Influence of nerve cuff channel count and implantation site on the separability of afferent ENG

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.

Spatio-temporal feature extraction in sensory electroneurographic signals

The recording and analysis of peripheral neural signal can provide insight for various prosthetic and bioelectronics medicine applications. However, there are few studies that investigate how informative features can be extracted from population activity electroneurographic (ENG) signals. In this study, five feature extraction frameworks were implemented on sensory ENG datasets and their classification performance was compared. The datasets were collected in acute rat experiments where multi-channel nerve cuffs recorded from the sciatic nerve in response to proprioceptive stimulation of the hindlimb. A novel feature extraction framework, which incorporates spatio-temporal focus and dynamic time warping, achieved classification accuracies above 90% while keeping a low computational cost. This framework outperformed the remaining frameworks tested in this study and has improved the discrimination accuracy of the sensory signals. Thus, this study has extended the tools available to extract features from sensory population activity ENG signals. This article is part of the theme issue 'Advanced neurotechnologies: translating innovation for health and well-being'.

A comparison of extraneural approaches for selective recording in the peripheral nervous system

The peripheral nervous system is a key target for the development of neural interfaces. However, recording from the peripheral nerves can be challenging especially when chronic implantation is desired. Nerve cuffs are frequently employed using either two or three contacts to provide a single recording channel. Advancements in manufacturing technology have enabled multi-contact cuffs, enabling measurement of both temporal (i.e., velocity) and spatial information (i.e., spatial location). Selective techniques have been developed with different time resolutions but it is unclear how the number of contacts and their spatial configuration affect their performance. Thus, this paper investigates two extraneural recording techniques (LDA and spatiotemporal signatures) and compares them using recordings made from the sciatic nerve of rats using high density (HD, 56 contact), reduced-HD (16 contacts), and low density (LD, 16 contact) datasets. Performance of the two techniques was evaluated using classification accuracy and F₁-score. Both techniques show an expected improvement in classification accuracy with the spatiotemporal signature approach showing a 21.6 (LD to HD) - 24.6% (reduced HD to HD) increase and the LDA approach showing a 2.9 (reduced HD to HD) - 41.3% (LD to HD) increase and had comparable results in both the LD and HD datasets.

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

The nervous system, through a combination of conscious and automatic processes, enables the regulation of the body and its interactions with the environment. The peripheral nervous system is an excellent target for technologies that seek to modulate, restore or enhance these abilities as it carries sensory and motor information that most directly relates to a target organ or function. However, many applications require a combination of both an effective peripheral nerve interface and effective signal processing techniques to provide selective and stable recordings. While there are many reviews on the design of peripheral nerve interfaces, reviews of data analysis techniques and translational considerations are limited. Thus, this tutorial aims to support new and existing researchers in the understanding of the general guiding principles, and introduces a taxonomy for electrode configurations, techniques and translational models to consider.

The Design of a Low Noise, Multi-Channel Recording System for Use in Implanted Peripheral Nerve Interfaces

In the development of implantable neural interfaces, the recording of signals from the peripheral nerves is a major challenge. Since the interference from outside the body, other biopotentials, and even random noise can be orders of magnitude larger than the neural signals, a filter network to attenuate the noise and interference is necessary. However, these networks may drastically affect the system performance, especially in recording systems with multiple electrode cuffs (MECs), where a higher number of electrodes leads to complicated circuits. This paper introduces formal analyses of the performance of two commonly used filter networks. To achieve a manageable set of design equations, the state equations of the complete system are simplified. The derived equations help the designer in the task of creating an interface network for specific applications. The noise, crosstalk and common-mode rejection ratio (CMRR) of the recording system are computed as a function of electrode impedance, filter component values and amplifier specifications. The effect of electrode mismatches as an inherent part of any multi-electrode system is also discussed, using measured data taken from a MEC implanted in a sheep. The accuracy of these analyses is then verified by simulations of the complete system. The results indicate good agreement between analytic equations and simulations. This work highlights the critical importance of understanding the effect of interface circuits on the performance of neural recording systems.

Compensation Strategies for Bioelectric Signal Changes in Chronic Selective Nerve Cuff Recordings: A Simulation Study

Peripheral nerve interfaces (PNIs) allow us to extract motor, sensory, and autonomic information from the nervous system and use it as control signals in neuroprosthetic and neuromodulation applications. Recent efforts have aimed to improve the recording selectivity of PNIs, including by using spatiotemporal patterns from multi-contact nerve cuff electrodes as input to a convolutional neural network (CNN). Before such a methodology can be translated to humans, its performance in chronic implantation scenarios must be evaluated. In this simulation study, approaches were evaluated for maintaining selective recording performance in the presence of two chronic implantation challenges: the growth of encapsulation tissue and rotation of the nerve cuff electrode. Performance over time was examined in three conditions: training the CNN at baseline only, supervised re-training with explicitly labeled data at periodic intervals, and a semi-supervised self-learning approach. This study demonstrated that a selective recording algorithm trained at baseline will likely fail over time due to changes in signal characteristics resulting from the chronic challenges. Results further showed that periodically recalibrating the selective recording algorithm could maintain its performance over time, and that a self-learning approach has the potential to reduce the frequency of recalibration.

Systematic analysis of wavelet denoising methods for neural signal processing

Objective.Among the different approaches for denoising neural signals, wavelet-based methods are widely used due to their ability to reduce in-band noise. All wavelet denoising algorithms have a common structure, but their effectiveness strongly depends on several implementation choices, including the mother wavelet, the decomposition level, the threshold definition, and the way it is applied (i.e. the thresholding). In this work, we investigated these factors to quantitatively assess their effects on neural signals in terms of noise reduction and morphology preservation, which are important when spike sorting is required downstream.Approach.Based on the spectral characteristics of the neural signal, according to the sampling rate of the signals, we considered two possible decomposition levels and identified the best-performing mother wavelet. Then, we compared different threshold estimation and thresholding methods and, for the best ones, we also evaluated their effect on clearing the approximation coefficients. The assessments were performed on synthetic signals that had been corrupted by different types of noise and on a murine peripheral nervous system dataset, both of which were sampled at about 16 kHz. The results were statistically analysed in terms of their Pearson's correlation coefficients, root-mean-square errors, and signal-to-noise ratios.Main results.As expected, the wavelet implementation choices greatly influenced the processing performance. Overall, the Haar wavelet with a five-level decomposition, hard thresholding method, and the threshold proposed by Hanet al(2007) achieved the best outcomes. Based on the adopted performance metrics, wavelet denoising with these parametrizations outperformed conventional 300-3000 Hz linear bandpass filtering.Significance.These results can be used to guide the reasoned and accurate selection of wavelet denoising implementation choices in the context of neural signal processing, particularly when spike-morphology preservation is required.

SenseBack - An Implantable System for Bidirectional Neural Interfacing

Chronic in-vivo neurophysiology experiments require highly miniaturized, remotely powered multi-channel neural interfaces which are currently lacking in power or flexibility post implantation. To resolve this problem we present the SenseBack system, a post-implantation reprogrammable wireless 32-channel bidirectional neural interfacing device that can enable chronic peripheral electrophysiology experiments in freely behaving small animals. The large number of channels for a peripheral neural interface, coupled with fully implantable hardware and complete software flexibility enable complex in-vivo studies where the system can adapt to evolving study needs as they arise. In complementary \textit{ex-vivo} and \textit{in-vivo} preparations, we demonstrate that this system can record neural signals and perform high-voltage, bipolar stimulation on any channel. In addition, we demonstrate transcutaneous power delivery and Bluetooth 5 data communication with a PC. The SenseBack system is capable of stimulation on any channel with 20 V of compliance and up to 315 A of current, and highly configurable recording with per-channel adjustable gain and filtering with 8 sets of 10-bit ADCs to sample data at 20 kHz for each channel. To our knowledge this is the first such implantable research platform offering this level of performance and flexibility post-implantation (including complete reprogramming even after encapsulation) for small animal electrophysiology. Here we present initial acute trials, demonstrations and progress towards a system that we expect to enable a wide range of electrophysiology experiments in freely behaving animals.

W:Ti Flexible Transversal Electrode Array for Peripheral Nerve Stimulation: A Feasibility Study

The development of hardware for neural interfacing remains a technical challenge. We introduce a flexible, transversal intraneural tungsten:titanium electrode array for acute studies. We characterize the electrochemical properties of this new combination of tungsten and titanium using cyclic voltammetry and electrochemical impedance spectroscopy. With an in-vivo rodent study, we show that the stimulation of peripheral nerves with this electrode array is possible and that more than half of the electrode contacts can yield a stimulation selectivity index of 0.75 or higher at low stimulation currents. This feasibility study paves the way for the development of future cost-effective and easy-to-fabricate neural interfacing electrodes for acute settings, which ultimately can inform the development of technologies that enable bi-directional communication with the human nervous system.

Automatic three-dimensional reconstruction of fascicles in peripheral nerves from histological images

Computational studies can be used to support the development of peripheral nerve interfaces, but currently use simplified models of nerve anatomy, which may impact the applicability of simulation results. To better quantify and model neural anatomy across the population, we have developed an algorithm to automatically reconstruct accurate peripheral nerve models from histological cross-sections. We acquired serial median nerve cross-sections from human cadaveric samples, staining one set with hematoxylin and eosin (H&E) and the other using immunohistochemistry (IHC) with anti-neurofilament antibody. We developed a four-step processing pipeline involving registration, fascicle detection, segmentation, and reconstruction. We compared the output of each step to manual ground truths, and additionally compared the final models to commonly used extrusions, via intersection-over-union (IOU). Fascicle detection and segmentation required the use of a neural network and active contours in H&E-stained images, but only simple image processing methods for IHC-stained images. Reconstruction achieved an IOU of 0.42±0.07 for H&E and 0.37±0.16 for IHC images, with errors partially attributable to global misalignment at the registration step, rather than poor reconstruction. This work provides a quantitative baseline for fully automatic construction of peripheral nerve models. Our models provided fascicular shape and branching information that would be lost via extrusion.

Neural signal recording and processing in somatic neuroprosthetic applications. A review

Neurointerfaces have acquired major relevance as both rehabilitative and therapeutic tools for patients with spinal cord injury, limb amputations and other neural disorders. Bidirectional neural interfaces are a key component for the functional control of neuroprosthetic devices. The two main neuroprosthetic applications of interfaces with the peripheral nervous system (PNS) are: the refined control of artificial prostheses with sensory neural feedback, and functional electrical stimulation (FES) systems attempting to generate motor or visceral responses in paralyzed organs. The results obtained in experimental and clinical studies with both, extraneural and intraneural electrodes are very promising in terms of the achieved functionality for the neural stimulation mode. However, the results of neural recordings with peripheral nerve interfaces are more limited. In this paper we review the different existing approaches for PNS signals recording, denoising, processing and classification, enabling their use for bidirectional interfaces. PNS recordings can provide three types of signals: i) population activity signals recorded by using extraneural electrodes placed on the outer surface of the nerve, which carry information about cumulative nerve activity; ii) spike activity signals recorded with intraneural electrodes placed inside the nerve, which carry information about the electrical activity of a set of individual nerve fibers; and iii) hybrid signals, which contain both spiking and cumulative signals. Finally, we also point out some of the main limitations, which are hampering clinical translation of neural decoding, and indicate possible solutions for improvement.

A review for the peripheral nerve interface designer

Informational density and relative accessibility of the peripheral nervous system make it an attractive site for therapeutic intervention. Electrode-based electrophysiological interfaces with peripheral nerves have been under development since the 1960s and, for several applications, have seen widespread clinical implementation. However, many applications require a combination of neural target resolution and stability which has thus far eluded existing peripheral nerve interfaces (PNIs). With the goal of aiding PNI designers in development of devices that meet the demands of next-generation applications, this review seeks to collect and present practical considerations and best practices which emerge from the literature, including both lessons learned during early PNI development and recent ideas. Fundamental and practical principles guiding PNI design are reviewed, followed by an updated and critical account of existing PNI designs and strategies. Finally, a brief survey of in vitro and in vivo PNI characterization methods is presented.

Evaluation of Time-Domain Features of Sensory ENG Signals

In the development of closed-loop prostheses that record from the patient's own nerves to provide sensory feedback, it is first necessary to determine the features of sensory signals that may help to identify different sensations. The aim of this work was to investigate different time-domain features for separation of sensory electroneurographic signals. To do this, sensory signals were elicited in response to mechanical stimulation of the rat hindpaw and these signals were recorded from a cuff electrode array placed on the sciatic nerve. Thirteen features were extracted, including: mean absolute value, variance, waveform length and ten time-domain descriptors that were recently proposed for classification of electromyographic signals. These features were individually fed into a linear discriminant analysis classifier. The results showed that the best overall performing features were the mean absolute value and waveform length. Additionally, six of the ten time-domain descriptors showed a comparable performance to these two features. This indicates that these features could be used as a tool to aid our understanding of the sensory neural signals recorded and further improve classification results. Enhanced classification of electroneurographic signals will provide the opportunity to develop more efficacious sensory-motor prostheses in the future.