Multiunit Activity-Based Real-Time Limb-State Estimation from Dorsal Root Ganglion Recordings

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Proprioceptive afferent signals are crucial for sensory feedback in closed-loop functional electrical stimulation (FES).
• Previous methods often rely on single-unit neuronal activity, limiting information extraction.
• A need exists for more robust methods to decode proprioceptive information for FES applications.

Purpose of the Study:

• To develop and validate a novel decoding method for estimating ankle and knee joint angles using multiunit activity (MUA).
• To assess the efficacy of the proposed method in providing real-time sensory feedback for closed-loop FES.

Main Methods:

• Proprioceptive afferent signals were recorded from dorsal root ganglia using microelectrodes during passive joint movements.
• Multiunit activity (MUA) was processed to extract the mean absolute value (MAV) feature.
• A dynamically driven recurrent neural network (DDRNN) was employed to decode joint angles from MAV features.

Main Results:

• The MAV feature derived from MUA effectively captured limb state information.
• The DDRNN demonstrated superior decoding performance compared to traditional linear estimators.
• The proposed method achieved processing time delays compatible with real-time FES control.

Conclusions:

• The developed MUA-based decoding method is a viable approach for real-time sensory feedback in closed-loop FES.
• This technique enhances the potential for more intuitive and effective FES control.
• The findings support the application of this method in improving FES system functionality and user experience.