前庭-视觉跨模态适应对认知影响的神经机制

Astronauts experience similar sensations of dizziness and disorientation during their first few days in the microgravity environment of space. Upon returning to Earth after prolonged exposure to microgravity, astronauts frequently have difficulty standing and walking upright, stabilizing their gaze, and walking or turning corners in a coordinated manner. An astronaut's sense of balance and body orientation takes time to re-adapt to Earth-normal conditions. While studies have identified the source of these changes to occur at the level of the central nervous system, the basic mechanisms underlying multisensory perceptual adaptation are currently unclear. To what extent does multisensory adaptation follow Bayesian-like (cue-reliability dependent) calibration? How does motivation/reward (reinforcement learning) influence multisensory adaptation? What is the neuronal substrate for multisensory adaptation? In this study I shall address these questions through psychophysics experiments and electrophysiological recording in non-human primates. The subject’s task will be to discriminate heading direction experienced during either single-cue or combined presentation of vestibular and visual (optic flow) input. During the trials, multisensory adaptation will be induced by introducing a discrepancy between the two (visual/vestibular) cues, and studied in the presence and absence of reinforcement. I hypothesize that dynamic adaptation of multisensory integration is driven by a combination of Bayesian theory and reinforcement learning, and intend to model it accordingly. Furthermore, I expect to find the neural correlate of Bayesian vs. reinforcement based adaptation in the cerebellum and basal ganglia, respectively. The dorsal medial superior temporal (MSTd) cortex, which has recently been associated with visual-vestibular integration, most likely integrates sub-cortical Bayesian and reinforcement-based adaptation, thereby manipulating the subject’s heading perception. The results of the proposed research will guide future investigations of sensorimotor changes in astronauts and assist their adaptation to life on earth.

宇航员进入太空之后会产生空间运动病，持续2~4天后症状可自动消失，而当宇航员从长期航天飞行环境回到地面后会发生前庭功能暂时紊乱，需要一段时间来重新适应地面生活所需要的平衡和空间定向能力。显然在这一过程中，来自前庭系统对重力变化的适应引发了其他系统（如视觉、体感等）产生相应的适应来抵消不同感觉之间冲突造成的不适。但是不同感觉之间适应是如何相互影响的神经机制尚不清楚。本项目通过在非人灵长类身上对前庭-视觉错觉适应的神经机制进行探索，将基于贝叶思理论的最优知觉行为整合理论和基于强化学习理论的奖励决策制定/行为选择结合起来，探索视觉、前庭觉及其相互作用影响认知加工的动态变化规律。这些研究将有助于对空间运动病及空间定向障碍中枢整合机制的了解，为探索新的航天员前庭功能训练方法提供理论依据。

“运动病”，也称为“晕动病”，即人们通常所说的晕船、晕车、晕空（乘飞机或高空作业头晕等）。目前广泛认可的发病机制是感觉冲突，尤其是视觉-前庭信息之间的冲突。为了探索视觉-前庭冲突的神经机制和两种信息的相互作用规律，我们对非人灵长类皮层视后裂区（VPS）在前庭、视觉以及不同视觉前庭组合下的神经元活动进行电生理记录分析，并对单个神经元放电以及群体神经元放电进行模型拟合，结果发现：1）与颞上内侧区（MSTd）相比，视后裂区（VPS）中的神经元放电以前庭为主，前庭和视觉一起作用时视觉反应被强烈抑制，而前庭反应有轻微的抑制，而MSTd中视觉前庭的相互作用呈现弱加性；2）对单模态的反应进行简单线性加和虽然能较好地预测MSTd中的多模态反应，但并不能很好地预测VPS中的多模态反应，部分原因是VPS中前庭反应在视觉刺激下会发生轻微的偏移；3）群体均一化模型可以比较好地同时解释MSTd和VPS中的视觉-前庭相互规律，当群体视觉和前庭反应相当时（如MSTd的视觉前庭反应）出现弱加性，而当两者差异较大时（如VPS的前庭视觉反应）则出现胜者全得（winner-take-all）的现象。通过这些研究，我们进一步理解了视觉和前庭信息在皮层相互作用的神经机制。