前额叶皮层和纹状体群体神经元编码视觉刺激和奖励信息

Stimulus and reward information encoded by population neurons in the primate prefrontal cortex and striatum

以往的文献报道了前额叶皮层和纹状体中的一些神经元编码视觉刺激(visual stimulus)和奖励(reward)信息,但从单个神经元角度来看,编码信息的方式很复杂,不同的神经元在不同时刻各自编码刺激或奖励信息。本文旨在研究这两个区域的群体神经元能否在整个试验期间(in a whole trial)稳定地表征与任务相关的信息。在一个刺激-奖励预测实验中,记录了日本猕猴(Macaca fuscata)前额叶皮层和纹状体中单个神经元的活动,采用多变量回归分析和线性支持向量机分类的方法研究了群体神经元编码的信息。结果显示,尽管与任务相关的神经元的比例在整个试验中变化较大,但前额叶皮层群体神经元稳定、可靠地编码了奖励和刺激信息,纹状体群体神经元在整个试验期间也编码了奖励信息,但没有编码刺激信息。这两个区域的神经元都编码了刺激和奖励的组合信息,并且前额叶皮层的神经元编码了一组刺激(一个集合)与奖励的关系,而纹状体神经元编码了一个特定刺激与奖励的关系。以上结果提示,前额叶皮层和纹状体群体神经元都能稳定地表达任务相关的参数,但各自编码的信息是不一样的,对应于它们在价值决策中不同的作用。

It has been reported that single-unit activity in the prefrontal cortex(PFC)and striatum represented visual stimulus and reward information.But how to encode these pieces of information is quite complex from the view of single-neuron activity.Different neurons represented stimulus or reward information in different task epochs with increasing or decreasing their activities relative to their baseline firing rates.The present paper was aimed to study whether population neurons in the two brain areas could stably encode task-relevant parameters in a whole trial period.We recorded single-unit activities in the lateral PFC(LPFC)and striatum while the monkey was performing a stimulus-reward prediction task,and analyzed the neuronal activities by the method of a multi-variable regression model and the linear support vector machine.The results showed that,although proportions of task-related neurons in the two areas varied largely in the whole trial period,LPFC population neurons encoded reward and stimulus information stably and reliably.Population neurons in the striatum encoded only reward information,not stimulus information.A group of neurons in the two areas represented combined information of stimulus and reward.Further analysis showed that LPFC neurons encoded reward information for a group of relevant stimuli,while striatal neurons encoded reward information for a specific stimulus.These results suggest that both LPFC and striatal population neurons are able to stably represent task-relevant information,but from different aspects of the task.The different strategies to encode information in the LPFC and striatum suggest their different contributions in reward-based decision making.