Energy-Based Phase-Locking State Analysis in Brain State Identification.

A new computational framework called Energy-based Phase-Locking State Analysis (EPLSA) integrates phase-coupling dynamics with energy landscape principles to analyze brain network states from fMRI data. Across two independent datasets, EPLSA outperformed existing methods in reliability, task-specific state differentiation, and individual classification. Sleep-wake analysis showed decreased primary state occupancy and increased minor state prevalence during sleep, with reduced direct transition probabilities. In Alzheimer's disease patients, the frontoparietal control network-default mode network co-activation state showed shortened dwell time and occurrence frequency, while the visual network-limbic network co-activation state showed prolonged dwell time and occurrence frequency, with these metrics correlating with cognitive impairment.

Study at a glance

Abstract

The human brain exhibits inherent multistability, with Energy Landscape Analysis (ELA) providing effective frameworks for investigating this property through BOLD signals. However, traditional amplitude-based approaches fundamentally neglect critical phase synchronization dynamics that mediate large-scale neural coordination, while existing phase-based methods like Leading Eigenvector Dynamic Analysis (LEiDA) lack thermodynamic formalism for state stability quantification. Here, we introduce Energy-based Phase-Locking State Analysis (EPLSA), a transformative computational framework that synergistically integrates instantaneous phase-coupling dynamics with rigorous energy landscape principles, addressing fundamental limitations of conventional methodologies. Comprehensive validation across two independent neuroimaging datasets (HCP and Natural Sleep) demonstrated EPLSA's marked superiority over LEiDA and conventional ELA in terms of test-retest reliability, task-specific brain state differentiation, and individual-level classification performance. To demonstrate the physiological and clinical utility of the proposed method, sleep-wake analysis was performed to reveal EPLSA's enhanced sensitivity to consciousness state transitions, identifying decreased primary state occupancy and increased minor state prevalence during sleep, with significantly reduced direct transition probabilities. Furthermore, application to patients with Alzheimer's disease using the OASIS-3 dataset identified shortened dwell time and occurrence frequency for the frontoparietal control network-default mode network (FPCN-DMN) co-activation state, and prolonged dwell time and occurrence frequency for the visual network-limbic network (VIS-LMN) co-activation state, with these metrics significantly correlating with cognitive impairment. By unifying phase-coupling and thermodynamic principles, EPLSA provides novel insights into neurodynamic mechanisms across cognitive tasks, consciousness states, and neurodegenerative conditions, offering a transformative analytical tool for investigating brain function in health and disease with particular promise for early detection and monitoring of neurological disorders.