Individualized AI-driven Neuromodulation Selectively Enhances and Reorganizes Human Tongue Sensorimotor Networks.

Abstract

Precise modulation of brain networks for tongue motor and sensory control (TMSC) is critical for neruorehabilitation. Existing neuromodulation approaches use fixed targets, ignoring individual network architecture. We present an AI-driven, individualized fMRI neuromodulation (iNM) platform adaptively targeting subject-specific TMSC networks in real-time. Using a neurofunctional twin built from each participant's anatomy and function, iNM reinforces targeted activity and suppresses competing states. Unsupervised machine learning and hidden Markov modeling showed iNM increased TMSC-state occupancy from 82% to 97% and suppressed swallow-state interference by 83%. The inferior parietal lobule emerged as a sensorimotor integration hub, with increased magnitude (d' 90.4 right, 84 left) and spatial expansion (134.8% right, 81.8% left). Cortico-subcortical connectivity strengthened 53% to 74% across motor, sensory, and attention networks and persisted into non-iNM blocks. Thus, iNM may offer precision neurorehabilitation for CNS or PNS disorders, such as amyotrophic lateral sclerosis, and treatment-induced cranial neuropathy in head-and-neck-cancer survivors, respectively.