Individualized AI-driven neuromodulation enhances tongue motor and sensory control: preliminary efficacy targeted towards the alleviation of chronic cranial neuropathies
Papageorgiou, T. D.; Webb, J.; Allam, A.; Froudarakis, E.; Rohren, E. M.; Sturgis, E. M.; Hutcheson, K. A.; Heilbronner, S. R.
Show abstract
Precise modulation of brain networks responsible for tongue motor and sensory control (TMSC) is critical for restoring functions, such as speech and swallowing in neurodegenerative disease or in treatment-induced chronic cranial neuropathy. We present an individualized, AI-driven fMRI neuromodulation (iNM) platform that adaptively targets subject-specific TMSC networks in real time. To enhance iNM precision and encodability --critical for neurorehabilitation--we mapped each healthy participants individualized TMSC selectivity network, creating a subject-specific TMSC digital twin. iNM increased signal strength, spatial expansion, and consistency across motor, sensory, and attention regions, while it reduced signal variability. The bilateral inferior parietal lobule emerged as key sensorimotor integration hub, as it exhibited exclusive activation under iNM along with highest discriminability, and largest spatial expansion. iNM also significantly strengthened and expanded motor, sensory, and attention-related networks -- medial-middle frontal areas, insula-claustrum, S1, M1, basal ganglia, motor cerebellum, and inferior temporal-- supporting interoceptive and proprioceptive-motor integration. Machine learning and unsupervised hidden Markov modeling revealed that iNM enhanced the decodability and stability of TMSC-neural states, while it suppressed competing swallow-neural state interference. Notably, the iNM effects extended beyond the neuromodulation window, indicating functional persistence--a key requirement for rehabilitation. iNM reconfigured TMSC networks by strengthening cortico-subcortical connectivity and adaptive circuit dynamics. Our findings show iNM as a non-invasive, personalized intervention capable of selectively enhancing sensorimotor control with high spatiotemporal specificity. By demonstrating mechanistic network-precision and functional carryover, iNM offers a promising intervention for individuals with limited treatment options, including head and neck cancer survivors and early-stage neurodegenerative disease patients.
Matching journals
The top 5 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Measuring the dynamic balance of integration and segregation underlying consciousness, anesthesia, and sleep 95%
- Myelination and excitation-inhibition balance synergistically shape structure-function coupling across the human cortex 95%
- Multiple Brain Activation Patterns for the Same Perceptual Decision-Making Task 95%
Similar papers in this journal
- Reconfigurations of cortical manifold structure during reward-based motor learning 95%
- Large-scale neural dynamics in a shared low-dimensional state space reflect cognitive and attentional dynamics 94%
- THINGS-data: A multimodal collection of large-scale datasets for investigating object representations in human brain and behavior 94%
Similar papers in this journal
- Simultaneous fMRI and fast-scan cyclic voltammetry bridges oxygenation and neurotransmitter dynamics across spatiotemporal scales 94%
- Shared subcortical arousal systems across sensory modalities during transient modulation of attention 94%
- Representation Learning of Resting State fMRI with Variational Autoencoder 94%
Similar papers in this journal
Similar papers in this journal
- Endogenous variation in ventromedial prefrontal cortex state dynamics during naturalistic viewing reflects affective experience 95%
- Cortical-subcortical structural connections support transcranial magnetic stimulation engagement of the amygdala 94%
- Chemogenetic Stimulation of Tonic Locus Coeruleus Activity Strengthens the Default Mode Network 94%
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.