2025年10月28日出版的《自然—神经科学》杂志发表了英国科学家的一项最新研究成果。来自英国伦敦帝国理工学院的Nir Grossman团队的研究显示，入睡遵循一个可预测的分岔动态。

在这里，研究人员报告了一个新的概念框架来分析和模拟这一现象。该框架将进入睡眠期间的脑电图活动的变化表示为标准化特征空间中的轨迹。该团队的主题框架表明，大脑的觉醒到睡眠的转变遵循分岔动力学与一个明显的临界点之前的一个关键减速。研究组在两个独立的数据集中验证了分岔动态，其中包括1000多名人类参与者。最后，课题组研究人员通过以秒时间分辨率和超过0.95的平均精度实时预测一个人进入睡眠的进程来证明该框架的实用性。

据悉，睡眠是他们生活的基本组成部分；然而，他们的大脑如何入睡仍然是神经科学中最持久的谜团之一。

附：英文原文

Title: Falling asleep follows a predictable bifurcation dynamic

Author: Li, Junheng, Ilina, Anastasia, Peach, Robert, Wei, Tianyu, Rhodes, Edward, Jaramillo, Valeria, Violante, Ines R., Barahona, Mauricio, Dijk, Derk-Jan, Grossman, Nir

Issue&Volume: 2025-10-28

Abstract: Sleep is a fundamental part of our lives; yet, how our brain falls asleep remains one of the most enduring mysteries of neuroscience. Here we report a new conceptual framework to analyze and model this phenomenon. The framework represents the changes in brain electroencephalogram activity during the transition into sleep as a trajectory in a normalized feature space. We use the framework to show that the brain’s wake-to-sleep transition follows bifurcation dynamics with a distinct tipping point preceded by a critical slowing down. We validate the bifurcation dynamics in two independent datasets, which include more than 1,000 human participants. Finally, we demonstrate the framework’s utility by predicting a person’s progression into sleep in real time with seconds temporal resolution and over 0.95 average accuracy.

DOI: 10.1038/s41593-025-02091-1

Source: https://www.nature.com/articles/s41593-025-02091-1