A framework for the discrimination of neural pathways using multi-contact nerve cuff electrodes

Monitoring the activity of specific neural pathways in a peripheral nerve is a task with numerous applications in implanted neuroprosthetic systems. Achieving selective recording using multi-contact nerve cuff electrodes is appealing because these devices are well suited for chronic use, but no viable general solution to the task of discriminating combinations of active pathways from extra-neural recordings has yet been proposed. Bioelectric source localization approaches have been suggested, but their effectiveness is limited by the accuracy of the nerve model used to solve the forward problem. We propose a model-free alternative to the pathway discrimination task, in which experimental data is used to estimate a solution to the forward problem. The method was evaluated using a 56-channel cuff placed on the rat sciatic nerve. 3 pathways were discriminated with a 94.2% success rate when individually active, whereas further improvements are needed in order to recover combinations of simultaneously active pathways.