Probing the flexible internal state transition and low-dimensional manifold dynamics of human brain with acupuncture

Acupuncture can improve the brain cognition and have therapeutic effects on neural disorders. However, the mechanism underlying the acupuncture for modulating functional brain activity is still unclear. Electroencephalography (EEG) signals of human subjects with different acupuncture manipulations at “Zusanli” were recorded. Machine learning and time-series analysis methods, including power spectral density, Gaussian Hidden Markov Model (GHMM), and Variational Auto-Encoder (VAE) were applied to extract dynamic features of brain activity, with the aim to probe the modulation effect of acupuncture on internal brain state and explain nonlinear dynamical response of the brain to acupuncture stimulations. Spatial and temporal analysis exhibited that brain activity in delta band (1–4 Hz) and alpha band (8–12 Hz) was enhanced significantly during acupuncture phase, especially in frontal and parietal lobe areas. Furthermore, internal states of the whole brain were inferred with GHMM, which transited along with acupuncture process. It was shown that acupuncture can activate new brain states and different acupuncture manipulations induced state transitions in independent pathways, which indicated the diversity of their regulation effects on brain activities. In addition, latent factors of multi-channel EEG signals were further estimated. Neural manifolds observed in a low-dimensional state space were periodic, which derived distinct attractor dynamics for different acupuncture manipulations and could explain the nonlinear dynamical response of the brain to external stimulation. The findings provide a new perspective to enhance the understanding of the dynamics of the human brain during acupuncture and improve its therapeutic effectiveness in clinical applications for neural disorders.