Journal of Neural Engineering

ObjectiveTo preserve the encoding of visual information in prosthetic vision as close to natural as possible, subretinal photovoltaic implants, which replace the lost photoreceptors, strive to stimulate the second-order retinal neurons, the bipolar cells, while avoiding direct activation of the downstream retinal ganglion cells. To assess the range of such selective subretinal activation, we implanted the devices in rodent models of retinal degeneration and measured the stimulation thresholds based on the visually evoked potentials. After assessment of the bipolar cell-mediated thresholds, direct activation of retinal ganglion cells was measured following intraocular injection of synaptic blockers. Since these chemicals are toxic to the retina, this procedure can only be done once in each animal. ApproachWe developed a machine-learning model that identifies the stimulation pathway directly from the recorded visually evoked potentials, eliminating the need for synaptic blockers. The model was trained on recordings from rats implanted with PRIMA subretinal arrays and evaluated on two additional implant architectures, a second rat species, and a different anesthesia protocol. Main ResultsThe classifier achieved a balanced accuracy of 92% in cross-validation on the training data. Generalization to all unseen experimental conditions yielded an average balanced accuracy of 91%. Integrated Gradients analysis showed that combined bipolar and ganglion cell responses were driven by the early P1 component, while bipolar cell responses relied on later waveform components, consistent with thalamocortical processing dynamics. SignificanceThe described computational alternative to pharmacological blockers should improve the experimental throughput, allow multiple recordings over the lifetime of the same animal, and might be applicable to optimization of the stimulation settings in patients.

ObjectiveOur prior studies have demonstrated that lateral spinal cord stimulation can evoke somatosensory percepts in the missing foot in individuals with a lower-limb amputation. However, subjects reported concurrent sensations in their residual limb. In this study, we evaluate the hypothesis that using high-density paddle electrodes with smaller contact sizes, and multipolar stimulation configurations could evoke more focal sensations in the foot over a wide range of stimulation amplitudes. ApproachWe used a combination of electrophysiology and computational modelling methods to investigate the selective activation of distal nerve branches in response to lateral spinal cord stimulation in cats. In six acute feline experiments, we performed an L3-S1 laminectomy and placed custom 32-electrode paddles laterally over the dura of the spinal cord. We recorded antidromic action potentials in the distal branches of the sciatic and femoral nerve trunks in response to stimulation using three contact diameters (150, 500 and 1000 {micro}m) and two stimulation configurations - monopolar and bipolar stimulation. We replicated the neural recruitment patterns from those experiments in a computational model of the feline lumbar spinal cord. We then used the model to examine neural recruitment with 1.8 mm and 2.5 mm contacts, as well as a tripolar guarded-cathode configuration. Main resultsIn the electrophysiology experiments, the 500 {micro}m-diameter electrodes achieved the most selective nerve activation (68%) compared to 62% for both 150 and 1000 {micro}m-diameter electrodes. The minimum amplitudes for recruiting nerve branches (i.e., threshold) as well as the dynamic ranges were largely similar for the different contact diameters (median: 35 {micro}A) and stimulation configurations (30 {micro}A for bipolar stimulation; 35 {micro}A for monopolar stimulation). The computational model reproduced the finding that selectivity did not differ significantly among the three contact sizes tested in cat experiments, though it revealed that increasing contact diameter above 1000 {micro}m raised the minimum amplitude required for selective activation and reduced spinal root selectivity. Across both approaches, we consistently recruited large-diameter afferents that are critical for somatosensory applications of spinal cord stimulation. SignificanceOur results indicate that, relative to clinical electrodes, reducing the contact diameter of stimulation electrodes can evoke focal sensations, but further reductions below 1000 {micro}m may fail to improve selectivity. This study highlights potential constraints with achieving focal selectivity that are not dependent on the design of the electrodes.

ObjectiveBiopotential measurement devices, such as EEG, ECG, and EMG recorders, are available in low-cost, open-source implementations with standard specifications. However, high-end systems remain expensive and predominantly proprietary, limiting accessibility and customization by research laboratories. In addition, neurophysiology techniques such as bioimpedance-based Fast Neural Electrical Impedance Tomography (FN-EIT) also rely on these systems for data acquisition. This work aimed to develop an open-source biopotential recording system using off-the-shelf components that achieves performance comparable to high-end devices. ApproachWe designed our system to provide simultaneous sampling over 32 channels, 24-bit resolution, 10 kHz bandwidth, 50 kHz sampling rate, and battery-powered operation while reducing cost by two orders of magnitude. System performance was evaluated comparatively against a reference device. Initial validation involved benchtop recordings in saline solution and standard non-invasive biopotential measurements (ECG and EMG). Further in-vivo validation was performed by recording evoked electrophysiological responses and FN-EIT traces from the sciatic nerve of a rat during tibial branch stimulation. Main resultsEMG recordings showed comparable RMS peak amplitudes (814{+/-}153 {micro}V vs. 897{+/-}113{micro}V, p=0.07), while ECG-derived heart rates closely matched between systems (64.8{+/-}4.0 bpm vs. 65.1{+/-}3.1bpm, p=0.54). During in-vivo recordings, compound action potentials exhibited comparable amplitudes and morphology (129{+/-}26 mV vs 128{+/-}25 mV, P=0.15). FN-EIT recordings showed strongly correlated baseline voltages (R>0.93, P=0.11), sub-microvolt noise levels (0.83{+/-}0.36{micro}V vs. 0.42{+/-}0.25{micro}V, p<0.05), and comparable impedance variations (0.006{+/-}0.002% vs 0.005{+/-}0.003, P>0.05). FN-EIT images of functional activity recorded with the novel device closely matched reference ones, exhibiting a 98.5% overlap in activated area. SignificanceThe proposed open-source device has the potential to broaden research access to customizable, high-specification data acquisition hardware and facilitate wider adoption of specialized neural recording techniques such as FN-EIT.

Extracellular sinusoidal low frequency alternating current (LFAC) stimulation of peripheral motor nerves has been observed to induce size wise activation of nerve fibers, unlike the inverse recruitment order typically seen in extracellular pulsed stimulation. This study aims to explore potential biophysical mechanisms responsible for this phenomenon using computational modeling. Volume conductor model was utilized with a bipolar cuff electrode encasing a single rat-sized fascicle. The extracellular potentials generated by LFAC and pulse stimulation were projected onto the McIntyre-Richardson-Grill models of myelinated motor nerve fibers to examine the activation of fibers ranging from 5.7 to 16m in diameter. Intracellular and extracellular stimulation were compared for strength-frequency relationships with LFAC (1-20Hz) and strength-duration curves for pulse stimulation. The threshold tracking technique was used to study membrane electrotonus and threshold electrotonus of different fibers to examine subthreshold accommodation in response to LFAC and prolonged pulse stimulation. The simulations revealed that the inverse order of fiber recruitment is an inherent characteristic of extracellular stimulation and is theoretically independent of the stimulation waveform. LFAC showed an inverse strength-frequency relationship (higher frequency, lower threshold current), similar to the inverse strength-duration relationship for pulsed stimulation. Analysis of subthreshold accommodation showed that larger fibers exhibit greater accommodation than smaller fibers, leading to increased activation thresholds as fast Na+ activation factor m3h decreases while slow K+ activation increases, supporting accommodation as a contributor to orderly recruitment. With increasing LFAC frequency (up to 20Hz), these accommodation characteristics were reduced and large-fiber state dynamics shifted toward those of smaller fibers. LFAC was also found to induce subthreshold oscillations that promoted spike initiation during slow depolarization. These findings suggest that LFAC provides a controlled and optimized method for achieving orderly recruitment without the need for complex selective blocking protocols. By leveraging intrinsic membrane properties, LFAC offers a neuromodulation strategy that preserves physiological recruitment order, with direct implications for selective nerve stimulation in clinical and neuroprosthetic applications. Author summaryElectrical stimulation is widely used to activate peripheral nerves in motor rehabilitation and neuroprosthetic devices, but conventional pulse stimulation activates larger nerve fibers first (with lower current intensity), which can induce rapid muscle fatigue and pain. We used well-established and validated computational models of motor nerve fibers (axons) to explore how sinusoidal low frequency alternating current (LFAC) stimulation can produce a more physiological, size-wise recruitment order. We simulated myelinated motor nerve fibers of different diameters individually inside a bipolar cuff electrode and analyzed how the membrane and ion channels changed during stimulation levels that are below thresholds for action potential firing. We found that larger fibers adapt (accommodate) more strongly during the slow depolarization of LFAC: their sodium channels become less open, while potassium activation increases, raising the current required to induce an action potential. Smaller fibers were less affected by this accommodation effect and could reach firing at lower current intensities. We also found that these effects depend on stimulation frequency; at lower frequencies, the accommodation characteristics were more defined (for all fibers), while at higher frequencies they were reduced (for large fibers) and all fiber responses became more similar. Our results suggest that the responses of intrinsic membrane dynamics to LFAC lead fiber recruitment toward a more physiological order, which facilitates the design of safer and more selective nerve stimulation strategies with LFAC.

Cochlear implants (CI) partially restore hearing in profoundly hearing impaired or deaf people by electrically stimulating the auditory nerve. A bottleneck of electrical CIs is the broad spread of electrical current from each electrode that limits the transfer of spectral information, which might be overcome by future spatially confined optogenetic stimulation. Here we established an in silico framework, FraSCO, to model sound encoding in the human cochlea by an optogenetic CI (oCI) for testing the potential of optogenetic hearing restoration. The biophysical modeling framework combined an optical raytracing model implementing a human cochlea implanted with a waveguide-based oCI with a single compartment model of optogenetically modified spiral ganglion neurons (SGNs). The input was an optogenetic sound coding strategy and the quality of the neural representation was evaluated based on comparison of neurograms evoked by optogenetic and electrical stimulation to the spectrogram of the sound applied. The model aimed for technologically feasible properties of the oCIs with 64 stimulation channels. The biophysical modeling framework successfully captured essential physiological features of optogenetic SGN stimulation with a minimal set of ion channel types expressed in the SGN soma. Working with a sample of 1000 SGNs distributed along the tonotopic axis to represent sound encoding, we found that improved spectral selectivity more than compensates for lower temporal fidelity of current implementations of optogenetic stimulation. The established computational framework enables in silico investigation and benchmarking of sound encoding in the cochlea by future oCI and state-of-the-art eCI. The results indicate that optogenetic sound encoding has potential to improve speech understanding in noisy environments for CI users.

Objective: Adaptive neuromodulation systems and implantable brain-computer interfaces (BCIs) are promising therapies for neurological and psychiatric disorders. However, their broader translation into research and clinical practice remains limited by technological complexity, restricted access to implantable research platforms, and the lack of standardized, reproducible experimental workflows. We therefore aimed to develop and validate an open, general-purpose translational ecosystem that enables rapid development, evaluation, and dissemination of novel neuromodulation and implantable BCI paradigms. Approach: The CorTec Brain Interchange (BIC) implantable neural sensing and stimulation device was integrated with the open-source BCI2000 platform to create a modular, extensible neuromodulation ecosystem. We established a standardized battery of quantitative assessments to characterize implantable neuromodulation systems to comprehensively evaluate the CorTec BIC device through benchtop characterization, long-term preclinical in vitro and in vivo validation, and a human proof-of-concept demonstration. Results: Benchtop and saline testing provided a comprehensive technical ex vivo characterization of the BIC device, independently validating previously reported performance while extending its characterization through quantification of the recording noise floor, stimulation and acquisition latencies and impedance measurement accuracy. Long-term in vivo validation in five canines, with the longest implantation exceeding three years, demonstrated stable chronic recordings while capturing progressive channel deterioration and its underlying mechanical causes. The ecosystem enabled active functional decoding more than two years after implantation, implementation of closed-loop stimulation using arbitrary spectral biomarkers, detection and modulation of epilepsy-associated biomarkers, and brain stimulation evoked potential recordings. In addition, we translated an established one-dimensional BCI cursor control paradigm to the BIC benchtop evaluation kit and demonstrated its feasibility in a human participant. Finally, we openly provide standardized surgical, imaging, and analysis pipelines together with datasets and software to facilitate reproducible neuromodulation research. Significance: We present a versatile, open-source translational ecosystem that supports a wide range of neuromodulation and implantable BCI applications with minimal modification. This battery of quantitative assessments can be applied generally as a blueprint for systematic characterization of implantable neuromodulation systems. By combining comprehensive hardware characterization with standardized software tools and experimental workflows, this work provides both an essential reference for researchers adopting the Brain Interchange platform. The ecosystem lowers technical barriers to implantable neurotechnology research, promotes reproducibility, and provides a foundation for accelerating the development and clinical translation of next-generation adaptive neuromodulation and implantable BCI therapies for patients with neurological and psychiatric disorders.

Understanding the neuromuscular properties that allow dexterous manipulation of objects remains a major challenge in neurorehabilitation, largely due to the difficulty of characterizing intrinsic hand muscle activity. These muscles are small, densely packed, and anatomically complex, making selective recordings with intramuscular electromyography (EMG) technically demanding and impractical for comprehensive studies. In this work, we present a custom, high-density (HD) surface EMG grid designed to non-invasively capture activity from intrinsic hand muscles from both dorsal and palmar surfaces. We evaluated the quality and spatial selectivity of the recordings by directly comparing them with intramuscular EMG signals obtained from the dorsal and palmar interossei. Surface EMG signals corresponded closely to the intramuscular recordings, with high correlation values for all subjects and tasks. Double differential spatial filtering significantly improved selectivity, although some residual volume conduction remained. The dorsal grid primarily captured dorsal interossei activity, while the palmar grid was more sensitive to lumbrical activation. The palmar interossei recordings were spatially more varied, with the second palmar interosseous predominantly detected on the dorsal grid and the third and fourth on the palmar grid. Together, these results demonstrate that non-invasive HD surface EMG will allow more complete measurement of intrinsic muscle activity, to provide a better understanding of the complex relation between the intrinsic and extrinsic hand muscles during dexterous movements. This basic information will allow refinement of biomechanical hand models and prosthetic devices, and the development of biomimetic brain computer interfaces aimed at restoring natural hand function after neurological injury.

P300 spellers convert electroencephalographic (EEG) activity into text by presenting users with a matrix of flickering characters. While these systems can achieve high classification accuracy, communication is severely slowed by the need for many stimulus repetitions to obtain a reliable signal. Reducing repetitions accelerates spelling but introduces character-level errors: insertions, deletions, and substitutions that degrade usability and increase user fatigue. Although substantial research has focused on improving performance at the signal acquisition and decoding stages, here we investigate a complementary text post-processing approach that leverages large language models (LLMs) to restore corrupted P300 speller output. We constructed a dataset derived from cLang-8 and simulated realistic P300-style text corruption using both random and empirically derived human-like error strategies. We evaluated several instruction-tuned LLMs alongside an optical character recognition (OCR)-fine-tuned ByT5 model under zero-shot and few-shot prompting conditions. We found that LLMs effectively recovered clean text from noisy inputs. Models employing SentencePiece tokenization consistently outperformed byte-pair encoding (BPE)-based counterparts, and few-shot in-context learning further improved restoration accuracy, with Gemma 3 achieving the strongest performance across all settings. These results suggest that LLM-based post-processing could enable P300 speller systems to operate with fewer repetitions and lower latency while maintaining or improving output accuracy, offering a practical path toward more efficient daily communication for users with motor and speech disabilities.

ObjectiveSpinal evoked compound action potentials (ECAPs) provide a quantitative measure of the neural response during spinal cord stimulation (SCS) and can be leveraged in closed-loop applications to control dose in response to spinal cord movement. However, interpretation of ECAPs recorded in vivo is limited by susceptibility to noise, inter-subject variability, and other confounding factors. As SCS systems evolve and the clinical and research applications of ECAPs expand, it is critical to understand how physiological and technical factors influence ECAP generation and morphology. ApproachWe used a computational modeling framework to systematically investigate the influence of anatomical (e.g., dorsal cerebrospinal fluid (dCSF) thickness), stimulation (e.g., pulse width, waveform shape, stimulation configuration, stimulation frequency), and recording configurations on the neural responses and ECAPs generated during SCS. We employed a hybrid computational modeling approach, coupling finite element method models with multicompartment axon models to simulate neural responses to SCS. Using these models, we characterized the spatiotemporal dynamics of neural recruitment and the resulting ECAP waveforms. Main resultsNeural responses and model ECAPs were strongly influenced by factors, such as dCSF thickness, pulse width, and stimulation waveform shape. Stimulation parameters introduced trade-offs between axonal recruitment thresholds, neural activation selectivity, and ECAP timing and morphology. Notably, similar ECAP amplitudes could obscure differences in the underlying neural recruitment. Complex ECAP morphologies also emerged in response to distinct stimulation paradigms, reflecting changes in the spatiotemporal properties of axonal activation. Additionally, we demonstrate that the selection of recording electrodes can be optimized to enhance recorded ECAP amplitudes. SignificanceOur findings provide a theoretical framework to advance our mechanistic understanding of SCS-induced ECAPs and offer insights into optimization strategies to improve closed-loop SCS therapies.

jaxon is an open-source, fully differentiable and GPU-native reimplementation of the canonical peripheral-nerve fiber models in JAX/Jaxley: the myelinated McIntyre-Richardson-Grill (MRG) and Sweeney axons and the un-myelinated Sundt and Rattay C-fibers. It reproduces NEURONs extracellular mechanism through a custom backward-Euler coupled intracellular/periaxonal double-cable solver, agreeing with PyFibers-wrapped NEURON on 99.6% of 943 activation-threshold configurations within 1% and matching conduction velocity to machine precision. Because the entire forward model is expressed in JAX, it is both vectorized--simulating whole fiber populations in parallel and reaching a geometric-mean [~]820x speedup at N = 100,000 fibers on a single GPU--and differentiable, so extracellular-stimulation parameters (per-contact amplitudes, waveform shape, and electrode position) can be optimized directly through the cable equation rather than grid-searched. jaxon slots into existing peripheral-nerve modeling pipelines as a gradient-enabled, population-scale replacement for the NEURON forward solver.

ObjectiveLateral spinal cord stimulation (LSCS) is a promising approach for restoring somatosensory feedback in lower-limb amputees, but its spatial selectivity remains limited. Percepts often spread to unintended regions of the residual limb, and reducing electrode contact size may not improve focality. This study investigated whether the anatomical organization of lumbar dorsal rootlets (DR) imposes fundamental constraints on LSCS selectivity. ApproachAcute neurophysiology experiments were performed in six adult cats. Both LSCS and individual DR stimulation were conducted in the same animals. For DR stimulation, bipolar hook electrodes were used to stimulate individual DR, while antidromic compound action potentials (CAPs) were recorded from femoral and sciatic nerve branches instrumented with nerve cuffs. For LSCS, custom 32-contact epidural paddle electrodes were placed over the lateral surface of the spinal cord at corresponding vertebral levels. Recruitment thresholds, dynamic ranges, and response patterns were analyzed across spinal levels, and DR recruitment patterns were directly compared to those evoked by LSCS within the same animals. Main resultsA clear rostrocaudal organization was observed across spinal levels during stimulation of individual DR, with femoral branches predominantly recruited at L4-L5 and sciatic branches at L6-L7. However, no somatotopic organization was found across DR within each spinal level; individual DR frequently co-activated multiple branches within the same group, and selective recruitment could only be maintained over a narrow dynamic range (median [~]10 {micro}A). LSCS exhibited even a narrower dynamic range ([~]5 {micro}A) but closely mirrored DR recruitment patterns, indicating that LSCS activates sensory afferents in a manner determined by the organizational structure of the DR. SignificanceThese findings demonstrate that the limited spatial selectivity of LSCS can largely be attributed to the coarse organization of DR within each root level rather than due to limitations of epidural electrode design. Moving electrodes intradurally or reducing contact size further is unlikely to substantially improve focality. Instead, improving paddle stability to ensure consistent placement over the appropriate spinal levels may be a more effective strategy for enhancing percept localization.

Brain-computer interfaces (BCIs) offer a promising solution to speech loss due to neurological injury by decoding intended speech directly from brain activity. While recent BCIs have restored high-accuracy text-based communication, they fail to provide instantaneous voice output essential for the natural flow of conversation. Brain-to-voice BCIs address this gap by decoding voice directly from neural signals. However, even the state-of-the-art (SOTA) BCI-synthesized voice is not yet intelligible enough for real-world adoption. We introduce brain2voice 2.0, a new multimodal Transformer-based BCI decoder architecture capable of synthesizing highly intelligible voice from intracortical neural signals in real-time. Brain2voice 2.0 is trained on continuous and custom-tokenized acoustic targets and phoneme targets, leveraging their complementary speech information. We use self-supervised and adversarial training objectives that enhance acoustic feature quality and improve synthesis intelligibility. At each 10 ms timestep, the model causally outputs continuous and tokenized acoustic features for real-time voice synthesis as well as time-aligned phoneme predictions (raw phoneme error rate: 7%, comparable to the latest brain-to-text models). We evaluated this new approach on our prior intracortical brain-to-voice benchmark dataset (Wairagkar et al. 2025). Naive human listeners transcribed brain2voice 2.0 synthesized voice with a word error rate of 5.24%--an 8x improvement in intelligibility over previous SOTA results (43.75%). Brain2voice 2.0 demonstrates that highly intelligible real-time voice synthesis from neural signals is achievable, for the first time crossing the intelligibility threshold necessary for clinically viable brain-to-voice BCIs for people with paralysis.

Non-invasive brain-computer interfaces (BCIs) have substantially advanced in the field of continuous cursor control over the past decade. Yet, current methods lack key control aspects such as initiation and termination of cursor movements as well as evaluation in real-world applications. In this study, we introduce a framework for continuous, electroencephalography-based cursor control that supports both active movement and no-movement states, thereby allowing for inactive periods of the user when no control input is desired. We demonstrate its applicability in healthy participants and show its performance in real-world application through the selection of targets on a screen. This demonstrates that participants can leverage the continuous control cursor control and the intentional starting and stopping of motions to effectively select targets on a screen through dwell-time selection. On average, 7.1 out of 40 targets were correctly selected (level of significant performance: 4.5 targets), while experienced BCI users achieved an average of 12.8 targets. The proposed framework additionally demonstrates compatibility with motor-impaired people without residual hand motions since it does not rely on observable movements for model training.

Deep brain stimulation (DBS) modeling relies heavily on biophysical neuron models to estimate neural activation thresholds and predict stimulation spread. In this study, we systematically compared a widely adopted axon model, the McIntyre-Richardson-Grill (MRG) model (Model I), with a more detailed biophysical model, the Cohen model (Model II), to assess how structural and electrophysiological differences affect predicted DBS outcomes. Electric field distributions generated by 2202 DBS lead were applied to the neuron models as extracellular input stimuli. Both models were simulated under biphasic pulse stimulation across varying axon-electrode distances, pulse widths, and stimulation frequencies. Activation distances ranged from approximately 2 to 10 mm depending on stimulation parameters and contact location. At 2 mA, Model I achieved an activation distance of 6 mm, whereas Model II reached 10 mm, indicating greater excitability. Across matched fiber tracts, threshold differences ranged from -1.40 mA to 0.27 mA, with Model II requiring lower thresholds in 97.7% of cases. Both models showed a strong inverse relationship between pulse width and activation threshold. However, frequency responses differed: Model II exhibited increasing thresholds at higher frequencies, while Model I showed a slight decrease. Machine learning regressors trained on distance, pulse width, and frequency achieved high predictive accuracy, with Gradient Boosting performing best. Model II demonstrated superior prediction metrics (R^2 = 0.986; RMSE = 0.045 mA; MAE = 0.034 mA) compared to Model I (R^2 = 0.977; RMSE = 0.089 mA; MAE = 0.068 mA). Overall, both models reliably estimate DBS-induced activation, but structural differences significantly affect excitability and frequency-dependent behavior. With appropriate awareness of their respective strengths and limitations, either model can be used to derive activation distances for estimating electric field isolevels and the volume of tissue activated in patient-specific DBS simulations.

Surface electromyography (sEMG) is a robust non-invasive modality for human-machine interaction, yet its application remains largely limited to coarse motor tasks such as grasping or rotation. The decoding of fine motor skills, specifically handwriting, remains a challenging problem with potential relevance for prosthetic control and natural communication interfaces. In this work, we explore a Transformer-based alternative to classical signal-processing pipelines that treats multi-channel sEMG signals as complex time series. We introduce DualMyo, a specialized model integrating Patch Embeddings and Rotary Positional Embeddings (RoPE) to capture the intricate spatio-temporal dynamics of myoelectric activity. Our experimental results show strong intra-session performance. Furthermore, we address the inherent challenges of signal drift and sensor displacement in cross-session applications. We show that a lightweight fine-tuning strategy of 10 epochs enables DualMyo to effectively adapt to session variability, achieving approximately 91\% accuracy with two examples per digit. These findings provide a promising step toward adaptive sEMG-based handwriting interfaces, although further validation is required for real-time and multi-subject deployment and neuromuscular control.

Computational models of peripheral nerve stimulation--coupling finite-element bioelectric fields to biophysical axon models--have become essential for designing electrodes and waveforms for neuromodulation therapies. Yet the established open tools are code-first and assume substantial modeling expertise, and several depend on commercial finite-element solvers, placing realistic nerve modeling out of reach for many experimentalists and clinicians. We present golgi, an open-source platform that takes a peripheral nerve from image to stimulated fiber population through a single graphical interface, with an equivalent scriptable Python API and command-line interface for batch studies. golgi integrates the full pipeline: image segmentation (or import of surfaces or masks), automated multi-region tetrahedral meshing, anisotropic finite-element solution of the extracellular field with explicit perineurium contact impedance, generation of realistic fiber populations and their three-dimensional trajectories (straight, or curved via a quasi-static streamline solver), and biophysical activation thresholds through interchangeable NEURON and a GPU-accelerated surrogate backend. We demonstrate golgi on extruded multifascicular swine and human cervical vagus nerves and on real three-dimensional, branching human and rabbit vagus nerves reconstructed from micro-computed tomography. It reproduces the physiological fiber-diameter recruitment order, quantifies fascicular selectivity and current steering with a multi-contact cuff electrode, and resolves anatomically defined nerve branches. Using this branch resolution, we find that selectively engaging a vagal cardiac branch from a proximal cuff depends on anatomy. In the rabbit, whose cardiac fibers are predominantly small and whose superior cardiac branch forms a discrete, spatially segregated tract, current steering isolates even its small cardiac (B-type) fibers; in the human cervical vagus only the large myelinated fibers are separable, because the high thresholds of the small cardiac fibers force stimulus currents that also recruit off-target fibers. Every study can be exported as an integrity-hashed, self-contained bundle whose finite-element-to-recruitment provenance is verifiable byte-for-byte with a single command--a reproducibility guarantee absent from existing tools. By combining non-specialist usability, anatomical realism, and verifiable reproducibility in one open package, golgi lowers the barrier to in-silico peripheral nerve stimulation modeling. golgi is freely available as open-source software. Author summaryElectrical stimulation of peripheral nerves treats a growing range of conditions, from epilepsy to inflammatory and cardiovascular disease. Deciding where to place an electrode and how to shape the stimulus increasingly relies on computer models that combine the electric field around the electrode with detailed models of how individual nerve fibers respond. We found that existing software for this is powerful but primarily designed for expert modelers: it generally requires programming, substantial modeling expertise, and sometimes expensive commercial software, which can limit its adoption by experimentalists and clinicians. We built golgi to remove that barrier. With golgi, a user can go from a nerve image all the way to predicted fiber recruitment through a single point-and-click interface, while advanced users keep full scripting control. golgi builds anatomically realistic nerve models, simulates how different fiber types and fascicles are recruited, and lets users compare electrode designs. Using golgi, we also found that whether a small but clinically important nerve branch--such as the cardiac branch of the vagus nerve--can be selectively stimulated depends on its anatomy. In a rabbit nerve, where this branch forms a discrete, spatially separated bundle and its fibers are mostly small, even its small fibers can be targeted from a cuff on the main trunk; in the human, only the large fibers can be reached selectively, because the small cardiac fibers are harder to excite and the stronger currents needed to reach them also activate off-target fibers. Critically, every result can be packaged so that anyone else can verify it reproduces exactly--making peripheral nerve stimulation models easier to build, share, and trust.

Brain-computer interfaces face a fundamental trade-off between the signal fidelity and stimulation precision of noninvasive systems and the surgical burden and scalability of invasive systems. Non-invasive BCIs suffer from low signal quality and poor stimulation accuracy due to the skull barrier and the variability introduced by the scalp and skull. Existing invasive BCIs rely on traumatic surgical procedures or brain-penetrating electrodes, which limits their spatial extensibility, application, and patient acceptance. Here, we introduce a minimally invasive hybrid BCI architecture that uses the skull as a distributed interface layer rather than treating it solely as a barrier. The hybrid BCI comprises four integrated components: (1) the safe and smart micro-hole craniotomy; (2) distributed microelectrodes subcutaneously implanted in micro-holes in the skull with the distal end in contact with the dura; (3) an external bi-directional wearable headset for coupling, recording, stimulation, and channel selection; and (4) an AI-assisted planning and control agent. Animal studies have shown that micro-holes with a diameter of 300-800 m can be safely and conveniently prepared at any predefined locations across the skull without impairing the dura. In vivo experiments on rats demonstrate that the hybrid BCI with skull-implanted microelectrodes evidently increases resting-state spectral power and improves the signal-to-noise ratio of somatosensory and steady-state visual evoked responses compared to the scalp EEG; the computational modelling shows that distributed skull-dura microelectrodes can increase the intracranial electrical field strength and steer focused temporal-interference fields towards predefined deep brain targets. These findings will lay a solid foundation for future endeavors in wireless integration, safety evaluation and clinical benefits of the hybrid BCI. In summary, we propose the hybrid BCI as a distinct minimally invasive BCI paradigm with the great potential as a distributed, scalable, and upgradable neural interface that can expand the clinical application of minimally invasive BCI techniques.

Kilohertz-frequency (kHz) electrical stimulation (1-100 kHz) is emerging as a powerful tool in both invasive and non-invasive neurostimulation applications, including functional electrical stimulation, spinal cord stimulation, and non-invasive brain stimulation. Most commonly these paradigms rely on amplitude modulation (AM-kHz)--achieved via burst or sinusoidal modulation--to produce phasic neural activation. Here, we propose, and validate, an alternative: frequency-modulated kilohertz stimulation (FM-kHz). This approach leverages the distinct strength-frequency dependence of kHz signals, whereby higher carrier frequencies are less efficient in depolarizing neurons than lower frequencies. By sweeping between sub- and suprathreshold frequencies, at a constant amplitude, FM-kHz generates a phasic neural activation envelope analogous to AM-kHz, without requiring amplitude modulation. Using both computational modelling and experimental data from Locusta migratoria (N5 nerve) and the human median nerve, we demonstrate that FM-kHz stimulation: 1. Produces reliable phasic evoked responses at the FM frequency; 2. Enables two degrees of control over stimulation--via FM frequency and frequency deviation. Across models tested, FM-kHz thresholds followed the same increasing strength-frequency relationship as AM-kHz, with FM-kHz requiring modestly higher thresholds at the upper end of the tested frequency range. These findings position FM-kHz as a viable and potentially advantageous alternative to AM-kHz strategies for future neuromodulation devices, and conceptually ground strength-frequency dependence as the key parameter in interpreting the effects of kHz electrical stimulation.

BackgroundDirect current (DC) stimulation can modulate neuronal activity in ways that differ from pulsatile stimulation, but its intracranial use has been limited by concerns about tissue injury at the electrode/tissue interface. Quantitative safety limits for DC delivered through metal electrodes directly to the brain remain poorly defined. ObjectiveTo estimate histological safety boundaries for DC stimulation delivered through metal electrodes in a large-brain gyrencephalic animal model. MethodsCathodal DC stimulation was applied to the exposed cortical surface of ten anesthetized sheep using platinum-iridium disc electrodes typically used in clinical applications (surface area [&le;] 4.15 mm2). Currents of 5 to 1000 {micro}A were delivered for 10 to 15 minutes at 36 cortical sites. Stimulation dose was quantified as charge density. Brains were removed shortly after stimulation and examined histologically for tissue damage, including necrosis, inflammation, gliosis, and demyelination. Lesion volumes were quantified and related to charge density. ResultsNo lesions were observed at sites where no current or a low charge density (0.7 mC/mm2) was delivered. With stimulation, lesion probability and volume increased with charge density, although variability was substantial. Lesions occurred in 3 of 18 sites at lower charge densities (1.4 to 10 mC/mm2) and in 5 of 9 sites at higher charge densities (14.4 to 144.4 mC/mm2). Linear regression of lesion volume against charge density yielded an estimated zero-lesion intercept of 2.3 mC/mm2, whereas alternative nonlinear models predicted thresholds up to 8.7 mC/mm2. ConclusionThese findings suggest that it may be possible to apply cathodal DC stimulation directly to the cortical surface through metal electrodes without detectable histological damage when current intensity, duration, and electrode size are appropriately constrained. These findings provide quantitative guidance for the safe application of DC directly to neural tissue in experimental and translational neuromodulation studies.

Low-cost and open-source neural recording systems are increasingly important for expanding access to electrophysiological research. However, many existing platforms still rely on specialized hardware or limited modularity, restricting flexibility for laboratories seeking customizable solutions. Here, we developed and evaluated a modular neural recording platform constructed entirely from commercially available components. Recordings were compared against the ground truth. The platform successfully recovered local field potential (LFP)-like waveforms in most conditions and detected spike-like activity during direct connection recordings. Principal component analysis and k-means clustering further demonstrated the ability to distinguish multiple simulated spike waveforms. Signal quality varied across configurations, with saline recordings and preamplifier integration introducing increased noise and reduced detectability. These findings demonstrate the feasibility of building affordable and modular electrophysiology systems using widely accessible hardware. Although the current implementation has limitations in sampling rate, noise performance, and in vivo validation, the presented framework provides a practical foundation for future customizable open-source neural recording.