Closed-Loop Neuromodulation for Brain Fatigue: From Real-Time Biomarkers to Adaptive Intervention

Brain fatigue is a debilitating condition whose molecular complexity-neurotransmitter imbalance, neuroinflammation, and metabolic failure-has long defied effective intervention. Conventional open-loop neuromodulation offers fixed stimulation, but fatigue fluctuates. Closed-loop neuromodulation, adjusting stimulation in real time, holds real promise. Yet its delivery hinges on a single, unresolved bottleneck: the sensing-decision chain. In this review, we deconstruct closed-loop systems into sensing, decision, and intervention, and argue that the true challenge is not technological but informational-how to integrate fast electrophysiological signals with slow molecular biomarkers (inflammatory cytokines, neurotrophic factors, adenosine) into a unified control framework. We show that a "fast-slow variable" architecture offers a practical path forward: fast signals guide immediate responses, slow variables set baselines and thresholds, and their integration enables predictive, pre-emptive intervention. We also examine the molecular correlates of neuromodulation-synaptic plasticity, anti-inflammatory signaling, and neurotrophic regulation-as the mechanistic foundation for therapeutic effect. Finally, we confront the translational triad of causality, inter-individual variability, and the information-energy-time trade-off. We conclude that realizing effective closed-loop neuromodulation for brain fatigue will require parallel advances in both stimulation hardware-improving spatial targeting, dose precision, and modality versatility-and individualized sensing-decision algorithms capable of reliably translating multi-modal biomarkers into timely, safe interventions. Both areas remain active fronts of development, and neither can be neglected in clinical translation.