175: Dynamic Brain Network Correlates of Prosthesis Embodiment in Patients with Phantom Limb Pain

Topic: Clinical Science
Subtopic: Itch, Physiotherapy (e.g. exercise, rehab), Physiotherapy, Electrophysiology/Clinical neurophysiology (e.g. microneurography, EEG, MEG, OPMs, TMS)

Background and Aims: Phantom limb pain (PLP) is linked to disturbances of the embodied self, particularly the sense of agency and ownership. Although agency-based approaches such as mirror therapy and virtual reality can reduce PLP, their effects are often limited. Evidence suggests that stronger prosthesis ownership is associated with lower PLP severity. We hypothesized that synchronized tactile feedback delivered to referred sensation areas would enhance prosthesis embodiment and alleviate PLP. We further aimed to identify underlying neural network mechanisms using EEG-based hidden Markov modeling of dynamic brain states.

Methods: Nine individuals with upper-limb amputation participated in this study. Participants underwent mapping of referred sensation areas around the stump. A prosthetic hand equipped with touch sensors delivered temporally synchronized vibrotactile stimulation during five minutes of gentle prosthetic stroking. 64 channel EEG was recorded continuously during this period. Preprocessing included filtering, artifact removal, source modeling, and parcellation. Brain state dynamics were inferred using a pre-trained hidden Markov model derived from a MEG dataset. For each inferred state, fractional occupancy was calculated and transition probabilities between states were estimated. Participants rated embodiment using an ownership questionnaire. Correlations with network metrics were examined.

Results: Higher embodiment scores were associated with increased fractional occupancy of a frontal network state (p<0.05). Stronger embodiment was also correlated with greater transition probability from sensory networks (visual and sensorimotor states) to the frontal network (p<0.05). These findings indicate that prosthesis embodiment relates not only to frontal network predominance but also to bottom-up information flow from sensory processing regions to higher-order cognitive systems. Five of nine participants showed partial reductions in PLP following the intervention.

Conclusions: Prosthesis embodiment was associated with greater frontal network occupancy and increased transitions from sensory to frontal networks. These findings suggest that dynamic network interactions may contribute to embodiment processes. Enhancing ownership through referred sensation–based tactile feedback may represent a promising therapeutic approach for PLP. Monitoring dynamic brain states could help identify neural markers of successful embodiment and guide future neuromodulation strategies targeting large-scale sensory–frontal network interactions. Ownership-based interventions may provide a more comprehensive rehabilitation framework for phantom limb pain.

References: 1) van Es MWJ, Higgins C, Gohil C, Quinn AJ, Vidaurre D, Woolrich MW. Large-scale cortical functional networks are organized in structured cycles. Nat Neurosci. 2025 Oct;28(10):2118-2128.
2) Chetan Gohil, Rukuang Huang, Cameron Higgins, Mats W.J. van Es, Andrew J. Quinn, Diego Vidaurre, Mark W. Woolrich. Canonical Hidden Markov Model Networks for Studying M/EEG. bioRxiv. 2025.

Presenting Author: Michihiro Osumi
Submitting Author: Michihiro Osumi

Senior Author: Michihiro Osumi

IASP 2026 Global Year theme of Neuropathic Pain: No

• Michihiro Osumi - Graduate School of Health Science, Kio University
• Masahiro Yoshikawa - Graduate School of Robotics and Design, Osaka Institute of Technology
• Noritaka Kawashima - Research Institute, National Rehabilitation Center for Persons with Disabilities