Department of Surgical, Medical, Molecular Pathology and Critical Medicine, University of Pisa, Pisa, Italy.
Coma Science Group, GIGA-Consciousness, University of Liège and University Hospital of Liège, Liège, Belgium.
J Sleep Res. 2020 Oct;29(5):e13117. doi: 10.1111/jsr.13117. Epub 2020 Jun 27.
We investigated changes of slow-wave activity and sleep slow oscillations in the night following procedural learning boosted by reinforcement learning, and how these changes correlate with behavioural output. In the Task session, participants had to reach a visual target adapting cursor's movements to compensate an angular deviation introduced experimentally, while in the Control session no deviation was applied. The task was repeated at 13:00 hours, 17:00 hours and 23:00 hours before sleep, and at 08:00 hours after sleep. The deviation angle was set at 15° (13:00 hours and 17:00 hours) and increased to 45° (reinforcement) at 23:00 hours and 08:00 hours. Both for Task and Control nights, high-density electroencephalogram sleep recordings were carried out (23:30-19:30 hours). The Task night as compared with the Control night showed increases of: (a) slow-wave activity (absolute power) over the whole scalp; (b) slow-wave activity (relative power) in left centro-parietal areas; (c) sleep slow oscillations rate in sensorimotor and premotor areas; (d) amplitude of pre-down and up states in premotor regions, left sensorimotor and right parietal regions; (e) sigma crowning the up state in right parietal regions. After Task night, we found an improvement of task performance showing correlations with sleep slow oscillations rate in right premotor, sensorimotor and parietal regions. These findings suggest a key role of sleep slow oscillations in procedural memories consolidation. The diverse components of sleep slow oscillations selectively reflect the network activations related to the reinforced learning of a procedural visuomotor task. Indeed, areas specifically involved in the task stand out as those with a significant association between sleep slow oscillations rate and overnight improvement in task performance.
我们研究了在强化学习促进的程序性学习之后，夜间慢波活动和睡眠慢波振荡的变化，以及这些变化如何与行为输出相关。在任务会话中，参与者必须到达视觉目标，通过适应光标运动来补偿实验中引入的角度偏差，而在控制会话中则不应用偏差。任务在睡前的 13:00 小时、17:00 小时和 23:00 小时以及睡眠后的 08:00 小时重复进行。偏差角度设定为 15°（13:00 小时和 17:00 小时），在 23:00 小时和 08:00 小时增加到 45°（强化）。对于任务和控制之夜，都进行了高密度脑电图睡眠记录（23:30-19:30 小时）。与控制之夜相比，任务之夜表现出以下变化：（a）整个头皮上的慢波活动（绝对功率）增加；（b）左中央顶区的慢波活动（相对功率）增加；（c）感觉运动和运动前区的睡眠慢波振荡率增加；（d）运动前区、左感觉运动区和右顶区的预下和上状态振幅增加；（e）右顶区上状态的 sigma 冠。任务之夜后，我们发现任务表现有所改善，与右运动前、感觉运动和顶区的睡眠慢波振荡率相关。这些发现表明睡眠慢波振荡在程序性记忆巩固中起着关键作用。睡眠慢波振荡的不同组成部分选择性地反映了与强化学习程序性视觉运动任务相关的网络激活。实际上，特别涉及任务的区域突出显示了睡眠慢波振荡率与夜间任务表现改善之间的显著关联。
