Neuro-fuzzy extraction of angular information from muscle afferents for ankle control during standing in paraplegic subjects: an animal model

Decoding Bilateral Hindlimb Kinematics From Cat Spinal Signals Using Three-Dimensional Convolutional Neural Network

To date, decoding limb kinematic information mostly relies on neural signals recorded from the peripheral nerve, dorsal root ganglia (DRG), ventral roots, spinal cord gray matter, and the sensorimotor cortex. In the current study, we demonstrated that the neural signals recorded from the lateral and dorsal columns within the spinal cord have the potential to decode hindlimb kinematics during locomotion. Experiments were conducted using intact cats. The cats were trained to walk on a moving belt in a hindlimb-only condition, while their forelimbs were kept on the front body of the treadmill. The bilateral hindlimb joint angles were decoded using local field potential signals recorded using a microelectrode array implanted in the dorsal and lateral columns of both the left and right sides of the cat spinal cord. The results show that contralateral hindlimb kinematics can be decoded as accurately as ipsilateral kinematics. Interestingly, hindlimb kinematics of both legs can be accurately decoded from the lateral columns within one side of the spinal cord during hindlimb-only locomotion. The results indicated that there was no significant difference between the decoding performances obtained using neural signals recorded from the dorsal and lateral columns. The results of the time-frequency analysis show that event-related synchronization (ERS) and event-related desynchronization (ERD) patterns in all frequency bands could reveal the dynamics of the neural signals during movement. The onset and offset of the movement can be clearly identified by the ERD/ERS patterns. The results of the mutual information (MI) analysis showed that the theta frequency band contained significantly more limb kinematics information than the other frequency bands. Moreover, the theta power increased with a higher locomotion speed.

Foreign Body Reaction to Implanted Biomaterials and Its Impact in Nerve Neuroprosthetics

The implantation of any foreign material into the body leads to the development of an inflammatory and fibrotic process-the foreign body reaction (FBR). Upon implantation into a tissue, cells of the immune system become attracted to the foreign material and attempt to degrade it. If this degradation fails, fibroblasts envelop the material and form a physical barrier to isolate it from the rest of the body. Long-term implantation of medical devices faces a great challenge presented by FBR, as the cellular response disrupts the interface between implant and its target tissue. This is particularly true for nerve neuroprosthetic implants-devices implanted into nerves to address conditions such as sensory loss, muscle paralysis, chronic pain, and epilepsy. Nerve neuroprosthetics rely on tight interfacing between nerve tissue and electrodes to detect the tiny electrical signals carried by axons, and/or electrically stimulate small subsets of axons within a nerve. Moreover, as advances in microfabrication drive the field to increasingly miniaturized nerve implants, the need for a stable, intimate implant-tissue interface is likely to quickly become a limiting factor for the development of new neuroprosthetic implant technologies. Here, we provide an overview of the material-cell interactions leading to the development of FBR. We review current nerve neuroprosthetic technologies (cuff, penetrating, and regenerative interfaces) and how long-term function of these is limited by FBR. Finally, we discuss how material properties (such as stiffness and size), pharmacological therapies, or use of biodegradable materials may be exploited to minimize FBR to nerve neuroprosthetic implants and improve their long-term stability.

Decoding hindlimb kinematics from descending and ascending neural signals during cat locomotion

OBJECTIVE

The main objective of this research is to record both sensory and motor information from the ascending and descending tracts within the spinal cord for decoding the hindlimb kinematics during walking on the treadmill.

APPROACH

Two different experimental paradigms (i.e., active and passive) were used in the current study. During active experiments, five cats were trained to walk bipedally while their hands kept on the front frame of the treadmill for balance or to walk quadrupedally. During passive experiments, the limb was passively moved by the experimenter. Local field potential (LFP) activity was recorded using a microwire array implanted in the dorsal column (DC) and lateral column (LC) of the L3-L4 spinal segments. The amplitude and frequency components of the LFP formed the feature set and the elastic net regularization was used to decode the hindlimb joint angles.

MAIN RESULTS

The results show that there is no significant difference between the information content of the signals recorded from the DC and LC regions during walking on the treadmill, but the information content of the DC is significantly higher than that of the LC during passively applied movement of the hindlimb in the anesthetized cats. Moreover, the decoding performance obtained using the recorded signals from the DC is comparable with that from the LC during locomotion. But, the decoding performance obtained using the recording channels in the DC is significantly better than that obtained using the signals recorded from the LC. The long-term analysis shows that robust decoding performance can be achieved over 2-3 months without a significant decrease in performance.

SIGNIFICANCE

This work presents a promising approach to developing a natural and robust motor neuroprosthesis device using descending neural signals to execute the movement and ascending neural signals as the feedback information for control of the movement.

A probabilistic recurrent neural network for decoding hind limb kinematics from multi-segment recordings of the dorsal horn neurons

OBJECTIVE

Providing accurate and robust estimates of limb kinematics from recorded neural activities is prominent in closed-loop control of functional electrical stimulation (FES). A major issue in providing accurate decoding the limb kinematics is the decoding model. The primary goal of this study is to develop a decoding approach to model the dynamic interactions of neural systems for accurate decoding. Another critical issue is to find reliable recording sites. Up to now, neural recordings from spinal neural activities were investigated. In this paper, the neural recordings from different vertebrae in decoding limb kinematics are investigated.

APPROACH

In the current study, a new generative probabilistic model with explicit considering the joint density is developed. Then, an adaptive discriminative learning algorithm is proposed for learning the model. It will be shown that the proposed generative process can be implemented by a recurrent neural network (RNN) with specific structure. We record the neural activities from dorsal horn neurons by using three electrodes placed in the L4, L5, and L6 vertebrae in anesthetized cats.

MAIN RESULTS

Information theoretic analysis on single-joint movement and multi-segment recordings implies the rostrocaudal distribution of kinematic information. It is demonstrated that during hip movement, best decoding performance is achieved by L4 recordings. For knee and ankle movements, best decodings are achieved by L5, and L6 recordings respectively. It is also shown that the decoding accuracy using multi-segment recordings outperforms decoding accuracy obtained by single-segment recording in multi-joint movement. The results also confirm the superiority of the proposed probabilistic recurrent neural network (PRNN) over the conventional recurrent neural network and Kalman filter ([Formula: see text]).

SIGNIFICANCE

Multi-segment recordings from dorsal horn neurons as well as the proposed probabilistic recurrent network model provide a promising approach for robust and accurate decoding limb kinematics.

An Implantable Wireless Neural Interface System for Simultaneous Recording and Stimulation of Peripheral Nerve with a Single Cuff Electrode

Recently, implantable devices have become widely used in neural prostheses because they eliminate endemic drawbacks of conventional percutaneous neural interface systems. However, there are still several issues to be considered: low-efficiency wireless power transmission; wireless data communication over restricted operating distance with high power consumption; and limited functionality, working either as a neural signal recorder or as a stimulator. To overcome these issues, we suggest a novel implantable wireless neural interface system for simultaneous neural signal recording and stimulation using a single cuff electrode. By using widely available commercial off-the-shelf (COTS) components, an easily reconfigurable implantable wireless neural interface system was implemented into one compact module. The implantable device includes a wireless power consortium (WPC)-compliant power transmission circuit, a medical implant communication service (MICS)-band-based radio link and a cuff-electrode path controller for simultaneous neural signal recording and stimulation. During in vivo experiments with rabbit models, the implantable device successfully recorded and stimulated the tibial and peroneal nerves while communicating with the external device. The proposed system can be modified for various implantable medical devices, especially such as closed-loop control based implantable neural prostheses requiring neural signal recording and stimulation at the same time.

Neuroprosthetic technology for individuals with spinal cord injury

CONTEXT

Spinal cord injury (SCI) results in a loss of function and sensation below the level of the lesion. Neuroprosthetic technology has been developed to help restore motor and autonomic functions as well as to provide sensory feedback.

FINDINGS

This paper provides an overview of neuroprosthetic technology that aims to address the priorities for functional restoration as defined by individuals with SCI. We describe neuroprostheses that are in various stages of preclinical development, clinical testing, and commercialization including functional electrical stimulators, epidural and intraspinal microstimulation, bladder neuroprosthesis, and cortical stimulation for restoring sensation. We also discuss neural recording technologies that may provide command or feedback signals for neuroprosthetic devices.

CONCLUSION/CLINICAL RELEVANCE

Neuroprostheses have begun to address the priorities of individuals with SCI, although there remains room for improvement. In addition to continued technological improvements, closing the loop between the technology and the user may help provide intuitive device control with high levels of performance.

Gait phase detection from sciatic nerve recordings in functional electrical stimulation systems for foot drop correction

Cutaneous afferent activities recorded by a nerve cuff electrode have been used to detect the stance phase in a functional electrical stimulation system for foot drop correction. However, the implantation procedure was difficult, as the cuff electrode had to be located on the distal branches of a multi-fascicular nerve to exclude muscle afferent and efferent activities. This paper proposes a new gait phase detection scheme that can be applied to a proximal nerve root that includes cutaneous afferent fibers as well as muscle afferent and efferent fibers. To test the feasibility of this scheme, electroneurogram (ENG) signals were measured from the rat sciatic nerve during treadmill walking at several speeds, and the signal properties of the sciatic nerve were analyzed for a comparison with kinematic data from the ankle joint. On the basis of these experiments, a wavelet packet transform was tested to define a feature vector from the sciatic ENG signals according to the gait phases. We also propose a Gaussian mixture model (GMM) classifier and investigate whether it could be used successfully to discriminate feature vectors into the stance and swing phases. In spite of no significant differences in the rectified bin-integrated values between the stance and swing phases, the sciatic ENG signals could be reliably classified using the proposed wavelet packet transform and GMM classification methods.

ESTIMATION OF ANKLE JOINT ANGLE FROM PERONEAL AND TIBIAL ELECTRONEUROGRAMS — A MUSCLE SPINDLE MODEL APPROACH

This study develops a method for estimating the angle of a passively stretched ankle joint from electroneurograms (ENGs) based on structural muscle spindle models of the tibial and peroneal nerves. Passive ramp-and-hold and alternating stretches of the ankle joint are performed on an anesthetized rabbit. Two cuff electrodes are employed to measure the ENGs of peroneal and tibial nerves simultaneously. From the two ENG signals and the joint angle trajectory, two intrafusal muscle fiber models are constructed and their inverse models are derived. The results of the two models are combined to generate the final angle estimate. An optimization method, called sequential quadratic programming, is employed to find the model parameters that minimize the squared errors between the ankle angles predicted by the model and the measured ankle angles. The performance of the proposed approach is compared with those of an adaptive neuro-fuzzy inference system and an artificial neural network model. The results reveal that the proposed model has the best performance in estimating the ankle joint angle in large-range movements and the smallest tracing error. The proposed method effectively estimates the passive ankle joint angle using the inverse physiological model of an intrafusal muscle fiber.

Selective recovery of fascicular activity in peripheral nerves

The peripheral nerves of an amputee's residual limb still carry the information required to provide the robust, natural control signals needed to command a dexterous prosthetic limb. However, these signals are mixed in the volume conductor of the body and extracting them is an unmet challenge. A beamforming algorithm was used to leverage the spatial separation of the fascicular sources, recovering mixed pseudo-spontaneous signals with normalized mean squared error of 0.14 ± 0.10 (n = 12) in an animal model. The method was also applied to a human femoral nerve model using computer simulations and recovered all five fascicular-group signals simultaneously with R(2) = 0.7 ± 0.2 at a signal-to-noise ratio of 0 dB. This technique accurately separated peripheral neural signals, potentially providing the voluntary, natural and robust command signals needed for advanced prosthetic limbs.

Neuro-fuzzy decoding of sensory information from ensembles of simultaneously recorded dorsal root ganglion neurons for functional electrical stimulation applications

Functional electrical stimulation (FES) is used to improve motor function after injury to the central nervous system. Some FES systems use artificial sensors to switch between finite control states. To optimize FES control of the complex behavior of the musculo-skeletal system in activities of daily life, it is highly desirable to implement feedback control. In theory, sensory neural signals could provide the required control signals. Recent studies have demonstrated the feasibility of deriving limb-state estimates from the firing rates of primary afferent neurons recorded in dorsal root ganglia (DRG). These studies used multiple linear regression (MLR) methods to generate estimates of limb position and velocity based on a weighted sum of firing rates in an ensemble of simultaneously recorded DRG neurons. The aim of this study was to test whether the use of a neuro-fuzzy (NF) algorithm (the generalized dynamic fuzzy neural networks (GD-FNN)) could improve the performance, robustness and ability to generalize from training to test sets compared to the MLR technique. NF and MLR decoding methods were applied to ensemble DRG recordings obtained during passive and active limb movements in anesthetized and freely moving cats. The GD-FNN model provided more accurate estimates of limb state and generalized better to novel movement patterns. Future efforts will focus on implementing these neural recording and decoding methods in real time to provide closed-loop control of FES using the information extracted from sensory neurons.

State-space decoding of primary afferent neuron firing rates

Kinematic state feedback is important for neuroprostheses to generate stable and adaptive movements of an extremity. State information, represented in the firing rates of populations of primary afferent (PA) neurons, can be recorded at the level of the dorsal root ganglia (DRG). Previous work in cats showed the feasibility of using DRG recordings to predict the kinematic state of the hind limb using reverse regression. Although accurate decoding results were attained, reverse regression does not make efficient use of the information embedded in the firing rates of the neural population. In this paper, we present decoding results based on state-space modeling, and show that it is a more principled and more efficient method for decoding the firing rates in an ensemble of PA neurons. In particular, we show that we can extract confounded information from neurons that respond to multiple kinematic parameters, and that including velocity components in the firing rate models significantly increases the accuracy of the decoded trajectory. We show that, on average, state-space decoding is twice as efficient as reverse regression for decoding joint and endpoint kinematics.

Estimation of ankle joint angle from peroneal and tibial electroneurograms based on muscle spindle model

The main goal of this study was to develop a new method of estimating the angle of the passively stretched ankle joint, based on structural muscle spindle models of the tibial and peroneal electroneurograms (ENG). Passive ramp-and-hold and alternating stretches of the ankle joint were performed in a rabbit. Simultaneously, two cuff electrodes were used to record the ENGs of peroneal and tibial nerves. Based on the two ENGs and the joint angle trajectory, two muscle spindle models were constructed and their inverse models were integrated to compute angle estimates. The model parameters were optimized. The performance of our approach was compared with those of the adaptive neuro-fuzzy inference system and artificial neural network model. The results revealed that our model had a better performance of estimating the ankle joint angle in large-range movements and smaller tracking errors. This study provides a new estimation algorithm to extract the joint angle from the information conveyed in a nerve.

Spike sorting of muscle spindle afferent nerve activity recorded with thin-film intrafascicular electrodes

Afferent muscle spindle activity in response to passive muscle stretch was recorded in vivo using thin-film longitudinal intrafascicular electrodes. A neural spike detection and classification scheme was developed for the purpose of separating activity of primary and secondary muscle spindle afferents. The algorithm is based on the multiscale continuous wavelet transform using complex wavelets. The detection scheme outperforms the commonly used threshold detection, especially with recordings having low signal-to-noise ratio. Results of classification of units indicate that the developed classifier is able to isolate activity having linear relationship with muscle length, which is a step towards online model-based estimation of muscle length that can be used in a closed-loop functional electrical stimulation system with natural sensory feedback.

Peripheral nerve signal recording and processing for artificial limb control

In order to take full advantage of modern multiple-degree of freedom prosthetic limbs, robust and natural control signals are needed. Previous work has shown that beamforming provides a method to extract such signals from peripheral nerve activity [1]. This paper describes in vivo experiments done to validate that method in a more realistic case. A 16-channel Flat Interface Nerve Electrode was used to record from the Sciatic nerve in Rabbit, while the distal Tibial and Peroneal branches were stimulated. Beamforming provided R(2)=0.7 ± 0.2, an improvement of 0.12 ± 0.06 over the a posteriori chosen best channels. When more realistic signals were generated using kHz-level stimulation, the beamforming filters were able to distinguish which branch was being stimulated, and in many cases how strongly, over a large range of stimulation intensities.

In vivo localization of fascicular activity

Users of modern high degree of freedom prosthetics need to provide a large number of natural, intuitive command signals in order to realize the high level of dexterity these devices offer. This level of natural control is beginning to be seen with new technologies like Targeted Muscle Reinnervation; however several serious drawbacks still exist. Flat Interface Nerve Electrode recordings provide a safe and stable means to record natural, intuitive volitional command signals. We investigate the use of Antenna Array techniques to separate command signals from different sources based on their spatial distribution within the nerve. Through a Rabbit sciatic model, it is shown that the system is able to separate compound action potentials elicited from the Tibial and Peroneal branches using 16-channel recordings made on the main sciatic nerve trunk.

Localization and recovery of peripheral neural sources with beamforming algorithms

The peripheral nervous system carries sensory and motor information that could be useful as command signals for function restoration in areas such as neural prosthetics and functional electrical stimulation (FES). Nerve cuff electrodes provide a robust and safe technique for recording nerve signals. However, a method to separate and recover signals from individual fascicles is necessary. Prior knowledge of the electrode geometry was used to develop an algorithm which assumes neither signal independence nor detailed knowledge of the nerve's geometry/conductivity, and is applicable to any wide-band near-field situation. When used to recover fascicular activities from simulated nerve cuff recordings in a realistic human femoral nerve model, this beamforming algorithm separates signals as close as 1.5 mm with cross-correlation coefficient, R > 0.9 (10% noise). Ten simultaneous signals could be recovered from individual fascicles with only a 20% decrease in R compared to a single signal. At high noise levels (40%), sources were localized to 180 +/- 170 microm in the 12 x 3 mm cuff. Localizing sources and using the resulting positions in the recovery algorithm yielded R = 0.66 +/- 0.10 in 10% noise for five simultaneous muscle-activation signals from synergistic fascicles. These recovered signals should allow natural, robust, closed-loop control of multiple degree-of-freedom prosthetic devices and FES systems.

On the use of longitudinal intrafascicular peripheral interfaces for the control of cybernetic hand prostheses in amputees

Significant strides have been recently made to develop highly sensorized cybernetic prostheses aimed at restoring sensorimotor limb functions to those who have lost them because of a traumatic event (amputation). In these cases, one of the main goals is to create a bidirectional link between the artificial devices (e.g., robotic hands, arms, or legs) and the nervous system. Several human-machine interfaces (HMIs) are currently used to this aim. Among them, interfaces with the peripheral nervous system and in particular longitudinal intrafascicular electrodes can be a promising solution able to improve the current situation. In this paper, the potentials and limits of the use of this interface to control robotic devices are presented. Specific information is provided on: 1) the neurophysiological bases for the use peripheral nerve interfaces; 2) a comparison of the potentials of the different peripheral neural interfaces; 3) the possibility of extracting and appropriately interpreting the neural code for motor commands and of delivering sensory feedback by stimulating afferent fibers by using longitudinal intrafascicular electrodes; 4) a preliminary comparative analysis of the performance of this approach with the ones of others HMIs; 5) the open issues which have to be addressed for a chronic usability of this approach.

Localization and control of activity in peripheral nerves

Interest in the field of the natural control of human limb using physiological signals has risen dramatically in the past 20 years due to the success of the brain machine interface. Cortical signals carry significant information but are difficult to access. The peripheral nerves of the body carry both command and sensory signals and are far more accessible. While numerous studies have documented the selective stimulation properties of, conventionally round, nerve cuff electrodes (i.e., transverse geometry) and even self-sizing electrodes, recording the activity levels from individual fascicles using these electrodes is still an unsolved problem. Moreover, the control algorithms for the control of joint movement with multiple contact electrodes such as the flat interface nerve electrode (FINE) have been difficult to implement. We propose solutions to both these problems by using beam forming techniques to detect the location and the activity in various fascicles. We also developed a control algorithm that separates the dynamic from the passive properties to solve the redundancy problem in multiple joint problems. This techniques could find application in the natural control of artificial limbs from peripheral nerve signals for patients with amputated limbs or to restore function in patients with stroke or paralyzed limbs.

Bidirectional interfaces with the peripheral nervous system

Considerable scientific and technological efforts have been devoted to develop neuroprostheses and hybrid bionic systems that link the human nervous system with electronic or robotic prostheses, with the main aim of restoring motor and sensory functions in disabled patients. Such developments have also the potential to be applied to normal human beings to improve their physical capabilities for bidirectional control and feedback of machines. A number of neuroprostheses use interfaces with peripheral nerves or muscles for neuromuscular stimulation and signal recording. This chapter provides a general overview of the peripheral neural interfaces available and their use from research to clinical application in controlling artificial and robotic prostheses and in developing neuroprostheses. Extraneural electrodes, such as cuff and epineurial electrodes, provide simultaneous interface with many axons in the nerve, whereas intrafascicular, penetrating, and regenerative electrodes may selectively contact small groups of axons within a nerve fascicle. Biological and technical issues are reviewed relative to the problems of electrode design and tissue injury. The last sections review different strategies for the use of peripheral neural interfaces in biomedical applications.

Solution space reduction in the peripheral nerve source localization problem using forward field similarities

Improving our ability to localize bioelectric sources within a peripheral nerve would help us to monitor the control signals flowing to and from any limb or organ. This technology would provide a useful neuroscience tool, and could perhaps be incorporated into a neuroprosthesis interface. We propose to use measurements from a multi-contact nerve cuff to solve an inverse problem of bioelectric source localization within the peripheral nerve. Before the inverse problem can be addressed, the forward problem is solved using finite element modeling. A fine mesh improves the accuracy of the forward problem solution, but increases the number of variables to be solved for in the inverse problem. To alleviate this problem, variables corresponding to mesh elements that are not distinguishable by the measurement setup are grouped together, thus reducing the dimension of the inverse problem without impacting on the forward problem accuracy. A quantitative criterion for element distinguishability is derived using the columns of the leadfield matrix and information about the uncertainty in the measurements. Our results indicate that the number of variables in the inverse problem can be reduced by more than half using the proposed method, without having a detrimental impact on the quality of the localization.

Classification of afferent signals recorded with a single cuff electrode

The implementation of systems to restore sensorimotor functions in person with neurological disabilities is an important research area. In the past, many studies have been carried out to develop closed-loop neuroprostheses based on the processing of electroneurographic (ENG) signals recorded from physiological sensors using cuff electrodes. However, the potential of this approach is not completely clear. In this paper, an artificial neural network is used to discriminate afferent ENG signals evoked by different mechanical stimuli and recorded with a single cuff electrode. The preliminary results indicate that even single cuff ENG signals can be useful to extract interesting information with good performance. In the future, the possibility of discriminating additional stimuli using additional channels and more advanced classification techniques will be investigated.

Model-based ankle joint angle tracing by cuff electrode recordings of peroneal and tibial nerves

The main goal of the present study was to estimate the ankle joint angle from the peroneal and tibial electroneurography (ENG) recordings. Two single-channel cuff electrodes for recording ENG were placed on the proximal part of rabbit peroneal and tibial nerves respectively and static positioning and ramp-and-hold stretches were performed to characterize the static and dynamic ENG responses. An ENG model, consisting of static and dynamic parts, was constructed to relate ENG to ankle angle trajectory and an inverse ENG model was derived to predict ankle angle. The results showed that the new model could accurately estimate large-range ankle angles during and after ramp-and-hold movements. Our study provides a basis for implementing joint angle tracing without using artificial angle sensors.

Design of a cybernetic hand for perception and action

Strong motivation for developing new prosthetic hand devices is provided by the fact that low functionality and controllability-in addition to poor cosmetic appearance-are the most important reasons why amputees do not regularly use their prosthetic hands. This paper presents the design of the CyberHand, a cybernetic anthropomorphic hand intended to provide amputees with functional hand replacement. Its design was bio-inspired in terms of its modular architecture, its physical appearance, kinematics, sensorization, and actuation, and its multilevel control system. Its underactuated mechanisms allow separate control of each digit as well as thumb-finger opposition and, accordingly, can generate a multitude of grasps. Its sensory system was designed to provide proprioceptive information as well as to emulate fundamental functional properties of human tactile mechanoreceptors of specific importance for grasp-and-hold tasks. The CyberHand control system presumes just a few efferent and afferent channels and was divided in two main layers: a high-level control that interprets the user's intention (grasp selection and required force level) and can provide pertinent sensory feedback and a low-level control responsible for actuating specific grasps and applying the desired total force by taking advantage of the intelligent mechanics. The grasps made available by the high-level controller include those fundamental for activities of daily living: cylindrical, spherical, tridigital (tripod), and lateral grasps. The modular and flexible design of the CyberHand makes it suitable for incremental development of sensorization, interfacing, and control strategies and, as such, it will be a useful tool not only for clinical research but also for addressing neuroscientific hypotheses regarding sensorimotor control.

A critical review of interfaces with the peripheral nervous system for the control of neuroprostheses and hybrid bionic systems

Considerable scientific and technological efforts have been devoted to develop neuroprostheses and hybrid bionic systems that link the human nervous system with electronic or robotic prostheses, with the main aim of restoring motor and sensory functions in disabled patients. A number of neuroprostheses use interfaces with peripheral nerves or muscles for neuromuscular stimulation and signal recording. Herein, we provide a critical overview of the peripheral interfaces available and trace their use from research to clinical application in controlling artificial and robotic prostheses. The first section reviews the different types of non-invasive and invasive electrodes, which include surface and muscular electrodes that can record EMG signals from and stimulate the underlying or implanted muscles. Extraneural electrodes, such as cuff and epineurial electrodes, provide simultaneous interface with many axons in the nerve, whereas intrafascicular, penetrating, and regenerative electrodes may contact small groups of axons within a nerve fascicle. Biological, technological, and material science issues are also reviewed relative to the problems of electrode design and tissue injury. The last section reviews different strategies for the use of information recorded from peripheral interfaces and the current state of control neuroprostheses and hybrid bionic systems.

Experiments on the development and use of a new generation of intra-neural electrodes to control robotic devices

The development of interfaces linking the human nervous system with artificial devices is an important area of research and several groups are now addressing it. Interfaces represent the key enabling technology for the development of devices usable for the restoration of motor and sensory function in subjects affected by neurological disorders, injuries or amputations. For example, current hand prostheses use electromyographic (EMG) signals to extract volitional commands but this limits the possibility of controlling several degrees of freedom and of delivering sensory feedback. To achieve these goals, implantable neural interfaces are required. Among the candidate interfaces with the peripheral nervous system intra-neural electrodes seem to be an interesting solution due to their bandwidth and ability to access volition and deliver sensory feedback. However, several drawbacks have to be addressed in order to increase their usability. In this paper, experiments to address many of these issues are presented as part of the development of a new generation of intra-neural electrodes. The results showed seem to confirm that these new interfaces seem to have interesting properties and that they can represent a significant improvement of the state of the art. Extensive experiments will be carried out in the future to validate these results.

On the Development of a Biomechatronic System to Record Tendon Sliding Movements

The main goal of this paper is to study the feasibility of a novel implantable micro-system able to record information about tendon sliding movements by using contactless measurement devices (magnetic sources and sensors). The system, named "Biomechatronic Position Transducer" (BPT), can be used for the implementation of advanced control strategies in neuroprostheses. After a preliminary analysis based on finite element model simulations, an experimental setup was developed in order to simulate the recording conditions (the sensors fixed to the bones and the magnetic sources placed on the tendons). In order to limit the number of implanted components of the system, a fuzzy Mamdani-like architecture was developed to extract the information from the raw data. The results confirm the possibility of using the presented approach for developing an implantable micro-sensor able to extract kinematic information useful for the control of neuroprostheses. Future works will go in the direction of integrating and testing the sensors and the electronic circuitry (to provide power supply and to record the data) during in vitro and in situ experiments.

Estimation of peroneal and tibial afferent activity from a multichannel cuff placed on the sciatic nerve

The afferent signals recorded with a multi-electrode cuff on the sciatic nerve were employed to investigate the possibility of extracting components ascending from the peroneal and tibial nerves. Two methods, an inverse regression model and principal component method, were studied. The parameters of inverse regression model, determined by data collected in semistatic conditions, were validated by data collected in dynamic conditions. The results showed that the regression model, which used only two channels of the sciatic recordings, was sufficient to separate the distal afferent components. The model, at the expense of requiring distal branch recordings for estimating model parameters, yielded better separation than the principal component method. In conclusion, peroneal and tibial afferent activity can be estimated from the sciatic nerve: the principal component method is suitable for applications focused on acquiring afferent information, whereas the inverse regression model is better for applications in which stimulations will be applied to the branches. The estimation technique provides a powerful tool for in vivo investigation of sensory information transmitted in a peripheral nerve and facilitates implementation of advanced functional neuromuscular stimulation systems.

A sensorized thumb for force closed-loop control of hand neuroprostheses

In this paper, we presented a sensorized thumb based on a matrix of piezoresistive force sensors, with an acquisition unit and a special wearing support. The sensor was calibrated and then the device was tested during different tasks simulating activities of daily living performed by seven able-bodied subjects. By means of these experiments, we verified that the device proposed can be used to extract force information during grasp. In fact, the device was able to provide useful force information in the 98% of the trials with a good repeatability during all the different conditions. Moreover, we evaluated the patterns obtained during the different grasping tasks. The palmar grasps were performed in a similar manner, whereas the lateral pinch and the spherical volar grip were more different. This device can provide force information with good performance and acceptability and it can be used for force closed-loop control of hand neuroprostheses.

On the intersubject generalization ability in extracting kinematic information from afferent nervous signals

In the recent past, many efforts have been carried out in order to evaluate the feasibility of implementing closed-loop controlled neuroprostheses based on the processing of sensory electroneurographic (ENG) signals. The success of these techniques mostly relies on the development of processing algorithms capable of extracting the necessary kinematic information from these signals. Soft-computing algorithms can be very useful when dealing with the complexity of the neuromuscular system because of their generalization ability and model-free structure. In this paper, these techniques were used to extract angular position information from the ENG signals recorded from muscle afferents in animal model using cuff electrodes. Specifically, a genetic algorithm-based dynamic nonsingleton fuzzy logic system (named GA-DNSFLS) was developed and tested on different types of angular trajectories (characterized by small or large angular excursions). In particular, two different Takagi-Sugeno-Kang (TSK)-like structures were used in the consequent part of the neuro-fuzzy model in order to verify which one could improve the generalization abilities (intrasubject and intersubject). The results showed that the GA-DNSFLS was able to reconstruct the trajectories giving interesting results in terms of correlation between the actual and the predicted trajectories for small excursion movements during intrasubject and intersubject tests. Particularly, one of the TSK models showed better results in terms of intersubject generalization. The simulations conducted with the large excursion movements led in some cases to interesting results but further experiments are necessary in order to analyze this point more in deep.

Biopotentials as command and feedback signals in functional electrical stimulation systems

Today Functional Electrical Stimulation (FES) is available as a clinical tool in muscle activation used for picking up objects, for standing and walking, for controlling bladder emptying, and for breathing. Despite substantial progress in development and new knowledge, many challenges remain to be resolved to provide a more efficient functionality of FES systems. The most important task of these challenges is to improve control of the activated muscles through open loop or feedback systems. Command and feedback signals can be extracted from biopotentials recorded from muscles (Electromyogram, EMG), nerves (Electroneurogram, ENG), and the brain (Electroencephalogram (EEG) or individual cells). This paper reviews work in which EMG, ENG, and EEG signals in humans have been used as command and feedback signals in systems using electrical stimulation of motor nerves to restore movements after an injury to the Central Nervous System (CNS). It is concluded that the technology is ready to push for more substantial clinical FES investigations in applying muscle and nerve signals. Brain-computer interface systems hold great prospects, but require further development of faster and clinically more acceptable technologies.

Improving signal reliability for on-line joint angle estimation from nerve cuff recordings of muscle afferents

Closed-loop functional electrical stimulation (FES) applications depend on sensory feedback, thus, it is important to continuously investigate new methods to obtain reliable feedback signals. The objective of the present paper was to examine the feasibility of using an artificial neural network (ANN) to predict joint angle from whole nerve cuff recordings of muscle afferent activity within a physiological range of motion. Furthermore, we estimated how small changes in joint angle that can be detected from the nerve cuff recordings. Neural networks were tested with data obtained from ten acute rabbit experiments in simulated, on-line experiments. The electroneurograms (ENG) of the tibial and peroneal nerves were recorded during passive ankle joint rotation. To decrease the joint angle prediction error with new rabbit data, we attempted to pretune the nerve signals and re-trained the ANNs with the pretuned data. With these procedures we were able to compensate for interrabbit variability. On average the mean prediction errors were less than 2.0 degrees (a total excursion of 20 degrees) and we were able to predict joint angles from muscle afferent activity with accuracy close to the best-estimated angular resolution. The angular resolution was found to depend on the initial joint angle and the actual step size taken and we found that there was a low probability of detecting joint angle changes less than 1.5 degrees. We thus suggest that muscle afferent activity is applicable as feedback in real-time closed-loop control, when the motion speed is restricted and when the movement is limited to a portion of the joint's physiological range.