Cortical representation of sensory predictions investigated with multivariate pattern analysis and high-field fMRI

理解自己的行为与其感官结果之间的关系对于与外部世界建立正确而有意义的互动至关重要。这是通过基于运动指令的副本，即所谓的参考副本，对我们自己的动作的感官结果产生预测来实现的。与外部刺激相比，这种预测通常会导致神经处理减少，尽管有时会观察到增强。尽管对基于行动的预测机制及其神经关联的认识不断增加，但仍有一些重要问题有待解决。在涉及3T和7T fMRI实验的三个工作包中，拟议的项目将回答有关在何处以及如何生成基于动作的预测，以及这些预测如何影响感知的问题。首先，目前还不清楚基于行动的预测从何而来。使用多元模式分析，在第一个工作包中将识别携带即将到来的刺激身份信息的大脑区域，并研究与外界产生的刺激预测的异同。在第二个工作包中，刺激特征将被操纵，以研究哪些因素导致行为和神经抑制，哪些因素导致增强。这个工作包还将回答这些预测在空间上有多精确的问题。在第三个工作包中，前两个工作包中的实验将在7T MRI扫描仪中进行高场fMRI优化，由于其优越的空间分辨率，该扫描仪允许对皮质层进行分割。这将有助于识别基于行动的预测机制的哪些方面是基于自下而上的过程，哪些是基于自上而下的过程，因为这些过程涉及不同的皮层。这些知识将为基于动作的预测处理的潜在神经机制提供重要的见解。总而言之，该项目旨在提供一个全面的框架，说明我们如何对人类大脑皮层水平的感觉动作后果进行预测。

Understanding the relationship between one’s own action and its sensory outcome is crucial to establish correct and meaningful interactions with the external world. This is achieved by generating predictions about the sensory consequences of our own actions based on a copy of the motor command, the so called efference copy. Such predictions often lead to reduced neural processing compared to externally generated stimuli, although sometimes enhancement is observed. Despite increasing knowledge about action-based predictive mechanisms and their neural correlates, some important questions remain to be resolved. In three work packages involving 3T and 7T fMRI experiments, the proposed project will answer questions regarding where and how action-based predictions are generated, and how these predictions impact perception. First of all, it is still unclear where action-based predictions originate in the first place. Using multivariate pattern analysis, brain areas carrying information regarding the identity of the upcoming stimulus will be identified in the first work package, and similarities and differences with predictions about externally generated stimuli will be studied. In the second work package, stimulus characteristics will be manipulated to investigate which factors lead to behavioural and neural suppression and which to enhancement. This work package will also answer the question how spatially precise these predictions are. In the third work package, the experiments from the first two work packages will be optimised for high-field fMRI in a 7T MRI scanner, which allows for segmentation of cortical layers due to its superior spatial resolution. This will allow to identify which aspects of action-based predictive mechanisms are based on bottom-up processes and which on top-down processes, as these processes involve different cortical layers. Such knowledge will provide important insight into the underlying neural mechanisms of action-based predictive processing. All in all, the proposed project aims to provide a comprehensive framework of how we generate predictions about sensory action consequences up to the level of cortical layers of the human brain.