Simulating Bioelectric Signal Drift in Chronically Implanted Nerve Cuff Electrodes
Stephen Sammut (1,2), Ryan G. L. Koh (1,2), José Zariffa (1,2,3,4)
1. Institute of Biomaterials and Biomedical Engineering, University of Toronto, Toronto, ON Canada
2. KITE, Toronto Rehabilitation Institute – University Health Network, Toronto, ON, Canada
3. Edward S. Rogers Sr. Department of Electrical and Computer Engineering, University of Toronto, Toronto, ON, Canada
4. Rehabilitation Sciences Institute, University of Toronto, Toronto, ON, Canada
Background: Recordings from the peripheral nervous system can be used to extract command signals in a variety of neuroprosthetic and neuromodulation applications. Typically, signal acquisition from peripheral nerves is achieved using intraneural (nerve penetrating) or extraneural (non-nerve penetrating) electrodes. While intraneural electrodes provide better signal to noise ratios, they are more likely to cause damage to the nerve and therefore may not be optimal for long-term use. Extraneural electrodes, such as nerve cuff electrodes, are known to be safe for chronic implantation, however they provide poorer recording selectivity and bioelectric signal drift is often seen in long-term implantation scenarios. As such, new signal processing approaches are needed that can compensate for changes over time and improve the chronic recording selectivity of extraneural electrodes.
Objective: The purpose of this research was to develop a computational model to simulate the changes observed over time in recordings from nerve cuff electrodes. This model will provide a tool to investigate adaptive signal processing strategies.
Methods: By modifying a previously validated computational model of a rat sciatic nerve, a new model was used to simulate two major factors known to cause bioelectric signal drift in nerve cuff electrode recordings: 1) the buildup of encapsulation tissue between the nerve and the cuff, and 2) shifts in the position of the electrode contacts.
Results: An isolated action potential was propagated through each of the models. Preliminary results show simulated encapsulation tissue results in an average normalized root mean square error (NRMSE) of 0.41 (SD=0.17) in recordings seen at the electrode contacts, compared to a base model. Similarly, a 20° shift in electrode position resulted in an average NRMSE of 0.09 (SD=0.06) in recordings seen at the electrode contacts, compared to a base model.
Conclusion: This new model allows for characterization of signal changes expected to occur over time and will enable us to test several approaches for maintaining high levels of performance during chronic implantation in future work. These capabilities will result in substantially more effective assistive technologies for a variety of clinical populations, including amputees and spinal cord injury patients.