Decoding Rat Movement Trajectories Based On Place Cells

Decoding rat movement trajectories from hippocampal place cells is one of the front of cross subject in neuroscience and engineering technology. It has important significance for studying on the neural mechanisms of spatial navigation. In the decoding of the trajectory, the issue about establish position decoding model is one of the important research content. However, due to non-stationary, nonlinear characteristics and complexity of the neural activity of the nervous system, making decoding animal trajectory become one of the most intractable problems in neuroscience, so how to build an model and using decoding algorithm is still the core and critical in current trajectory reconstruction.This article according to the response characteristics of neurons of rat hippocampus place cells, and establish a cluster position coding model of rats trajectory, and research the performance of rats trajectory reconstruction with particle filter(PF), and compared with the extended Kalman filter(EKF) and unscented Kalman filter algorithm(UKF). The main contents include:(1)Based on the feature of place cells in rat hippocampus, establishment position encode model of a single place cell, that means position model of.place cell In order to decode the trajectory of rats, based on state space model, established a cluster location coding model of place cell.(2)This article get the spike data of place cells through simulation the place field of place cells and simulation trajectory. Measured data from a common data platform at New York University, as well as operations and training, by using mathematical morphological filtering method of detected spikes from the original data, last uses principal components analysis to classify peak potential.(3)The trajectory encoding model established in this article belong to non-linear model, first use the traditional nonlinear filtering algorithms: EKF to decode rat trajectory, because EKF linearization the non-linear function directly after the Taylor series expansion. Make the decoding accuracy in bottlenecks, To solve this problem, use the UKF by symmetrical Sigma sampling makes decoding accuracy has improved. In order to achieve more accurate decoding, this paper mainly introduces the PF algorithm, PF based on the point process estimation, denote system state using particle set, get rid of the random quantity when solving the problem of nonlinear decoding must meet Gaussian distribution, can embody the characteristics of neural information parsing coherence, improves the accuracy of trajectory decoding.(4)Evaluation index of decoding accuracy is given. By simulation data, and using the above three algorithms decode rat movement, to demonstrate feasibility of decoding algorithms. Also discusses the effect of different number of place cells on decoding accuracy.(5)Preprocess the experimental data, find the number of place cells and it’s spike data. Then decode the trajectory, calculated evaluation data. The results of simulated and measured data have shown that correlation coefficient(mean square error)(0.92(1.98), 0.94(1.64)) between decoded trajectories by the PF algorithm and real trajectories is sensible higher(lower) than by the EKF algorithm(0.71(4.51), 0.75(3.93)) and by the UKF algorithm(0.87(3.16), 0.92(2.25)) from four rats. Moreover, the number of place cells was needed by the PF algorithm is more less than the others under the same reconstruction precision.