Tutorial: A guide to techniques for analysing recordings from the peripheral nervous system

A Scoping Review of Machine Learning Applied to Peripheral Nerve Interfaces

Peripheral nerve interfaces (PNIs) can enable communication with the peripheral nervous system and have a broad range of applications including in bioelectronic medicine and neuroprostheses. They can modulate neural activity through stimulation or monitor conditions by recording from the peripheral nerves. The recent growth of Machine Learning (ML) has led to the application of a wide variety of ML techniques to PNIs, especially in circumstances where the goal is classification or regression. However, the extent to which ML has been applied to PNIs or the range of suitable ML techniques has not been documented. Therefore, a scoping review was conducted to determine and understand the state of ML in the PNI field. The review searched five databases and included 63 studies after full-text review. Most studies incorporated a supervised learning approach to classify activity, with the most common algorithms being some form of neural network (artificial neural network, convolutional neural network or recurrent neural network). Unsupervised, semi-supervised and reinforcement learning (RL) approaches are currently underutilized and could be better leveraged to improve performance in this domain.

Ex-vivo systems for neuromodulation: A comparison of ex-vivo and in-vivo large animal nerve electrophysiology

BACKGROUND

Little research exists on extending ex-vivo systems to large animal nerves, and to the best of our knowledge, there has yet to be a study comparing these against in-vivo data. This paper details the first ex-vivo system for large animal peripheral nerves to be compared with in-vivo results.

NEW METHOD

Detailed ex-vivo and in-vivo closed-loop neuromodulation experiments were conducted on pig ulnar nerves. Temperatures from 20 °C to 37 °C were evaluated for the ex-vivo system. The data were analysed in the time and velocity domains, and a regression analysis established how evoked compound action potential amplitude and modal conduction velocity (CV) varied with temperature and time after explantation.

MAIN RESULTS

Pig ulnar nerves were sustained ex-vivo up to 5 h post-explantation. CV distributions of ex-vivo and in-vivo data were compared, showing closer correspondence at 37 °C. Regression analysis results also demonstrated that modal CV and time since explantation were negatively correlated, whereas modal CV and temperature were positively correlated.

COMPARISON WITH EXISTING METHODS

Previous ex-vivo systems were primarily aimed at small animal nerves, and we are not aware of an ex-vivo system to be directly compared with in-vivo data. This new approach provides a route to understand how ex-vivo systems for large animal nerves can be developed and compared with in-vivo data.

CONCLUSION

The proposed ex-vivo system results were compared with those seen in-vivo, providing new insights into large animal nerve activity post-explantation. Such a system is crucial for complementing in-vivo experiments, maximising collected experimental data, and accelerating neural interface development.

Fork-shaped neural interface with multichannel high spatial selectivity in the peripheral nerve of a rat

Objective.This study aims to develop and validate a sophisticated fork-shaped neural interface (FNI) designed for peripheral nerves, focusing on achieving high spatial resolution, functional selectivity, and improved charge storage capacities. The objective is to create a neurointerface capable of precise neuroanatomical analysis, neural signal recording, and stimulation.Approach.Our approach involves the design and implementation of the FNI, which integrates 32 multichannel working electrodes featuring enhanced charge storage capacities and low impedance. An insertion guide holder is incorporated to refine neuronal selectivity. The study employs meticulous electrode placement, bipolar electrical stimulation, and comprehensive analysis of induced neural responses to verify the FNI's capabilities. Stability over an eight-week period is a crucial aspect, ensuring the reliability and durability of the neural interface.Main results.The FNI demonstrated remarkable efficacy in neuroanatomical analysis, exhibiting accurate positioning of motor nerves and successfully inducing various movements. Stable impedance values were maintained over the eight-week period, affirming the durability of the FNI. Additionally, the neural interface proved effective in recording sensory signals from different hind limb areas. The advanced charge storage capacities and low impedance contribute to the FNI's robust performance, establishing its potential for prolonged use.Significance.This research represents a significant advancement in neural interface technology, offering a versatile tool with broad applications in neuroscience and neuroengineering. The FNI's ability to capture both motor and sensory neural activity positions it as a comprehensive solution for neuroanatomical studies. Moreover, the precise neuromodulation potential of the FNI holds promise for applications in advanced bionic prosthetic control and therapeutic interventions. The study's findings contribute to the evolving field of neuroengineering, paving the way for enhanced understanding and manipulation of peripheral neural functions.

Resource-Efficient Neural Network Architectures for Classifying Nerve Cuff Recordings on Implantable Devices

BACKGROUND

Closed-loop functional electrical stimulation can use recorded nerve signals to create implantable systems that make decisions regarding nerve stimulation in real-time. Previous work demonstrated convolutional neural network (CNN) discrimination of activity from different neural pathways recorded by a high-density multi-contact nerve cuff electrode, achieving state-of-the-art performance but requiring too much data storage and power for a practical implementation on surgically implanted hardware.

OBJECTIVE

To reduce resource utilization for an implantable implementation, with minimal performance loss for CNNs that can discriminate between neural pathways in multi-contact cuff electrode recordings.

METHODS

Neural networks (NNs) were evaluated using rat sciatic nerve recordings previously collected using 56-channel cuff electrodes to capture spatiotemporal neural activity patterns. NNs were trained to classify individual, natural compound action potentials (nCAPs) elicited by sensory stimuli. Three architectures were explored: the previously reported ESCAPE-NET, a fully convolutional network, and a recurrent neural network. Variations of each architecture were evaluated based on F1-score, number of weights, and floating-point operations (FLOPs).

RESULTS

NNs were identified that, when compared to ESCAPE-NET, require 1,132-1,787x fewer weights, 389-995x less memory, and 6-11,073x fewer FLOPs, while maintaining macro F1-scores of 0.70-0.71 compared to a baseline of 0.75. Memory requirements range from 22.69 KB to 58.11 KB, falling within on-chip memory sizes from published deep learning accelerators fabricated in ASIC technology.

CONCLUSION

Reduced versions of ESCAPE-NET require significantly fewer resources without significant accuracy loss, thus can be more easily incorporated into a surgically implantable device that performs closed-loop responsive neural stimulation.

Microneurography as a minimally invasive method to assess target engagement during neuromodulation

Objective.Peripheral neural signals recorded during neuromodulation therapies provide insights into local neural target engagement and serve as a sensitive biomarker of physiological effect. Although these applications make peripheral recordings important for furthering neuromodulation therapies, the invasive nature of conventional nerve cuffs and longitudinal intrafascicular electrodes (LIFEs) limit their clinical utility. Furthermore, cuff electrodes typically record clear asynchronous neural activity in small animal models but not in large animal models. Microneurography, a minimally invasive technique, is already used routinely in humans to record asynchronous neural activity in the periphery. However, the relative performance of microneurography microelectrodes compared to cuff and LIFE electrodes in measuring neural signals relevant to neuromodulation therapies is not well understood.Approach.To address this gap, we recorded cervical vagus nerve electrically evoked compound action potentials (ECAPs) and spontaneous activity in a human-scaled large animal model-the pig. Additionally, we recorded sensory evoked activity and both invasively and non-invasively evoked CAPs from the great auricular nerve. In aggregate, this study assesses the potential of microneurography electrodes to measure neural activity during neuromodulation therapies with statistically powered and pre-registered outcomes (https://osf.io/y9k6j).Main results.The cuff recorded the largest ECAP signal (p< 0.01) and had the lowest noise floor amongst the evaluated electrodes. Despite the lower signal to noise ratio, microneurography electrodes were able to detect the threshold for neural activation with similar sensitivity to cuff and LIFE electrodes once a dose-response curve was constructed. Furthermore, the microneurography electrodes recorded distinct sensory evoked neural activity.Significance.The results show that microneurography electrodes can measure neural signals relevant to neuromodulation therapies. Microneurography could further neuromodulation therapies by providing a real-time biomarker to guide electrode placement and stimulation parameter selection to optimize local neural fiber engagement and study mechanisms of action.