Modelling Of Decision Brain Network Based On The Task-state FMRI Data

Risky decision-making refers to deciding in situations where the objective value or probability of options is uncertain.People can infer the value or probability of different options through feedback or personal experience.Risky decision-making,as an advanced cognitive function,integrates multiple cognitive components such as working memory and emotion and involves cooperation between multiple brain regions.Previous studies have mainly focused on the role of individual brain regions in risky decision-making,with less research focusing on constructing brain networks to study the dynamic changes of brain connections in different brain regions under different value feedbacks.We propose an improved fusion dynamic causal method based on multivariate Granger and physiological psychological interaction and construct a risky decision-making brain network using this method,focusing on studying the connection patterns of the brain during the risky decision-making process and how feedback from two types of valence signals dynamically affects the connections between brain regions.We used the dynamic causal model to build the brain network based on the task-state fMRI data of 465 subjects under the gambling task paradigm in the HCPS1200 database.Considering the limitations of this method in terms of computational complexity and hypothesis dependency,we propose an improved dynamic causal algorithm.First,a reliable prior model was constructed based on the analysis results of multivariate Granger causality and physiological psychological interaction,and then build the dynamic causal model based on the prior model.We also verify the effectiveness of the improved fusion method.It can balance the complexity and effectiveness of the model.Among the six constructed models,the variational free energy of the default connection matrix constructed by multivariate Granger causality is generally smaller than that of the fully connected model.On the connection adjustment matrix,among the twelve matrics under reward and punishment conditions,the variational free energy of nine models based on the physiological psychological interaction is less than that of the fully connected model.The posterior probability of the models based on physiological and psychological interaction effects is six higher than that of the fully connected model.The effectiveness of the improved dynamic causal method proposed in this paper is related to the number of brain regions.A stricter statistical test should be adopted when the number of brain regions is large.Higher connection numbers and weights should be artificially given for certain brain regions essential to specific cognitive functions.The modelling results of the decision brain network reveal a possible risk assessment mechanism in the brain,where reward-type value stimuli strengthen the connection between the ventromedial prefrontal cortex and the cingulate cortex to consolidate the preferences formed during the previous risky decision-making process while weakening the connection strength between the cingulate cortex and the inferior frontal gyrus to weaken the error signal generated by conflict detection and lower the risk assessment level of the current decision-making environment in the inferior frontal gyrus.In contrast,the value stimulation of punishment type is precisely the opposite.At the same time,the inhibition state of the brain connection between the inferior frontal gyrus and other brain regions under value stimulation may provide a new explanation for the cognitive function of this brain region in task switching and response inhibition.