| Decision-making is an important cognitive function of the human brain which runs through the entire life cycle of an individual.It can be divided into three processes,including stimulus information coding,motion behavior selection and execution and feedback learning.Although event-related potentials,such as readiness potentials(RP),provide a window to observe brain activities during decision-making,the functions of it are completed by multiple brain regions.It is difficult to conduct functional and structural studies from neurons to neural networks to multiple levels of the brain with existing research techniques.With this in mind,this paper establishes a bio-realistic basal ganglia(BG)neural circuit model,and designs the corresponding reinforcement learning experimental paradigm to obtain EEG and behavioral data.Then,with the psychological drift diffusion model(DDM)as a bridge,the correlation between RP,behavioral data and activity of BG brain region is systematically analyzed,and the connection between basal ganglia decision circuit and RP is established.The main contents include:(1)The important biological characteristics are instantiated after analyzing the structure and function of the basal ganglion nerve circuit,such as neuron activation function,two-way propagation of neural network information,network error-driven and self-organizing learning rules,and dopamine reward and punishment mechanism,etc.Based on the emergent software,the neural circuit decision computation model of cortex-basal ganglia-thalamus-cortex(CBGTC)related to behavior selection is constructed.With this model,the simulation experiments about binary choice task are carried out.The activities of the neurons in each layer of the model are observed.The results show that the model satisfies the main biological constraints and functions.(2)After designing a probabilistic reinforcement learning experiment paradigm,we set three different conflict levels to carry out simulations of the CBGTC model and human experiments to collect the discharge rate of neuronal groups and behavior data of the simulations,EEG and behavior data of the human experiments.Firstly,the traditional statistical analysis is used to compare the simulated and the experimental behavior data.It is found that the trends of these two sets of data are similar.As the conflict increases,the accuracy decreases and the response time increases,which further validates that this model is able to predict accuracy and response time qualitatively.In addition,according to the CBGTC model,two levels of dopamine(DA)and subthalamic nucleus(STN)are adjusted.The results show that the response time prolonged with the increase of the level of conflict and the STN.The response time decreased with the increase of DA levels.The accuracy rate is greatly affected by the conflict,and the correlation with the other two factors is not significant.(3)To reflect the adaptive response of the decision process to evidence and the trade-off between decision speed and accuracy,and reveal its potential neuro-processing mechanism,The DDM is used to fit the behavior data generated by the CBGTC model with changing the three factors.As a result,we can build the relationship between the three factors and the lparameters of the DDM,including decision boundary,drift rate and non-decision time.The results show that the decision boundary is positively correlated with the level of STN and negatively correlated with the level of DA.The drift rate is negatively correlated with the conflict level.(4)The RP is extracted by the methods of data preprocessing and averaging,and the lateralized readiness potential-stimulus locked(LRP-S)and the lateralized readiness potentialresponse locked(LRP-R)are calculated.After analyzing LRP and the discharge rate of STN,we find that the negative drift slope of LRP-S is positively correlated with the drift rate,which indicates that LRP-S is related to stimulation and reflects the level of the conflict,and the negative peak of LRP-R is inversely related to STN activity and positively correlated with DA level,which confirms that patients with Parkinson’s disease have the lower RP amplitude and the slower action due to lower DA level and in turn explains the treatment of reducing the discharge rate by using STN.This findings suggest that we may be able to identify early PD through RP in the future.In summary,this paper builds a decision-making neural circuit calculation model based on the CBGTC.Through the DDM,the bridge between the CBGTC model and the RP is set up,and the correlation between the decision model and the RP is obtained which provides a new idea for decision-making research. |
