日時: 2015年7月17日(金)13:00-14:30
場所: 電気通信大学 東4号館802会議室
講師: 田嶋 達裕氏(ジュネーブ大学 Lab of Cognitive Computational Neuroscience 博士研究員)
司会: 宮脇陽一 准教授
題目: Untangling complex brain dynamics(複雑な脳のダイナミクスを”解きほぐす”)
概要: The brain is a complex dynamical system. While a variety of innovative large-scale recording technologies are yielding a plethora of biological data at unprecedented spatiotemporal resolution, it may sometimes seem hopeless to derive a reduced, intuitive model from the nonlinear, heterogeneous and high-dimensional dynamics of a large neural population. However, extracting simple structures that govern complex neurophysiological phenomena is still possible by focusing on their deterministic aspects—constraints on the neural signals in the temporal domain. I will present two studies where theoretically-motivated analyses of intrinsic brain dynamics shed light on the underlying neural processing mechanisms.
First, using a nonlinear mapping from neural state to stimulus space, we find flexibly-modulated attractor dynamics during task-switching in monkey visual cortex. This is in contrast to a prevailing view that dynamics outside of sensory cortices solely account for the flexible sensory-action association. The temporal evolution of neural modulation is not explained by static gain-control mechanisms, suggesting an involvement of sensory cortex in recurrent dynamics that underlies the flexible perceptual abilities.
Next, I introduce a more generic method based on dynamical systems theory, which is capable of analyzing large-scale and heterogeneous neural interaction. Applying this method to whole-brain electrocardiography data from behaving monkey reveals a universal relationship between dynamical complexity and area-to-area interaction, which dissociates conscious from unconscious brain states. Remarkably, the method captures state-dependent structures of cross-area interaction that can be missed by conventional correlation-based analysis. These results underscore the importance of intrinsic dynamics in understanding complex neuronal systems.