Neuromorphic hardware for somatosensory neuroprostheses.

In individuals with sensory-motor impairments, missing limb functions can be restored using neuroprosthetic devices that directly interface with the nervous system. However, restoring the natural tactile experience through electrical neural stimulation requires complex encoding strategies. Indeed, they are presently limited in effectively conveying or restoring tactile sensations by bandwidth constraints. Neuromorphic technology, which mimics the natural behavior of neurons and synapses, holds promise for replicating the encoding of natural touch, potentially informing neurostimulation design. In this perspective, we propose that incorporating neuromorphic technologies into neuroprostheses could be an effective approach for developing more natural human-machine interfaces, potentially leading to advancements in device performance, acceptability, and embeddability. We also highlight ongoing challenges and the required actions to facilitate the future integration of these advanced technologies.

The efficacy of hybrid neuroprostheses in the rehabilitation of upper limb impairment after stroke, a narrative and systematic review with a meta-analysis.

BACKGROUND: Paresis of the upper limb (UL) is the most frequent impairment after a stroke. Hybrid neuroprostheses, i.e., the combination of robots and electrical stimulation, have emerged as an option to treat these impairments. METHODS: To give an overview of existing devices, their features, and how they are linked to clinical metrics, four different databases were systematically searched for studies on hybrid neuroprostheses for UL rehabilitation after stroke. The evidence on the efficacy of hybrid therapies was synthesized. RESULTS: Seventy-three studies were identified, introducing 32 hybrid systems. Among the most recent devices (n = 20), most actively reinforce movement (3 passively) and are typical exoskeletons (3 end-effectors). If classified according to the International Classification of Functioning, Disability and Health, systems for proximal support are expected to affect body structures and functions, while the activity and participation level are targeted when applying Functional Electrical Stimulation distally plus the robotic component proximally. The meta-analysis reveals a significant positive effect on UL functions (p < 0.001), evident in a 7.8-point Mdiff between groups in the Fugl-Meyer assessment. This positive effect remains at the 3-month follow-up (Mdiff = 8.4, p < 0.001). CONCLUSIONS: Hybrid neuroprostheses have a positive effect on UL recovery after stroke, with effects persisting at least three months after the intervention. Non-significant studies were those with the shortest intervention periods and the oldest patients. Improvements in UL functions are not only present in the subacute phase after stroke but also in long-term chronic stages. In addition to further technical development, more RCTs are needed to make assumptions about the determinants of successful therapy.

A high-performance speech neuroprosthesis.

Speech brain-computer interfaces (BCIs) have the potential to restore rapid communication to people with paralysis by decoding neural activity evoked by attempted speech into text1,2 or sound3,4. Early demonstrations, although promising, have not yet achieved accuracies sufficiently high for communication of unconstrained sentences from a large vocabulary1-7. Here we demonstrate a speech-to-text BCI that records spiking activity from intracortical microelectrode arrays. Enabled by these high-resolution recordings, our study participant-who can no longer speak intelligibly owing to amyotrophic lateral sclerosis-achieved a 9.1% word error rate on a 50-word vocabulary (2.7 times fewer errors than the previous state-of-the-art speech BCI2) and a 23.8% word error rate on a 125,000-word vocabulary (the first successful demonstration, to our knowledge, of large-vocabulary decoding). Our participant's attempted speech was decoded  at 62 words per minute, which is 3.4 times as fast as the previous record8 and begins to approach the speed of natural conversation (160 words per minute9). Finally, we highlight two aspects of the neural code for speech that are encouraging for speech BCIs: spatially intermixed tuning to speech articulators that makes accurate decoding possible from only a small region of cortex, and a detailed articulatory representation of phonemes that persists years after paralysis. These results show a feasible path forward for restoring rapid communication to people with paralysis who can no longer speak.

A high-performance neuroprosthesis for speech decoding and avatar control.

Speech neuroprostheses have the potential to restore communication to people living with paralysis, but naturalistic speed and expressivity are elusive1. Here we use high-density surface recordings of the speech cortex in a clinical-trial participant with severe limb and vocal paralysis to achieve high-performance real-time decoding across three complementary speech-related output modalities: text, speech audio and facial-avatar animation. We trained and evaluated deep-learning models using neural data collected as the participant attempted to silently speak sentences. For text, we demonstrate accurate and rapid large-vocabulary decoding with a median rate of 78 words per minute and median word error rate of 25%. For speech audio, we demonstrate intelligible and rapid speech synthesis and personalization to the participant's pre-injury voice. For facial-avatar animation, we demonstrate the control of virtual orofacial movements for speech and non-speech communicative gestures. The decoders reached high performance with less than two weeks of training. Our findings introduce a multimodal speech-neuroprosthetic approach that has substantial promise to restore full, embodied communication to people living with severe paralysis.

Automatic calibration of electrode arrays for dexterous neuroprostheses: a review.

Background. Electrode arrays can simplify the modulation of shape, size, and position for customized stimulation delivery. However, the intricacy in achieving the desired outcome stems from optimizing for the myriad of possible electrode combinations and stimulation parameters to account for varying physiology across users.Objective. This study reviews automated calibration algorithms that perform such an optimization to realize hand function tasks. Comparing such algorithms for their calibration effort, functional outcome, and clinical acceptance can aid with the development of better algorithms and address technological challenges in their implementation.Methods. A systematic search was conducted across major electronic databases to identify relevant articles. The search yielded 36 suitable articles; among them, 14 articles that met the inclusion criteria were considered for the review.Results. Studies have demonstrated the realization of several hand function tasks and individual digit control using automatic calibration algorithms. These algorithms significantly improved calibration time and functional outcomes across healthy and people with neurological deficits. Also, electrode profiling performed via automated algorithms was very similar to a trained rehabilitation expert. Additionally, emphasis must be given to collecting subject-specific a priori data to improve the optimization routine and simplify calibration effort.Conclusion. With significantly shorter calibration time, delivering personalized stimulation, and obviating the need for an expert, automated algorithms demonstrate the potential for home-based rehabilitation for improved user independence and acceptance.

Dense, 11 V-Tolerant, Balanced Stimulator IC with Digital Time-Domain Calibration for 100 nA Error.

This article presents a multichannel neurostimulator implementing a novel charge balancing technique to achieve maximal integration. Safe neurostimulation demands accurate charge balancing of the stimulation waveforms to prevent charge build-up on the electrode-tissue interface. We propose digital time-domain calibration (DTDC), which adjusts the second phase of the biphasic stimulation pulses digitally, based on a one-time characterization of all stimulator channels with an on-chip ADC. Accurate control of the stimulation current amplitude is loosened in exchange for time-domain corrections, relieving circuit matching constraints and consequentially saving channel area. A theoretical analysis of DTDC is presented, establishing expressions for the required time resolution and the new, relaxed circuit matching constraints. To validate the DTDC principle, a 16-channel stimulator was implemented in 65 nm CMOS, requiring only 0.0141 mm 2 area/channel. Despite being implemented in a standard CMOS technology, 10.4 V compliance is achieved for compatibility with high-impedance microelectrode arrays typical for high-resolution neural prostheses. To the authors' knowledge, this is the first stimulator in a 65 nm low-voltage process achieving over 10 V output swing. Measurements after calibration show the DC error is successfully reduced below 96 nA on all channels. Static power consumption is 20.3 µW/channel.

Advanced electrochemical potential monitoring for improved understanding of electrical neurostimulation protocols.

Objective.Current-controlled neurostimulation is increasingly used in the clinical treatment of neurological disorders and widely applied in neural prostheses such as cochlear implants. Despite its importance, time-dependent potential traces of electrodes during microsecond-scale current pulses, especially with respect to a reference electrode (RE), are not precisely understood. However, this knowledge is critical to predict contributions of chemical reactions at the electrodes, and ultimately electrode stability, biocompatibility, and stimulation safety and efficacy.Approach.We assessed the electrochemistry of neurostimulation protocolsin vitrowith Pt microelectrodes from millisecond (classical electroanalysis) to microsecond (neurostimulation) timescales. We developed a dual-channel instrumentation amplifier to include a RE in neurostimulation setups. Uniquely, we combined potential measurements with potentiostatic prepolarization to control and investigate the surface status, which is not possible in typical stimulation setups.Main results.We thoroughly validated the instrumentation and highlighted the importance of monitoring individual electrochemical electrode potentials in different configurations of neurostimulation. We investigated electrode processes such as oxide formation and oxygen reduction by chronopotentiometry, bridging the gap between milli- and microsecond timescales. Our results demonstrate how much impact on potential traces the electrode's initial surface state and electrochemical surface processes have, even on a microsecond scale.Significance.Our unique use of preconditioning in combination with stimulation reveals that interpreting potential traces with respect to electrode processes is misleading without rigorous control of the electrode's surface state. Especiallyin vivo, where the microenvironment is unknown, simply measuring the voltage between two electrodes cannot accurately reflect the electrode's state and processes. Potential boundaries determine charge transfer, corrosion, and alterations of the electrode/tissue interface such as pH and oxygenation, particularly in long-termin vivouse. Our findings are relevant for all use-cases of constant-current stimulation, strongly advocating for electrochemicalin situinvestigations in many applications like the development of new electrode materials and stimulation methods.

Invisible consequences of paralysis.

Neuroprosthetic technologies can control blood pressure and restore walking.

Conformable Electrode Arrays for Wearable Neuroprostheses.

Wearable electrode arrays can selectively stimulate muscle groups by modulating their shape, size, and position over a targeted region. They can potentially revolutionize personalized rehabilitation by being noninvasive and allowing easy donning and doffing. Nevertheless, users should feel comfortable using such arrays, as they are typically worn for an extended time period. Additionally, to deliver safe and selective stimulation, these arrays must be tailored to a user's physiology. Fabricating customizable electrode arrays needs a rapid and economical technique that accommodates scalability. By leveraging a multilayer screen-printing technique, this study aims to develop personalizable electrode arrays by embedding conductive materials into silicone-based elastomers. Accordingly, the conductivity of a silicone-based elastomer was altered by adding carbonaceous material. The 1:8 and 1:9 weight ratio percentages of carbon black (CB) to elastomer achieved conductivities between 0.0021-0.0030 S cm-1 and were suitable for transcutaneous stimulation. Moreover, these ratios maintained their stimulation performance after several stretching cycles of up to 200%. Thus, a soft, conformable electrode array with a customizable design was demonstrated. Lastly, the efficacy of the proposed electrode arrays to stimulate hand function tasks was evaluated by in vivo experiments. The demonstration of such arrays encourages the realization of cost-effective, wearable stimulation systems for hand function restoration.

Significantly reduced inflammatory foreign-body-response to neuroimplants and improved recording performance in young compared to adult rats.

The multicellular inflammatory encapsulation of implanted intracortical multielectrode arrays (MEA) is associated with severe deterioration of their field potentials' (FP) recording performance, which thus limits the use of brain implants in basic research and clinical applications. Therefore, extensive efforts have been made to identify the conditions in which the inflammatory foreign body response (FBR) is alleviated, or to develop methods to mitigate the formation of the inflammatory barrier. Here, for the first time, we show that (1) in young rats (74±8 gr, 4 weeks old at the onset of the experiments), cortical tissue recovery following MEA implantation proceeds with ameliorated inflammatory scar as compared to adult rats (242 ± 18 gr, 9 weeks old at the experimental onset); (2) in contrast to adult rats in which the Colony Stimulating factor 1 Receptor (CSF1R) antagonist chow eliminated ∼95% of the cortical microglia but not microglia adhering to the implant surfaces, in young rats the microglia adhering to the implant were eliminated along with the parenchymal microglia population. The removal of microglia adhering to the implant surfaces was correlated with improved recording performance by in-house fabricated Perforated Polyimide MEA Platforms (PPMP). These results support the hypothesis that microglia adhering to the surface of the electrodes, rather than the multicellular inflammatory scar, is the major underlying mechanism that deteriorates implant recording performance, and that young rats provide an advantageous model to study months-long, multisite electrophysiology in freely behaving rats. STATEMENT OF SIGNIFICANCE: Multisite electrophysiological recordings and stimulation devices play central roles in basic brain research and medical applications. The insertion of multielectrode-array platforms into the brain's parenchyma unavoidably injures the tissue, and initiates a multicellular inflammatory cascade culminating in the formation of an encapsulating scar tissue (the foreign body response-FBR). The dominant view, which directs most current research efforts to mitigate the FBR, holds that the FBR is the major hurdle to effective electrophysiological use of neuroprobes. By contrast, this report demonstrates that microglia adhering to the surface of a neuroimplants, rather than the multicellular FBR, underlie the performance deterioration of neuroimplants. These findings pave the way to the development of novel and focused strategies to overcome the functional deterioration of neuroimplants.

Neurorobotics and neuroprostheses: Towards a new anatomy.

The idea of this Special Issue arose from the technological advances in bionic, robotic, and neural rehabilitation systems and the common need to comprehend in detail how human anatomical structures can be replicated or controlled. Motor control theories, among others, include the generalized control program theory, the equilibrium point hypothesis, or the optimal control approach in which neural commands to the muscles are a result of the central nervous system solving an optimization problem for a specific cost function. No matter the alternative interpretation selected to replicate biological control of human movements, artificial "anatomies" should consider not only motor capabilities from the central nervous system but integrate bioinspired mechanical features (such as compliance) in artificial limbs. The development of wearable robotics and neuroprosthetic systems for human movement compensation and control is naturally inspired by human anatomy and biology. Cutting-edge technological advances in the field of biomedical and neural engineering are bringing us more and more close to a new artificial anatomy with which humans could augment their motor capabilities or replace them after they are compromised. Either augmentative/assistive or rehabilitation technologies in the near future will require engineering solutions based on novel approaches to create usable neurorobotic and neuroprosthetic systems for the most relevant societal needs.

A Comparison of Delay-and-Add and Maximum Likelihood Estimation for Velocity-Selective Recording Using Multi-Electrode Cuffs.

Extracting information from the peripheral nervous system with implantable devices remains a significant challenge that limits the advancement of closed-loop neural prostheses. Linear electrode arrays can record neural signals with both temporal and spatial selectivity, and velocity selective recording using the delay-and-add algorithm can enable classification based on fibre type. The maximum likelihood estimation method also measures velocity and is frequently used in electromyography but has never been applied to electroneurography. Therefore, this study compares the two algorithms using in-vivo recordings of electrically evoked compound action potentials from the ulnar nerve of a pig. The performance of these algorithms was assessed using the velocity quality factor (Q-factor), computational time and the influence of the number of channels. The results show that the performance of both algorithms is significantly influenced by the number of channels in the recording array, with accuracies ranging from 77% with only two channels to 98% for 11 channels. Both algorithms were comparable in accuracy and Q-factor for all channels, with the delay-and-add having a slight advantage in the Q-factor.

Ethical implications of visual neuroprostheses-a systematic review.

Objective. The aim of this review was to systematically identify the ethical implications of visual neuroprostheses.Approach. A systematic search was performed in both PubMed and Embase using a search string that combined synonyms for visual neuroprostheses, brain-computer interfaces (BCIs), cochlear implants (CIs), and ethics. We chose to include literature on BCIs and CIs, because of their ethically relavant similarities and functional parallels with visual neuroprostheses.Main results. We included 84 articles in total. Six focused specifically on visual prostheses. The other articles focused more broadly on neurotechnologies, on BCIs or CIs. We identified 169 ethical implications that have been categorized under seven main themes: (a) benefits for health and well-being; (b) harm and risk; (c) autonomy; (d) societal effects; (e) clinical research; (f) regulation and governance; and (g) involvement of experts, patients and the public.Significance. The development and clinical use of visual neuroprostheses is accompanied by ethical issues that should be considered early in the technological development process. Though there is ample literature on the ethical implications of other types of neuroprostheses, such as motor neuroprostheses and CIs, there is a significant gap in the literature regarding the ethical implications of visual neuroprostheses. Our findings can serve as a starting point for further research and normative analysis.

Brain implant enables man in locked-in state to communicate.

Despite complete paralysis from amyotrophic lateral sclerosis, person used neural signals to spell out thoughts.

A Wireless Power and Data Transfer IC for Neural Prostheses Using a Single Inductive Link With Frequency-Splitting Characteristic.

This paper presents a frequency-splitting-based wireless power and data transfer IC that simultaneously delivers power and forward data over a single inductive link. For data transmission, frequency-shift keying (FSK) is utilized because the FSK modulation scheme supports continuous wireless power transmission without disruption of the carrier amplitude. Moreover, the link that manifests the frequency-splitting characteristic due to a close distance between coupled coils provides wide bandwidth for data delivery without degrading the quality factors of the coils. It results in large power delivery, high data rate, and high power transfer efficiency. The presented IC fabricated in a 180-nm BCD process simultaneously achieves up-to-115-mW wireless power delivery to the load and 2.5-Mb/s downlink data rate over the single inductive link. The measured overall power efficiency from the DC power supply at the transmitter module to the load at the receiver module reaches 56.7 % at its maximum, and the bit error rate is lower than 10 -6 at 2.5 Mb/s. As a result, the figure of merit (FoM) for data transmission is enhanced by 2 times, and the FoM for power delivery is improved by 38.7 times compared to prior state-of-the-arts using a single inductive link.

Real-time linear prediction of simultaneous and independent movements of two finger groups using an intracortical brain-machine interface.

Modern brain-machine interfaces can return function to people with paralysis, but current upper extremity brain-machine interfaces are unable to reproduce control of individuated finger movements. Here, for the first time, we present a real-time, high-speed, linear brain-machine interface in nonhuman primates that utilizes intracortical neural signals to bridge this gap. We created a non-prehensile task that systematically individuates two finger groups, the index finger and the middle-ring-small fingers combined. During online brain control, the ReFIT Kalman filter could predict individuated finger group movements with high performance. Next, training ridge regression decoders with individual movements was sufficient to predict untrained combined movements and vice versa. Finally, we compared the postural and movement tuning of finger-related cortical activity to find that individual cortical units simultaneously encode multiple behavioral dimensions. Our results suggest that linear decoders may be sufficient for brain-machine interfaces to execute high-dimensional tasks with the performance levels required for naturalistic neural prostheses.

Real-time synthesis of imagined speech processes from minimally invasive recordings of neural activity.

Speech neuroprosthetics aim to provide a natural communication channel to individuals who are unable to speak due to physical or neurological impairments. Real-time synthesis of acoustic speech directly from measured neural activity could enable natural conversations and notably improve quality of life, particularly for individuals who have severely limited means of communication. Recent advances in decoding approaches have led to high quality reconstructions of acoustic speech from invasively measured neural activity. However, most prior research utilizes data collected during open-loop experiments of articulated speech, which might not directly translate to imagined speech processes. Here, we present an approach that synthesizes audible speech in real-time for both imagined and whispered speech conditions. Using a participant implanted with stereotactic depth electrodes, we were able to reliably generate audible speech in real-time. The decoding models rely predominately on frontal activity suggesting that speech processes have similar representations when vocalized, whispered, or imagined. While reconstructed audio is not yet intelligible, our real-time synthesis approach represents an essential step towards investigating how patients will learn to operate a closed-loop speech neuroprosthesis based on imagined speech.

Closed-Loop Neural Prostheses With On-Chip Intelligence: A Review and a Low-Latency Machine Learning Model for Brain State Detection.

The application of closed-loop approaches in systems neuroscience and therapeutic stimulation holds great promise for revolutionizing our understanding of the brain and for developing novel neuromodulation therapies to restore lost functions. Neural prostheses capable of multi-channel neural recording, on-site signal processing, rapid symptom detection, and closed-loop stimulation are critical to enabling such novel treatments. However, the existing closed-loop neuromodulation devices are too simplistic and lack sufficient on-chip processing and intelligence. In this paper, we first discuss both commercial and investigational closed-loop neuromodulation devices for brain disorders. Next, we review state-of-the-art neural prostheses with on-chip machine learning, focusing on application-specific integrated circuits (ASIC). System requirements, performance and hardware comparisons, design trade-offs, and hardware optimization techniques are discussed. To facilitate a fair comparison and guide design choices among various on-chip classifiers, we propose a new energy-area (E-A) efficiency figure of merit that evaluates hardware efficiency and multi-channel scalability. Finally, we present several techniques to improve the key design metrics of tree-based on-chip classifiers, both in the context of ensemble methods and oblique structures. A novel Depth-Variant Tree Ensemble (DVTE) is proposed to reduce processing latency (e.g., by 2.5× on seizure detection task). We further develop a cost-aware learning approach to jointly optimize the power and latency metrics. We show that algorithm-hardware co-design enables the energy- and memory-optimized design of tree-based models, while preserving a high accuracy and low latency. Furthermore, we show that our proposed tree-based models feature a highly interpretable decision process that is essential for safety-critical applications such as closed-loop stimulation.

Neuroprosthesis for Decoding Speech in a Paralyzed Person with Anarthria.

BACKGROUND: Technology to restore the ability to communicate in paralyzed persons who cannot speak has the potential to improve autonomy and quality of life. An approach that decodes words and sentences directly from the cerebral cortical activity of such patients may represent an advancement over existing methods for assisted communication. METHODS: We implanted a subdural, high-density, multielectrode array over the area of the sensorimotor cortex that controls speech in a person with anarthria (the loss of the ability to articulate speech) and spastic quadriparesis caused by a brain-stem stroke. Over the course of 48 sessions, we recorded 22 hours of cortical activity while the participant attempted to say individual words from a vocabulary set of 50 words. We used deep-learning algorithms to create computational models for the detection and classification of words from patterns in the recorded cortical activity. We applied these computational models, as well as a natural-language model that yielded next-word probabilities given the preceding words in a sequence, to decode full sentences as the participant attempted to say them. RESULTS: We decoded sentences from the participant's cortical activity in real time at a median rate of 15.2 words per minute, with a median word error rate of 25.6%. In post hoc analyses, we detected 98% of the attempts by the participant to produce individual words, and we classified words with 47.1% accuracy using cortical signals that were stable throughout the 81-week study period. CONCLUSIONS: In a person with anarthria and spastic quadriparesis caused by a brain-stem stroke, words and sentences were decoded directly from cortical activity during attempted speech with the use of deep-learning models and a natural-language model. (Funded by Facebook and others; ClinicalTrials.gov number, NCT03698149.).