Single neuron firing properties impact correlation-based population coding.
Single neuron firing properties impact correlation-based population coding.
复制标题
DOI：
10.1523/jneurosci.3735-11.2012
复制
发表时间：
2012-01-25
期刊：
影响因子：
--
通讯作者：
De Schutter E
中科院分区：
文献类型：
--
作者：
Hong S;Ratté S;Prescott SA;De Schutter E
相关的尖峰信号已经被广泛观察到，但它对神经编码的影响仍然存在争议。由神经元之间的速率共同调制产生的相关性已被证明随着单个神经元的放电速率而不同。这意味着频率和相关性与刺激的调谐是等价的；在这些条件下，相关的尖峰放电不会提供除了单个神经元放电速率已有的信息之外的信息。这种相关性是不相关的，并且会因为引入冗余而降低编码效率。利用模拟和在大鼠海马神经元上的实验，我们在这里表明，接收相关输入的神经元对也显示出源于精确的尖峰时间同步的相关性。与速率共调制相反，尖峰时间同步不受发射速率的影响，从而使基于同步和基于速率的编码能够独立运行。输出相关的类型取决于神经元的内在属性是否促进积分或符合检测：“理想”积分器(尖峰产生对刺激均值敏感)表现出速率共调制，而“理想”符合检测器(尖峰产生对刺激方差敏感)表现出精确的尖峰-时间同步。锥体神经元对刺激均值和方差都很敏感，因此表现出两种类型的输出相关性，这两种类型的输出相关性与哪种操作模式占主导地位成比例。我们的结果解释了不同类型的相关性是如何根据单个神经元产生尖峰的方式产生的，以及为什么尖峰时间同步和速率共同调制可以编码不同的刺激属性。我们的结果也强调了神经元属性对于种群水平编码的重要性，因为神经网络可以根据其组成神经元的主要操作模式采用不同的编码方案。
Correlated spiking has been widely observed but its impact on neural coding remains controversial. Correlation arising from co-modulation of rates across neurons has been shown to vary with the firing rates of individual neurons. This translates into rate and correlation being equivalently tuned to the stimulus; under those conditions, correlated spiking does not provide information beyond that already available from individual neuron firing rates. Such correlations are irrelevant and can reduce coding efficiency by introducing redundancy. Using simulations and experiments in rat hippocampal neurons, we show here that pairs of neurons receiving correlated input also exhibit correlations arising from precise spike-time synchronization. Contrary to rate co-modulation, spike-time synchronization is unaffected by firing rate, thus enabling synchrony- and rate-based coding to operate independently. The type of output correlation depends on whether intrinsic neuron properties promote integration or coincidence detection: “ideal” integrators (with spike generation sensitive to stimulus mean) exhibit rate co-modulation whereas “ideal” coincidence detectors (with spike generation sensitive to stimulus variance) exhibit precise spike-time synchronization. Pyramidal neurons are sensitive to both stimulus mean and variance, and thus exhibit both types of output correlation proportioned according to which operating mode is dominant. Our results explain how different types of correlations arise based on how individual neurons generate spikes, and why spike-time synchronization and rate co-modulation can encode different stimulus properties. Our results also highlight the importance of neuronal properties for population-level coding insofar as neural networks can employ different coding schemes depending on the dominant operating mode of their constituent neurons.