soma is often driven by the current obtained by the weighted sum of all dendritic inputs,a single dendrite of the continuous motion-sensitive neuron has the ability to induce the somatic spikes,which means they are more suitable for the detection of local small stimuli.Integrating and learning from these three types of continuous motion-sensitive neurons have discovered neural mechanisms,and studying their information encoding methods will deepen the understanding of the mechanism of biological perception of small motion targets,and have important theoretical and practical significance for the construction of new brain-like computing models.Aiming at the characteristics and information integration of continuous motion-sensitive neurons,this paper used pigeons as experimental animals to collect the neural signals of continuous motion-sensitive neurons in the intermediate and deep layers of the Optic Tectum(800-1500?m)through electrophysiological experiments.Taking into account the sensitivity of small targets and the preference for continuous motion described in the literature,multi-size stimulus and continuous sensitivity stimuli were designed for physiological verification.Subsequently,the hypothesis of the directed energy accumulation of the dendritic field of this kind of neurons and the sequential probabilistic activation hypothesis of the soma were proposed respectively for this kind of neurons' preference for continuous motion and the ability of single dendrites to induce somatic spikes.Combining these two hypotheses and other related neural mechanisms,the information encoding model of this kind of neurons was constructed.Then,the results of the encoding model on multi-size stimulus and continuous sensitivity stimuli were compared with the physiological experiments,and a qualitatively acceptable fitting result was obtained.After proper testing and analysis of the encoding model and rationalization of the hypothesis,a suitable scene was designed to realize the population decoding of the encoding neurons' somatic spikes,and the deviation between the decoded trajectory and the actual trajectory was only about 1.5°.The main work and research results of this paper are as follows:(1)Design of the experimental procedures and preparation of the electrophy-siological experimentsThe experiment plan and the electrophysiological signal collection work were designed and prepared,including writing the suitable stimulation program and determining the parameter adjustment range before the experiment,selecting healthy pigeons as the experimental animals,performing acute surgery and implanting the microelectrode array,and the pre-processing of the neural signals,etc.(2)Verification of the characteristics of continuous motion-sensitive neuronsFirstly,combined with movement point stimulus,sparse noise stimulus and ON/OFF stimulus,it is analyzed whether the neurons collected by the microelectrode array were continuous motion-sensitive neurons.For those neurons that have relatively large motion receptive fields,sparsely small sparse receptive fields,and are not sensitive to ON/OFF stimulus,they were identified as the continuous motion-sensitive neurons needed in this paper.After that,multi-size stimulus and continuous sensitivity stimuli were used to test whether the continuous motion-sensitive neurons collected were sensitive to small targets and continuous motion,and the numbers of somatic spikes were counted and their tuning law were obtained.(3)Construction of the encoding model of continuous motion-sensitive neuronsFirst,the simulation rules for the generation of this kind of neurons' dendritic field was given,then FSL(First-Spike Latency)model and the inhibitory effect of the simulated suppressor cells were combined to obtain the information processing from visual stimulation input to the integration of the dendrites of the continuous motion-sensitive neuron to generate the dendritic spikes.Aiming at the characteristics of continuous motion-sensitive neurons prefer continuous motion,the hypothesis of directed energy accumulation of dendritic field was proposed.Aiming at the ability of single dendrites to induce somatic response,the hypothesis of sequential probabilistic activation of soma was proposed.Based on the above steps,a complete simulation process from visual information input to somatic response of continuous motion-sensitive neuron was constructed.(4)The analysis of test and decoding of the encoding modelAfter adjusting the model parameters,the model's response to multi-size stimulus and continuous sensitivity stimuli were first tested,and the numbers of the somatic spikes were counted to obtain its response tuning curve.Based on the existing experimental data,the encoding model qualitatively fitted the physiological response characteristics of continuous motion-sensitive neurons.After testing the influence of dendritic field distribution and directed energy accumulation on the response of the encoding model,a suitable scene was designed to realize the population decoding of the simulated soma response,and the deviation between the decoded trajectory and the actual trajectory was only about an average of 1.5°.