Journal of Neural Engineering

Brain-machine interfaces (BMIs) for vision restoration require models that accurately simulate the anatomy and electrical properties of visual pathways. However, current models focus only on isolated structures, such as the retina or brain, and overlook surrounding tissues. Here, we present a comprehensive computational model of the human head, incorporating the entire visual pathway--including the eye, optic nerve, and brain--along with critical neighboring tissues such as the orbit, paranasal sinuses, enabling precise simulations. Validation using human and large animal data demonstrated a strong correlation between the simulated and measured electrical potentials. Component elimination analysis revealed that the optimized comprehensive model outperformed simplified versions. The models utility was demonstrated through multiple applications: (1) comparative analysis of electrical neuromodulation technologies for optic neuropathy, revealing the filed intensity limitations of noninvasive approaches and the safety concerns of invasive intraorbital approach; (2) identification of optimal stimulation site, revealing that transnasal stimulation at the optic chiasm outperformed traditional approaches; and (3) in silico design of electrode arrays for optic nerve prosthetics, demonstrating theoretical advantages in invasiveness and visual field coverage compared to existing retinal and cortical prosthetics. This validated and versatile computational resource supports the development of neuromodulation strategies and visual BMI technologies.

BackgroundPeripheral nerve stimulation (PNS) is most effective when specific nerve fiber subpopulations are activated, while minimizing off-target activation, which may cause undesirable side effects. This selectivity depends primarily on electrode design and charge delivery. We hypothesized that selective PNS could be achieved through electrode placement and intrafascicular electric field steering using Longitudinal Intrafascicular Electrodes (LIFEs). MethodsLIFEs were implanted into the tibial fascicle of the sciatic nerve of 17 anesthetized adult rats. We tested whether electrodes positioned at different cross-sectional and longitudinal locations within the same fascicle, together with different electric field-steering approaches produced distinct activation patterns in the gastrocnemius lateralis muscle. Muscle responses were measured using high-density epimysial electromyography (HD-eEMG). ResultsElectrodes placed at different locations within the same fascicle activated distinct muscle regions, demonstrating intrafascicular selectivity. Bipolar stimulation recruited nerve fibers differently than monopolar stimulation, showing that electric field steering can further shape the selective recruitment. In both configurations, increasing the stimulation amplitude produced a graded increase in muscle activation. Furthermore, our findings demonstrated that HD-eEMG is an effective tool for evaluating intrafascicular selectivity. ConclusionThese findings suggest that improving on-target selectivity may support next-generation bioelectronic therapies with better outcomes and fewer side effects, potentially enabling more precise, organ-specific neuromodulation. Using multiple intrafascicular electrodes may provide two complementary strategies for enhancing selectivity: strategic intrafascicular placement to access different fiber subpopulations and bipolar configurations to steer recruitment beyond what a single electrode can achieve.

Reconstructing speech envelopes from electroen-cephalography (EEG) signals is a challenging but valuable task for brain-computer interfaces (BCIs), with applications in assistive communication for individuals with speech impairments. While deep learning has improved reconstruction accuracy, most existing approaches are restricted to single-layer architectures such as convolutional neural networks (CNNs). This limits their ability to capture the full complexity of spatio-temporal and structural EEG patterns. In this work, we systematically extend the VLAAI framework by evaluating 26 architectures that integrate CNNs, long short-term memory networks (LSTMs), and graph convolutional networks (GCNs) in both single-layer and hybrid configurations. Experiments on the 64-channel Spar-rKULee dataset demonstrate that CNNs remain the strongest standalone models, but hybrid designs--particularly CNN-LSTM and CNN-GCN-LSTM--achieve competitive or superior performance. These results highlight the importance of combining spatial, temporal, and graph-based processing, and provide practical guidelines for hybrid architecture design. Our study offers the first large-scale comparative analysis of hybrid models for EEG-based speech envelope reconstruction, advancing robust BCI systems for non-invasive speech decoding.

Brain-Computer interfaces (BCIs) offer a link between neural signals and external computation, enabling control of devices for the purposes of restoring function to motor-affected individuals and enhancing capabilities of a wider set of populations. Electroencephalography (EEG) offers a high temporal resolution for dynamic and potential real-time feedback for non-invasive systems. However, its practical efficacy remains limited due to low spatial resolution and poor signal-to-noise ratio, leading to insufficient decoding accuracy and unintuitive control paradigms that hinder reliable user interaction. In this study, we present a framework for an online EEG foundation model by creating a custom foundation model through spectrogram reconstruction of compact temporal windows and online constraints during pretraining. We evaluate the performance of the model in a challenging control paradigm of single-arm, directional motor imagery with dynamic movements for guided and free movement cursor control tasks. Our foundation model approach achieved a final average accuracy of 51.3% during a goal-oriented guided control task. This represents a 15.8% increase over a conventional deep learning framework and a 26.3% increase above chance level, evaluated in a cohort of 11 human participants. During the free movement task, the foundation model invoked a higher rate of completion and lower completion times. Furthermore, the custom EEG foundation model demonstrated superior adaptability from same-session finetuning and indicated an enhanced capability to assist subject learning. These findings highlight the potential of EEG foundation models to support more robust and intuitive non-invasive BCI systems, providing a promising modelling framework for future BCI development.

BackgroundEar-EEG-based brain-computer interfaces (BCIs) provide improved wearability and comfort compared to traditional scalp-EEG systems. However, their performance is constrained by low signal-to-noise ratios (SNRs) and high rates of BCI illiteracy under conventional luminance-modulated steady-state visual evoked potential (SSVEP) paradigms. MethodsThis study introduces a text-sequence stimulation paradigm to address these limitations by leveraging ventral visual pathway responses that are more accessible to electrodes near the ear. Using offline frequency-sweeping experiments across 4-8 Hz, we identified optimal stimulus parameters (4.6-6.8 Hz with 0.25{pi} phase shifts) and integrated them into a 12-target BCI system. We further conducted online experiments to compare the response characteristics and real-time spelling performance between the proposed text-sequence paradigm and conventional luminance stimulation. ResultsComparative experiments with 14 participants demonstrate that text sequence stimuli achieve an average information transfer rate (ITR) of 44.59 {+/-} 10.50 bits/min, outperforming luminance modulation by 76.18% in ITR. Notably, text sequence stimulation effectively mitigated BCI illiteracy, with all participants achieving near or above 70% accuracy (mean: 86.37 {+/-} 9.61%). This represents a significant improvement over luminance modulation, where 50% of users fell below 70% accuracy. ConclusionsBy reducing the flicker area by 14% and mimicking the natural luminance variations that occur during reading, the proposed method enhanced visual comfort. The online results further validate text-sequence stimulation as a high-performance and user-friendly paradigm for ear-EEG BCIs, supporting their practicality for assistive applications.

Machine-learning surrogate models are positioned to help optimize deep brain stimulation (DBS) usage by predicting neural activation in response to electrical stimulation, while minimizing tradeoffs between computational expense and accuracy. Previous work has developed high accuracy artificial neural network (ANN) and convolutional neural network (CNN) surrogate models that predict activation of individual, myelinated axons, to extracellular electrical stimulation for subsets of DBS programming configurations. Moreover, more traditional machine learning methods including extreme gradient boosting (XGBoost) have been used effectively for peripheral-nerve single-fiber activation predictions. We build upon the previous work and compare ANN, CNN and XGBoost methods to a much expanded set of electrode programming configurations including: monopolar, bipolar, tripolar, quadrupolar, multiple monopolar, and multiple cases of directional leads. Training used datasets generated from a finite-element model of an implanted DBS lead together with multi-compartment cable models of synthetically generated axons. We evaluated the machine learning predictors using white matter pathways derived from group-averaged connectome data within a patient-specific tissue conductivity field, comparing both predicted stimulus activation thresholds and pathway recruitment across a clinically relevant range of stimulus amplitudes and pulse widths. Our ANN and CNN models successfully predicted neural fiber activation for almost all electrode configurations with low error, expanding the scope of our previous predictor model. Results also showed key limitations of XGBoost models and superior performance of CNNs for more complex electrostatic fields of the directional leads.

ObjectiveConvolutional neural networks (CNNs) have shown promise in decoding neural drive from high-density surface electromyography (HD-sEMG) signals. However, the effects of convolutional kernel dimensionality on the generalizability and computational efficiency of CNN-based neural drive decoding remain unclear. This study systematically examined how the dimensionality of convolutional kernels (1D, 2D, and 3D) affects both the generalizability and computational efficiency of CNN-based neural drive decoding. ApproachThree CNN architectures differing only in the dimensionality of their convolutional kernels were implemented to extract temporal (1D), spatial (2D), or spatiotemporal (3D) features from HD-sEMG signals of isometric knee extension, ankle plantarflexion at three intensities. Each CNN was repeatedly trained using subsets of a pooled training dataset with varying sizes. Cross-intensity and cross-muscle generalizability were assessed by the correlation coefficient between neural drive from deep CNN and that from golden standard blind source separation (BSS) algorithms. Computational efficiency was assessed by measuring inference time on both CPU and GPU platforms. Main ResultsAll CNN architectures demonstrated generalizability across contraction intensities and muscles. For cross contraction intensities, the 1D, 2D, and 3D CNNs achieved mean correlation coefficients of 0.986 {+/-} 0.009, 0.987 {+/-} 0.010, and 0.987 {+/-} 0.010, respectively. For cross-muscle generalizability, the corresponding correlation coefficients were 0.961 {+/-} 0.051, 0.965 {+/-} 0.049, and 0.968 {+/-} 0.046. In terms of efficiency, the 3D CNN was the least computationally efficient, with inference times of 4.1 ms per sample on the CPU and 1.2 ms per sample on the GPU. SignificanceThese findings demonstrate that increased CNN architectural complexity does not necessarily yield superior generalizability in neural drive decoding from HD-sEMG signals. The results provide practical guidance for balancing decoding performance and computational efficiency in HD-sEMG-based neural-machine interfaces.

Vagus nerve stimulation (VNS) is an established clinical therapy for drug-resistant epilepsy and shows potential for treating other conditions, including depression, rheumatoid arthritis, diabetes, and heart failure. However, stimulation often produces unwanted side effects such as hoarseness, coughing, and paraesthesia. In some cases, these effects limit the delivery of therapeutic stimulation levels and hinder the development of new neuromodulation therapies. Selective VNS (sVNS) offers a strategy to reduce off-target organ activation. MethodsThis work presents an NFC-controlled, wirelessly powered, battery-free, temporary implantable multichannel stimulation device, made with off-the-shelf components, enabling selective stimulation of specific regions of the vagus nerve (VN). The encapsulated device is suitable for short-term implantation in animals. Main resultThe sVNS device was investigated in a porcine (n = 4) trial and an n = 1 pilot human experiment. Selective bradycardia of 23.28 {+/-} 12.91% was achieved in pigs and 7.5% in the human participant. In the human, a clear separation of cardiac efferent and afferent stimulation was observed, with additional selectivity in laryngeal activity. Physiological effects of laryngeal and cardiac fibre separation were measured to be 231{degrees}. SignificanceGeometrically selective stimulation of VN fascicles has the potential to improve clinical outcomes, enhance therapeutic efficacy, and reduce stimulation-related side effects. This strategy may enhance neuromodulation approaches for the treatment of heart failure using VNS.

ObjectiveComputational models and visualization toolboxes for Deep Brain Stimulation (DBS) increasingly rely on pre-computed electric field libraries to estimate the Volume of Tissue Activated (VTA). However, the boundary conditions (BCs) and source models used to generate these fields vary widely across studies, and there is currently no experimental consensus regarding which parameters most accurately reflect the physical device output. The objective of this study was to experimentally validate the electric potential distribution of directional DBS leads in order to determine the optimal Finite Element Method (FEM) configuration. ApproachThe voltage distribution surrounding a Boston Scientific Vercise Gevia directional lead was mapped in a saline phantom using a custom high-precision robotic scanning system. Experimental measurements were compared against six FEM configurations that varied in source formulation (Dirichlet vs. Neumann boundary conditions) and ground definitions. For each configuration, the resulting VTA volume was computed to assess the clinical impact of modeling assumptions. ResultsThe FEM configuration implementing a Dirichlet (voltage) boundary condition on the active contact with a grounded implantable pulse generator (IPG) surface demonstrated the highest accuracy, achieving a Symmetric Mean Absolute Percent Error (SMAPE) of less than 9% across all contact levels. In contrast, conventional current-controlled simulations employing Neumann boundary conditions with disparate ground definitions substantially overestimated electric field spread. Suboptimal boundary condition selection resulted in an approximate 67% overestimation of VTA volume (137 mm3 vs. 82 mm3) relative to the experimentally validated model. SignificanceAlthough clinical DBS systems operate as current sources, standard Neumann (current density) boundary conditions do not adequately represent the equipotential behavior of the electrode-tissue interface, resulting in nearly a two-fold error in predicted VTA volume. To improve the validity of predictive clinical models, we recommend the use of Dirichlet boundary conditions derived from the device operating impedance (V = Itarget x Zmeasured) rather than conventional current density specifications.

Transcranial temporal interference stimulation (tTIS) is a promising non-invasive brain stimulation technique that has the potential to selectively modulate deep brain regions by delivering two high-frequency alternating currents that interfere to produce a low-frequency amplitude-modulated envelope at the target. A key challenge in deploying tTIS is finding electrode current patterns that are simultaneously effective, focal, and safe. This is a fundamentally non-convex optimization problem for which a number of methods have recently been proposed. However, no systematic comparison of these methods across a large and diverse set of brain targets has been performed, leaving practitioners without clear guidance on how best to optimize for a particular experimental setting. In this work, we present a comprehensive benchmarking study comparing seven tTIS optimization methods that have appeared in the literature in recent years: exhaustive search, genetic algorithm, multi-objective evolutionary algorithm (MOVEA), min-max optimization, convex TI (CVXTI), non-convex optimization with convex relaxations, and an unsupervised neural network. All methods were evaluated across 250 brain targets spanning cortical and subcortical gray matter and white matter regions in five finite element head models. Each method was evaluated on two key metrics: mean electric field strength within the target region of interest, and off-target stimulated brain volume. Results were stratified by tissue type and target depth to identify systematic performance differences. Based on these results, we provide practical evidence-based recommendations for optimization method selection among these seven methods depending on target location, tissue type, and available computation time. Moreover we provide the code base that will allow other investigators to use these methods for their own applications. Our goal is to provide researchers and clinicians with a clear, evidence-based framework for choosing a tTIS optimization method suited to their specific target and application.

ObjectivesEpidural spinal cord stimulation (SCS) is an emerging therapy for motor rehabilitation following spinal cord injury (SCI) and other motor disorders. Conventionally, SCS leads are placed along the dorsal spinal cord (SCSD), where stimulation activates large diameter afferent fibers, which indirectly activate motoneurons through reflex pathways. This leads to broad activation of flexor and extensor muscles and limited fine-tuned control of motor output. Targeting the ventral spinal cord (SCSV) may enable more direct activation of motoneuron pools, potentially improving the specificity of muscle activation; however, there is currently no established method to place leads ventrally. To address this, we evaluated the feasibility of four modified percutaneous implantation techniques to target the ventrolateral thoracolumbar spinal cord. Materials and methodsPercutaneous SCSV implantation was performed in three human cadaver torso specimens under fluoroscopic guidance. The following approaches were evaluated: sacral hiatus, transforaminal, interlaminar contralateral, and interlaminar ipsilateral. The leads in the latter 3 approaches were inserted between L1 and L5. Eighteen implants were attempted, with nine leads retained for analysis. Lead and electrode position were assessed using computed tomography (CT) with three-dimensional reconstruction, along with anatomical dissection to verify lead and electrode placement within the epidural space. ResultsSuccessful ventral epidural lead placement was achieved using all four implantation approaches. The sacral hiatus (16/16 electrodes) and transforaminal (8/8 electrodes) approaches resulted in exclusively ventrolateral placement. The interlaminar contralateral approach led to 27/32 electrodes positioned ventrolaterally and 5/32 dorsally. The interlaminar ipsilateral implantation approach led to 14/32 electrodes positioned ventrolaterally and 18/32 positioned ventromedially. ConclusionsThese findings demonstrate that ventral epidural SCS lead placement can be achieved using modified percutaneous implant techniques. The four approaches outlined here provide a clinically feasible pathway to SCSV and establishes a foundation for future clinical studies investigating SCSV for motor rehabilitation following SCI.

ObjectivesDeep brain stimulation (DBS) is an established therapy for neurological disorders such as Parkinsons disease (PD). Modern DBS devices can record local field potentials (LFPs) to guide DBS therapy. LFPs from these devices are typically limited to bipolar configurations to suppress common-mode noise and reject artifacts. However, bipolar recordings also attenuate some local physiological signals. Methods that convert bipolar to monopolar power offer more spatially precise estimates of LFPs. Herein, we develop a model to estimate monopolar power from bipolar recordings. Materials and MethodsThis retrospective study analyzed 64 patients with PD undergoing STN (11) or GPi (53) DBS implantation. Intraoperatively, LFPs were recorded from all contacts and filtered. Bipolar montages were generated for each combination. Power spectral density (PSD) was calculated from each monopolar and bipolar signal, averaged over canonical frequency bands, and processed as log PSD. A common set of bipolar configurations was selected to minimize the Condition Number (CN), maximizing model stability. Monopolar and bipolar powers were related using robust OLS regression. Observations were randomly partitioned into training and validation sets. ResultsSixty-four PD patients yielded 640 observations. The configuration group with the lowest CN (7.45) was {C03, C12, C23}. The models demonstrated adjusted R2s of 0.9015, 0.9055, 0.8853, and 0.8764, and RMSEs (dB) of 3.2663, 3.2801, 3.5815, and 3.7035 when predicting C0, C1, C2, and C3 (N = 500; all p < 0.0001). The STN, GPi, and combined cohorts performed comparably. Weights transferred from the combined model to the validation set retained high performance. ConclusionsThis study demonstrates that monopolar LFP power can be accurately estimated from bipolar power using a linear regression model with strong generalizability across targets and validation sets. This approach offers a hardware-agnostic solution to spatially disambiguate signals and better inform DBS programming and adaptive stimulation in chronically implanted devices.

BackgroundLower-limb EEG is a rehabilitation-facing control signal for stroke neurorehabilitation and future non-invasive brain-spine interfaces, but a public external benchmark that jointly audits source construction, minimal adaptation burden, and confound sensitivity is lacking. We therefore tested whether lower-limb effort-versus-rest decoders trained on healthy public EEG transport to a stroke target domain. MethodsWe conducted a retrospective public-data external benchmark using three public EEG datasets harmonised to a common lower-limb effort-versus-rest target. Classical and deep models were compared under zero-shot transport, 10-shot temperature calibration, and 10-shot fine-tuning. For few-shot analyses, each target participant contributed a trial-disjoint subject-internal support set of 10 labelled trials per class and a held-out remainder test set. Prespecified analyses audited source construction, support-resampling sensitivity, and montage controls. Uncertainty was summarised with participant-level bootstrap confidence intervals. ResultsWithin this benchmark, healthy-to-stroke zero-shot transport was weak. The best zero-shot result was classical rather than deep, with CSP+LDA reaching area under the receiver operating characteristic curve (AUROC) 0.603, whereas EEGNet remained near chance (AUROC 0.527). Ten-shot calibration improved operating behaviour more than discrimination: for CSP+LDA, expected calibration error fell from 0.267 to 0.035 and specificity increased from 0.180 to 0.485, whereas AUROC remained essentially unchanged (0.603 to 0.604). Ten-shot fine-tuning produced only modest gains; the best overall AUROC was 0.605 for pooled dataset-balanced CSP+LDA, numerically tied with pooled raw CSP+LDA (0.605). MILimbEEG-only source training was consistently weak, exploratory deep domain-generalisation variants did not rescue transport, and frontal and temporal montage controls remained relatively competitive. ConclusionsWithin this public benchmark, source construction and minimal adaptation burden mattered more than model novelty, and retrospective montage controls limited motor-specific interpretation. The results support harmonised prospective validation of lower-limb EEG transport over further retrospective model iteration.

Kilosort 2.0 is a widely adopted spike sorting algorithm recognized for its efficiency and accuracy on planar electrode arrays, such as Neuropixels. To adapt its robust architecture to emerging three-dimensional (3D) neural probes, we present Kilosort 2.0-3D, a modified version that leverages 3D spatial information. Our modification specifically redefines the spatial processing components of Kilosort 2.0 to operate in 3D space while leaving the core template-matching process unchanged. By using synthetic extracellular recording data with ground-truth neuron positions and firing times, we demonstrate that Kilosort 2.0-3D effectively resolves spatial ambiguities and unit misclassifications inherent in 2D spatial assumptions. Our results show that Kilosort 2.0-3D achieves rotational invariance and maintains full backward compatibility with planar arrays. This work establishes a validated, scalable tool for spike sorting of high-density 3D neural electrophysiology data.

Neural decoders serve as probabilistic interfaces in co-control brain-to-text BCIs, where predicted uncertainty shapes hypothesis generation and language model integration, enabling decisions to be made safely under uncertainty. However, it remains unclear whether these decoders produce reliable and informative uncertainty, or how training objectives shape these properties. This work characterizes and improves uncertainty representations in brain-to-text decoding. We extend two metrics, calibration error (ECE) and resolution (RES), to evaluate sequential probabilistic predictions from frame-level phoneme estimates to word-level hypotheses, quantifying the reliability and informativeness of model uncertainty. Using this framework, we analyze neural decoders trained with connectionist temporal classification (CTC). To isolate the causal role of uncertainty independent of accuracy, we manipulate predicted probability distributions while holding predicted sequences fixed. Motivated by the observed failures, we further examine the role of the training objective and propose a two-stage cross-entropy (CE) formulation that decouples alignment inference from classification. We show that widely used CTC-trained neural decoders in brain-to-text BCIs produce systematically over-confident predictions, with high confidence persisting even when predictions are incorrect. Controlled manipulations of the prediction reveal that improved ECE and RES enhance hypothesis generation and language-model integration by promoting diverse alternatives and more effective re-ranking of hypotheses aligned with user intent. Mechanistically, CTC relies on over-confident predictions to resolve alignment ambiguity. Replacing CTC with CE loss yields significantly more reliable and informative probabilistic predictions without degrading decoding accuracy. Uncertainty emerges as a system-level design variable in brain-to-text interfaces. Calibrated uncertainty from neural decoders enables effective integration with independently trained language models and reliable error detection. This work reframes uncertainty from a passive output into an active control signal, identifies key components and evaluation criteria for probabilistic co-control, and outlines a pathway toward next-generation BCIs that supports increasingly complex interactions with the world.

IntroductionVagus nerve stimulation modulates laryngeal, cardiac, pulmonary, and gastrointestinal functions. Knowledge of where along the vagal trunk organ-specific branches emerge may support alternative surgical placements of stimulation devices to engage targeted functions while avoiding off-target effects. However, no quantified map of how vagal branches emerge and how they relate to surgically relevant anatomical landmarks exists in humans. MethodsFifty-eight vagus nerves (29 left, 29 right) from 29 embalmed donor bodies (15 females) were dissected from the jugular foramen through the thoracic cavity. Branches were traced to end organs and allocated to seven groups -- sympathetic, muscular, vascular, cardiac, pulmonary, esophageal, and multiple targets -- and several sub-groups. Distances between branch emergence and the jugular foramen (JF) were normalized to three anatomical landmarks: carotid bifurcation, laryngeal prominence, and superior border of clavicle. ResultsBranch emergence follows a proximal-to-distal order: sympathetic (5.28 cm from JF), muscular (9.59 cm), vascular (10.70 cm), cardiac (19.65 cm), pulmonary (25.36 cm), and esophageal (26.57 cm). Vagal branches emerge into two embryological domains separated near the clavicle: pharyngeal arch-targeting branches cluster proximally (9.34 cm) and primitive mediastinum-targeting branches cluster distally (23.74 cm), with sympathetic, muscular, and vascular sub-groups occupying distinct zones within the proximal domain. The largest branch-free intervals occur above the left clavicle (2.33 {+/-} 2.80 cm) and below the left carotid bifurcation (2.58 {+/-} 3.17 cm). Alternate placement regions separating targeted organs from off-targets: sympathetic vs. cervical visceral at 6/8 cm (L/R), cardiac vs. carotid sinus/bifurcation at 14/10 cm, and recurrent laryngeal vs. other cervical visceral at 18/13 cm from JF. Overall, no differences were found between male and female donors. ConclusionsThis study provides a quantified, landmark-registered map of cervical and thoracic vagal branch emergence, offering a standardized anatomical template that may inform strategies for more function-selective vagal neuromodulation.

BackgroundPersonalized optimization of 4x1 high-definition transcranial electrical stimulation (HD-tES) faces inherent trade-offs between montage flexibility, computational efficiency, and implementation accessibility. Conventional 10-10 electrode systems constrain placement to discrete landmark positions, while unconstrained optimization relies on stochastic algorithms that risk converging to local optima and requires neuronavigation equipment often unavailable in rehabilitation settings. Here we introduce a scalp geometry-based parameter space (SGP) that parameterizes 4x1 HD-tES montages using three intuitive scalp-defined parameters--position, radius, and orientation--and characterize parameter-performance regularities through exhaustive electric field simulations across 30 subjects and 624 cortical targets (>3.6 million configurations). ResultsPosition primarily determined proximity to optimal performance, radius governed the intensity-focality trade-off, and orientation served as fine-tuning. Exploiting these regularities, a minimal search space (SGP-MSS) was constructed that reduced computational complexity by over 90% while guaranteeing global optima identification. Compared with standard 10-10 montages, SGP-MSS achieved up to 99% higher targeting intensity and 126% higher focality (all p < 0.0001). Compared with lead-field-free optimization, SGP-MSS achieved comparable performance with greater cross-subject stability. ConclusionsThe SGP framework enables efficient individualized HD-tES optimization without neuronavigation. Its scalp-based parameterization supports electrode positioning via standard cranial landmark measurements, facilitating translation to routine clinical and home-based rehabilitation settings.

Conventional temporal interference stimulation (TI, TIS, or tTIS) leverages two pairs of electrodes to induce an interfering electrical field in the brain. Both computational and experimental studies show that TI can stimulate deep brain regions without significantly affecting shallow areas. While promising, optimization of the locations and dosages on these two pairs of electrodes for maximal focal modulation remains computationally challenging. We are the first to propose two arrays of electrodes instead of two or multiple pairs of electrodes to boost modulation focality. However, the optimization algorithm outputs too many electrodes with overlaps across two frequencies, making it difficult to implement in practice. Based on recent progress in developing multi-channel TI devices and computational work on TI optimization, here we again advocate two-array TI, but with solid software and hardware evidence to show the feasibility. Specifically, we show that the latest optimization algorithm for two-pair TI innately works for two-array TI with the fastest speed (under 30s) among all major algorithms. With a similar amount of electrodes, two-array TI could achieve better focality (3.03 cm) at the hippocampus even than TI using up to 16 pairs of electrodes (3.19 cm) that takes days to optimize. We also show a hardware implementation of two-array TI using 10 electrodes on our 8-channel TI device. We argue that two-pair TI is only preferred when one does not care about modulation focality and promote two-array TI for its advantages in focality and lower cost in terms of both optimization time and electrodes needed. We restate the focality-intensity tradeoff but in the context of TI and provide a first voxel-level map of achievable focality and modulation strength by TI in the MNI-152 head template. We hope this work will pave the way for future adoptions of two-array TI for more focal non-invasive deep brain stimulation.

BackgroundElectrophysiological recordings from chronically implanted Deep Brain Stimulation (DBS) electrodes can greatly advance understanding of disease and treatment mechanisms of motor and psychiatric disorders. The Medtronic Percept system allows for chronic recordings of local field potentials (LFP) from DBS target regions. However, these systems lack an inbuilt synchronization option to align LFP recordings to other recording modalities and consequently events in computerized tasks. ObjectiveWe propose and evaluate a synchronization method based on Transcutaneous Electrical Stimulation (TES) with low amplitudes to precisely align recorded LFP signals from the DBS electrodes to EEG recordings. MethodsThe TES-based synchronization approach was implemented and tested in 11 participants implanted with the Medtronic Percept for treatment of Parkinsons disease. ResultsThe proposed method provides high reliability, precise alignment and usability across all Medtronic Percept recording modes. Notably, the method enables recordings during adaptive DBS and with stimulation turned off. In this recording mode, LFP signals can be acquired from all recording contact pairs simultaneously, with a high signal-to-noise ratio. We provide detailed setup plans and share Python and Matlab scripts for signal alignment to enable easy application of our approach. ConclusionBy enabling reliable, well-aligned LFP recordings from all DBS contacts, our method provides a robust tool for studying neural dynamics and refining therapeutic interventions in diverse neurological conditions.

Visual prostheses face a critical miniaturisation challenge: converting photoreceptor signals to biologically appropriate retinal ganglion cell (RGC) stimulation patterns within the spatial constraints of intraocular implants. Existing systems rely on external microcontrollers for signal processing, limiting scalability for high-density pixel arrays. This paper presents an integrated per-pixel circuit architecture that directly converts photocurrent into frequency-modulated current pulses that match RGC activation thresholds. The design targets are established through NEURON computational modelling of red-green colour-opponent midget RGCs, identifying stimulation thresholds of +0.1nA to +3.5nA for depolarisation and -0.1nA for repolarisation. The proposed circuit combines a transimpedance amplifier, a voltage-controlled oscillator with a Schmitt trigger, and a current-controlled output stage to generate biphasic pulses within these thresholds. A complementary output provides lateral inhibition, reducing crosstalk between adjacent RGC stimulation sites. Photoreceptor integration is achieved using P3HT:PCBM organic photodiodes for cone-associated RGCs and phototransistors for rod-associated RGCs, validated through OghmaNano finite element simulations. The photodiode circuit produces output frequencies of 2.5Hz (dark) to 600Hz (100 W/m2), matching reported RGC response ranges. This architecture eliminates external processing requirements, enabling scalable high-density retinal prostheses design.