The Neural Basis of Functional MRI Responses

去年，我的实验室发表了三篇与自发功能磁共振信号的神经基础相关的论文(Leopold和Maier，NeuroImage，2012；Hutchison et al.NeuroImage，2013；Scholvinck et al，NeuroImage，2013)。这些论文的要点是，通常用于评估人脑连接结构的自发fMRI相关性表现良好，与电生理信号有关，但它们的起源仍然神秘。通过最近的两项研究，我们在这个问题上也取得了实验进展。在Spaak等人(Current Biology，2012)中，我们证明了静息状态电生理过程表现出一种层特定的耦合，包括与功能磁共振相关的时间尺度。同样，在一项合作研究(Fukushima et al，Neuron，2012)中，我们证明了在休息时在听觉皮质观察到的自发活动的空间模式与正常声刺激时发现的非常相似。这些发现是自发电生理相关性和fMRI静息状态相关性之间存在联系的证据。 在一个新的项目中，我们已经开始调查相关的自发fMRI波动是否不仅源于两个区域之间的直接联系，还源于基于基底前脑输入的区域的共享神经支配。虽然解剖学上的联系显然会限制活动相关性的模式，但它们的存在并不能解释为什么自发活动一开始就会出现缓慢的波动。一种可能性是，来自几个基底前脑结构的深远的调节性胆碱能和GABA能输入共同驱动皮质兴奋性高和低的时期，这将导致在人类fMRI中通常测量的功能相关性模式。我们计划的第一步是测试这一假说，包括将几个基底前脑结构的失活与休息时和稍后自然观看时的fMRI测量结合起来。这项实验的零假设是明确的：这些投射神经元的失活应该不会对皮质表面上的相关波动模式产生影响。 在另一个项目中，我们询问了在高级视觉皮质中使用fMRI观察到的反应在多大程度上类似于使用微电极记录在同一脑区观察到的神经反应。我们使用自然观看社交视频的方式进行了这个项目，记录了功能磁共振成像和电生理反应，而猴子则重复观看了一些五分钟的视频片段。使用这种方法，我们测量了微电极放电数据和血流动力学fMRI数据中每部电影诱发的一致时间过程。鉴于fMRI信号中的每个体素包含数十万个神经元，这些结果提出了一个问题：体素中fMRI信号的活动时间进程在多大程度上反映了该体素内的尖峰时间进程？作为第一步，我们评估了微丝阵列中相邻神经元之间的相关性。神经元之间的距离不超过400微米，因此是同一皮质柱的一部分。我们发现，尽管单个细胞非常可靠地被同一电影的多次重复所驱动，但附近的细胞在它们的反应中却显著地不相关。我们计划的这项工作的扩展是检查集体神经活动的某个方面，例如体素中所有神经元的平均活动，或者特定的、已识别的神经元的活动，是否最能反映血流动力学的时间进程。人们希望这种方法不仅能让我们了解大脑这一部分的编码原理，还能了解这种编码是如何反映在fMRI信号中的。 最后，我们已经开始了另一项实验，在这项实验中，我们测量了植入的磁共振兼容电极在接受功能磁共振的猴子身上的单个单位活动。如上所述，相邻神经元在自然观看过程中表现出非常不同的时间进程。在自发活动中也是如此。因此，一项计划中的实验是确定两个显示不同时间进程的神经元是否在功能上与大脑中不同的功能磁共振网络有关。因此，使用两个不同神经元的尖峰波动来揭示两个不同的网络可能会阐明不同的大脑网络对受限皮质区域的局部神经机制的相对贡献。

In the last year, my laboratory has published three papers related to the neural basis of the spontaneous fMRI signal (Leopold and Maier, NeuroImage, 2012; Hutchison et al. NeuroImage, 2013; Scholvinck et al, NeuroImage, 2013). The gist of these papers is that the spontaneous fMRI correlations that are commonly used to evaluate the connectional structure of the human brain are well-behaved and link to electrophysiogical signals, but their origins remain mysterious. We also have made experimental headway on this topic with two recent studies. In Spaak et al, (Current Biology, 2012), we show that resting state electrophysiological processes exhibit a layer-specific coupling, including over time scales relevant for fMRI. Likewise, in a collaborative study (Fukushima et al, Neuron, 2012), we demonstrate that the spatial patterns of spontaneous activity observed in the auditory cortex during rest closely resemble those found during normal acoustic stimulation. These findings are present evidence of a link between spontaneous electrophysiological correlations and fMRI resting state correlations. In a new project, we have begun to investigate whether correlated spontaneous fMRI fluctuations may have their origins not only in the direct connections between two areas, but also in the shared innervation of areas based on input from the basal forebrain. While it is clear that anatomical connections will constrain patterns of activity correlations, their existence does not explain why slow fluctuations in spontaneous activity emerges in the first place. One possibility is that the far-reaching modulatory cholinergic and GABAergic inputs emanating from the several basal forebrain structures drives periods of high and low cortical excitability in concert, which would result in a pattern of functional correlations commonly measured in human fMRI. Our planned first steps to test this hypothesis involves combining inactivation of several basal forebrain structures with fMRI measurements during rest, and later during natural viewing. The null hypothesis of this experiment is clear: inactivation of these projection neurons should have no effect on the pattern of correlated fluctuations over the cortical surface. In another project, we asked to what extent responses observed in high-level visual cortex using fMRI would resemble neural responses observed in the same piece of cortex using microelectrode recordings. We approached this project using natural viewing of social videos, recording both fMRI and electrophysiological responses while monkeys repeatedly watched a number of five-minute video clips. Using this method, we measured a consistent time course evoked for each movie in the microelectrode spiking data and the hemodynamic fMRI data. Given that each voxel in the fMRI signal contains hundreds of thousands of neurons, these results raise the question: to what extent does the activity time course of the fMRI signal in a voxel reflect the time course of spiking within that voxel? As a first step, we have assessed the correlation between adjacent neurons within our microwire array. Neurons were separated by no more than 400 microns, and were thus part of the same cortical column. We found that, even though individual cells are very reliably driven by multiple repetitions of the same movie, nearby cells are remarkably uncorrelated in their responses. Our planned extension of this work is to examine whether some aspect of collective neural activity, such as the mean activity of all neurons in a voxel, or the activity of special, identified neurons, most closely reflects the hemodynamic time course. The hope is that this approach can not only allow us to understand about the coding principles in this part of the brain, but also about how this coding is reflected in the fMRI signal. Finally, we have begun another experiment in which we measure single unit activity from implanted, MR-compatible electrodes, in monkeys that are undergoing fMRI. As described above, adjacent neurons show very different time courses during natural viewing. The same is true during spontaneous activity. Thus one planned experiment is to determine whether two neurons showing different time courses are functionally linked with different fMRI networks in the brain. Thus using spiking fluctuations of two different neurons to reveal two different networks may elucidate the relative contributions of different brain networks to the local neural machinery in a confined cortical area.