Modeling Hierarchical Brain Dynamics Outperforms Hormonal Biomarkers in Predicting Menstrual Cycle Phases

Abstract

Hormonal fluctuations across the menstrual cycle influence large-scale brain dynamics, yet the underlying neurobiological mechanisms remain poorly understood. In this study, 60 nat-urally cycling women were scanned using resting-state fMRI during the early follicular, pre-ovulatory, and mid-luteal phases. We then applied a thermodynamics-inspired framework to explore the functional hierarchical organization of whole-brain dynamics across these phases. First, we found that brain dynamics are significantly modulated by estradiol, progesterone, and age across multiple resting-state networks. Second, to elucidate underlying mechanisms, we es-timated generative effective connectivity (GEC) matrices using whole-brain models and trained support vector machine classifiers to predict menstrual phases. These model-based biomarkers outperformed traditional functional connectivity and hormone measures in classifying men-strual cycle phases. These findings reveal that menstrual cycle-related changes modulate the hierarchical reorganization of brain dynamics, highlighting the potential of model-based ap-proaches to advance womens brain health research.