Journal of Neural Engineering

Objective. Effective seizure prediction can reduce patient burden, improve clinical treatment accuracy, and lower healthcare costs. However, existing deep learning-based seizure prediction methods primarily rely on single models, which have limitations in feature extraction. This study aims to develop a hybrid model that integrates the advantages of convolutional neural networks (CNNs) and Transformer to enhance seizure prediction performance. Approach. We propose FusionXNet, a hybrid model inspired by CNNs and Transformer architectures, for seizure prediction. Specifically, we design a token synthesis unit to extract local features using convolution operations and capture global electroencephalography (EEG) representations via attention mechanisms. By merging local and global features extracted from the EEG segments, FusionXNet enhances feature representations, which are subsequently fed into a classifier for final seizure prediction. Main results. We evaluate the model on the publicly available Boston Children's Hospital and the Massachusetts Institute of Technology dataset, conducting segment-based and event-based experiments in both patient-specific and cross-patient settings. In event-based patient-specific experiments, FusionXNet achieves a sensitivity of 97.602% and a false positive rate (FPR) of 0.059 h⁻¹. The results demonstrate that the proposed model effectively predicts seizures with high sensitivity and a low FPR, outperforming existing methods. Significance. The proposed FusionXNet model provides a robust and efficient approach for seizure prediction by leveraging both local and global feature extraction. The high sensitivity and low FPR indicate its potential for real-world clinical applications, improving patient management and reducing healthcare costs.

Objective. To evaluate the effectiveness of a novel depth-based vision processing (VP) method, local background enclosure (LBE), in comparison to the comprehensive VP method, Lanczos2 (L2), in suprachoroidal retinal prosthesis implant recipients during navigational tasks in laboratory and real-world settings. Approach. Four participants were acclimatized to both VP methods. Participants were asked to detect and navigate past five of eight possible obstacles in a white corridor across 20–30 trials. Randomized obstacles included black or white mannequins, black or white overhanging boxes, black or white bins and black or white stationary boxes. The same four participants underwent trials at three different real-word urban locations using both VP methods (randomized order). They were tasked with navigating a complex, dynamic pre-determined scene while detecting, verbally identifying, and avoiding obstacles in their path. Main results. The indoor obstacle course showed that the LBE method (63.6 ± 10.7%, mean ± SD) performed significantly better than L2 (48.5 ± 11.2%), for detection of obstacles (p < 0.001, Mack–Skillings). The real-world assessment showed that of the objects detected, 50.2% (138/275) were correctly identified using LBE and 41.7% (138/331) using L2, corresponding to a risk difference of 8 percentage points, p = 0.081). Significance. Real world outcomes can be improved using an enhanced VP algorithm, providing depth-based visual cues (#NCT05158049).

Objective. A novel method is introduced to regress over the sign language finger movements from human electrocorticography (ECoG) recordings. Approach. The proposed graph-optimized block-term tensor regression (Go-BTTR) method consists of two components: a deflation-based regression model that sequentially Tucker-decomposes multiway ECoG data into a series of blocks, and a causal graph process (CGP) that accounts for the complex relationship between finger movements when expressing sign language gestures. Prior to each regression block, CGP is applied to decide which fingers should be kept separate or grouped and should therefore be referred to BTTR or its extended version eBTTR, respectively. Main results. Two ECoG datasets were used, one recorded in five patients expressing four hand gestures of the American sign language alphabet, and another in two patients expressing all gestures of the Flemish sign language alphabet. As Go-BTTR combines fingers in a flexible way, it can better account for the nonlinear relationship ECoG exhibits when expressing hand gestures, including unintentional finger co-activations. This is reflected by the superior joint finger trajectory predictions compared to eBTTR, and predictions that are on par with BTTR in single-finger scenarios. For the American sign language alphabet (Utrecht dataset), the average correlation across all fingers for all subjects was 0.73 for Go-BTTR, 0.719 for eBTTR and 0.70 for BTTR. For the Flemish sign language alphabet (Leuven dataset), the average correlation across all fingers for all subjects was 0.37 for Go-BTTR, 0.34 for eBTTR and 0.33 for BTTR. Significance. Our findings show that Go-BTTR is capable of decoding complex hand gestures taken from the sign language alphabet. Go-BTTR also demonstrates computational efficiency, providing a notable benefit when intracranial electrodes are inserted during a patient's pre-surgical evaluation. This efficiency helps reduce the time required for developing and testing a brain–computer interface solution.

Objective. With the recent development of visual evoked potential (VEP) based brain–computer interfaces (BCIs), the stimulus paradigm has been continuously innovated, in which the pursuit of higher BCI performance and better user experience has become indispensable. Approach. To optimize the stimulus paradigm, a 12-target online BCI system was designed in this study by adopting flicker for steady-state VEPs, Newton's ring for steady-state motion VEP, and frame rate based video stimulus, respectively. The signal characteristics of VEP, classification accuracy, and user experience of the three stimulus paradigms were quantitatively evaluated and compared. Main results. The online information transfer rates for the three stimulus paradigms were 53.77 bits min⁻¹, 51.41 ± 3.55 bits min⁻¹, and 52.07 ± 3.09 bits min⁻¹, respectively. The video stimulus had a significantly better user experience, while the flicker stimulus showed the worst. Significance. These results demonstrate the advantage of the proposed video stimulus paradigm and have significant theoretical and applied implications for developing VEP-based BCI systems.

Objective. Visual prostheses can provide partial visual function in patients with retinal degenerative diseases. However, in clinical trials, patients implanted with retinal prostheses have reported perceptual fading, which is thought to be related to response desensitization. Additionally, natural stimuli consist of aperiodic events across a short temporal span, whereas periodic stimulation (fixed inter-pulse intervals (IPIs)) is the standard approach in retinal prosthesis research. In this study, we investigated how aperiodic stimulation of the epiretinal surface affects electrically evoked responses in the primary visual cortex (V1) compared with periodic stimulation. Approach. In vivo experiments were conducted in healthy and retinal-degenerated rats. Periodic stimulation consisted of constant IPIs, whereas aperiodic stimulation was provided by mixed IPIs. We calculated the spike time tiling coefficient to assess response consistency across trials, the significant response ratio, and the spike rate to analyze response desensitization. Main results. The results showed a significantly lower consistency of cortical responses in retinal degenerated rats than in healthy rats at 5 Hz. The consistency of the response to periodic stimulation decreased considerably as the frequency was increased to 10 Hz and 20 Hz in both groups and was greatly improved by applying aperiodic stimulation. In addition, aperiodic stimulation evoked a significantly higher spike rate in response to continuous stimulation at high frequencies (e.g. 10 and 20 Hz). Significance. By applying electrical stimulation with varying IPIs directly on the epiretinal surface, we observed promising results in terms of enhancing cortical response consistency and reducing desensitization. This finding presents a potential approach to enhance the effectiveness of retinal prostheses.

Transcranial magnetic resonance-guided focused ultrasound (MRgFUS) represents a transformative modality in treating neurological disorders and diseases, offering precise, minimally invasive interventions for conditions such as essential tremor and Parkinson's disease. Objective. This paper presents an industry-focused perspective on the current state of MRgFUS, highlighting recent advancements, challenges, and emerging opportunities within the field. Approach. We review key clinical applications and therapeutic mechanisms, focusing on targeted ablation, while discussing technological innovations that support new indications. Current regulatory frameworks, challenges in device development, and market trends are examined to provide an understanding of the industry landscape. Main results. We indicate some limitations in MRgFUS and suggest potential strategies for overcoming these limitations to optimize treatment outcomes. Significance. We conclude with an outlook on promising developments, including artificial intelligence-enhanced targeting, low and high-field magnetic resonance imaging integration, and multimodal imaging techniques, that could potentially drive further innovation and adoption of MRgFUS in brain therapy.

Objective. Machine learning's (MLs) ability to capture intricate patterns makes it vital in neural engineering research. With its increasing use, ensuring the validity and reproducibility of ML methods is critical. Unfortunately, this has not always been the case in practice, as there have been recent retractions across various scientific fields due to the misuse of ML methods and validation procedures. To address these concerns, we propose the first version of the neural engineering reproducibility and validity essentials for ML (NERVE-ML) checklist, a framework designed to promote the transparent, reproducible, and valid application of ML in neural engineering. Approach. We highlight some of the unique challenges of model validation in neural engineering, including the difficulties from limited subject numbers, repeated or non-independent samples, and high subject heterogeneity. Through detailed case studies, we demonstrate how different validation approaches can lead to divergent scientific conclusions, highlighting the importance of selecting appropriate procedures guided by the NERVE-ML checklist. Effectively addressing these challenges and properly scoping scientific conclusions will ensure that ML contributes to, rather than hinders, progress in neural engineering. Main results. Our case studies demonstrate that improper validation approaches can result in flawed studies or overclaimed scientific conclusions, complicating the scientific discourse. The NERVE-ML checklist effectively addresses these concerns by providing guidelines to ensure that ML approaches in neural engineering are reproducible and lead to valid scientific conclusions. Significance. By effectively addressing these challenges and properly scoping scientific conclusions guided by the NERVE-ML checklist, we aim to help pave the way for a future where ML reliably enhances the quality and impact of neural engineering research.

Objective. Sharing computational models offers many benefits, including increased scientific rigor during project execution, readership of the associated paper, resource usage efficiency, replicability, and reusability. In recognition of the growing practice and requirement of sharing models, code, and data, herein, we provide guidance to facilitate sharing of computational models by providing an accessible resource for regular reference throughout a project's stages. Approach. We synthesized literature on good practices in scientific computing and on code and data sharing with our experience in developing, sharing, and using models of neural stimulation, although the guidance will also apply well to most other types of computational models. Main results. We first describe the '6 R' characteristics of shared models, leaning on prior scientific computing literature, which enforce accountability and enable advancement: re-runnability, repeatability, replicability, reproducibility, reusability, and readability. We then summarize action items associated with good practices in scientific computing, including selection of computational tools during project planning, code and documentation design during development, and user instructions for deployment. We provide a detailed checklist of the contents of shared models and associated materials, including the model itself, code for reproducing published figures, documentation, and supporting datasets. We describe code, model, and data repositories, including a list of characteristics to consider when selecting a platform for sharing. We describe intellectual property (IP) considerations to balance permissive, open-source licenses versus software patents and bespoke licenses that govern and incentivize commercialization. Finally, we exemplify these practices with our ASCENT pipeline for modeling peripheral nerve stimulation. Significance. We hope that this paper will serve as an important and actionable reference for scientists who develop models—from project planning through publication—as well as for model users, institutions, IP experts, journals, funding sources, and repository platform developers.

The Tenth International brain–computer interface (BCI) meeting was held June 6–9, 2023, in the Sonian Forest in Brussels, Belgium. At that meeting, 21 master classes, organized by the BCI Society's Postdoc & Student Committee, supported the Society's goal of fostering learning opportunities and meaningful interactions for trainees in BCI-related fields. Master classes provide an informal environment where senior researchers can give constructive feedback to the trainee on their chosen and specific pursuit. The topics of the master classes span the whole gamut of BCI research and techniques. These include data acquisition, neural decoding and analysis, invasive and noninvasive stimulation, and ethical and transitional considerations. Additionally, master classes spotlight innovations in BCI research. Herein, we discuss what was presented within the master classes by highlighting each trainee and expert researcher, providing relevant background information and results from each presentation, and summarizing discussion and references for further study.

The brain is a highly complex physical system made of assemblies of neurons that work together to accomplish elaborate tasks such as motor control, memory and perception. How these parts work together has been studied for decades by neuroscientists using neuroimaging, psychological manipulations, and neurostimulation. Neurostimulation has gained particular interest, given the possibility to perturb the brain and elicit a specific response. This response depends on different parameters such as the intensity, the location and the timing of the stimulation. However, most of the studies performed so far used previously established protocols without considering the ongoing brain activity and, thus, without adaptively targeting the stimulation. In control theory, this approach is called open-loop control, and it is always paired with a different form of control called closed-loop, in which the current activity of the brain is used to establish the next stimulation. Recently, neuroscientists are beginning to shift from classical fixed neuromodulation studies to closed-loop experiments. This new approach allows the control of brain activity based on responses to stimulation and thus to personalize individual treatment in clinical conditions. Here, we review this new approach by introducing control theory and focusing on how these aspects are applied in brain studies. We also present the different stimulation techniques and the control approaches used to steer the brain. Finally, we explore how the closed-loop framework will revolutionize the way the human brain can be studied, including a discussion on open questions and an outlook on future advances.

Objective. Effective smoothing of electroencephalogram (EEG) signals while maintaining the original signal's features is important in EEG signal analysis and brain-computer interface (BCI). This paper proposes a novel EEG signal-smoothing algorithm and its potential application in cognitive conflict processing. Approach. Instead of being processed in the time domain, the input signal is visualized in increasing line width, the representation frame of which is converted into a binary image. An effective thinning algorithm is employed to obtain a unit-width skeleton as the smoothed signal. Main results. Experimental results on data fitting have verified the effectiveness of the proposed approach on different levels of signal-to-noise (SNR) ratio, especially on high noise levels (SNR ≤ 5 dB), where our fitting error is only 86.4%-90.4% compared to that of its best counterpart. The potential application of the proposed algorithm in EEG-based cognitive conflict processing is comprehensively evaluated in a classification and a visual inspection task. The employment of the proposed approach in pre-processing the input data has significantly boosted the F1 score of state-of-the-art models by more than 1%. The robustness of our algorithm is also evaluated via a visual inspection task, where specific cognitive conflict peaks, i.e. the prediction error negativity (PEN) and error-related positive potential (Pe), can be easily observed at multiple line-width levels, while the insignificant ones are eliminated. Significance. These results demonstrated not only the advance of the proposed approach but also its impact on classification accuracy enhancement.

Objective: Motor-evoked potentials (MEPs) in response to brain stimulation, such as transcranial magnetic stimulation (TMS), allow quantification of corticospinal excitability and have served in the design of almost all available neuromodulatory interventions. So-called thresholding of MEPs at a point not too far above the noise floor establishes the reference point for dosage and safety. Despite the fundamental importance of distinguishing true MEPs from background noise, the statistical properties of the noise floor are hardly known or characterised. Furthermore, detecting pre-activation of the motor system by endogenous signals before a stimulus---which substantially distorts the subsequent stimulation response---practically involves distinguishing spontaneous activity from the background noise. However, current methods for this detection are largely ad hoc. This study aims to determine the probability distribution of the noise floor.&#xD;&#xD;Approach: We tested four probability distribution models (log-normal, gamma, generalized extreme value (GEV), and normal) in experimental data from 19 healthy subjects. Additionally, we employed a mixture model of Gaussian and Laplacian distributions to simulate background electromyography signals and tested these models on the resulting distributions.&#xD;&#xD;Main results: The distribution of the background noise floor was highly skewed, which contradicted the common assumption of normality or even log-normality. The GEV distribution model consistently outperformed other models in describing both experimental and simulated data. The gamma distribution model performed similarly to the GEV model in simulations but emerged as the second-best option with experimental data.&#xD;&#xD;Significance: The GEV and gamma distribution models enable more accurate characterisation of the background noise floor. Improvements by using these models could enhance the precision and reliability of MEP detection criteria, pre-activation identification, and variability analysis in research and motor thresholding in clinical TMS for a more accurate interpretation of neurophysiological data.

Objective: Reversible block of peripheral nerve conduction using kilohertz-frequency (KHF) electrical signals has substantial potential for treating diseases. However, onset response, i.e., KHF-induced excitation en route to producing nerve block, is an undesired outcome of neural block protocols. Previous studies of KHF nerve block observed increased onset responses when KHF signal amplitude was linearly ramped for up to 60 s at frequencies up to 30 kHz. Here, we evaluated the onset response across a broad range of ramp durations and frequencies.&#xD;Approach: In experiments on the rat tibial nerve and biophysical axon models, we quantified nerve responses to linearly ramped KHF signals applied for durations from 16 to 512 s and at frequencies from 10 to 83.3 kHz. We also investigated the role of slow inactivation on onset response during linear ramps by using lacosamide to enhance slow inactivation pharmacologically and by introducing a slow inactivation gating variable in computational models.&#xD;Main results: In experiments, sufficiently high frequencies (≥20.8 kHz) with amplitudes that were ramped sufficiently slowly (4.4 to 570 µA/s) generated conduction block without onset response, and increasing frequency enabled shorter ramps to block without onset response. Experimental use of lacosamide to enhance slow inactivation also eliminated onset response. In computational models, the effects of ramp duration/ramp rate on onset response only occurred after introducing a slow inactivation gating variable, and the models did not account for frequency effects.&#xD;Significance: The results reveal, for the first time, the ability to use charge-balanced linearly ramped KHF signals to block without onset response. This novel approach enhances the precision of neural blocking protocols and enables coordinated neural control to restore organ function, such as in urinary control after spinal cord injury.&#xD;

Objective.&#xD;Decoding neural activity from ventral (speech) motor cortex is known to enable high-performance speech brain-computer interface (BCI) control. It was previously unknown whether this brain area could also enable computer control via neural cursor and click, as is typically associated with dorsal (arm and hand) motor cortex.&#xD;Approach.&#xD;We recruited a clinical trial participant with ALS and implanted intracortical microelectrode arrays in ventral precentral gyrus (vPCG), which the participant used to operate a speech BCI in a prior study. We developed a cursor BCI driven by the participant's vPCG neural activity, and evaluated performance on a series of target selection tasks.&#xD;Main results.&#xD;The reported vPCG cursor BCI enabled rapidly-calibrating (40 seconds), accurate (2.90 bits per second) cursor control and click. The participant also used the BCI to control his own personal computer independently.&#xD;Significance.&#xD;These results suggest that placing electrodes in vPCG to optimize for speech decoding may also be a viable strategy for building a multi-modal BCI which enables both speech-based communication and computer control via cursor and click.&#xD;(BrainGate2 ClinicalTrials.gov ID NCT00912041)

Objective: Understanding the neural mechanisms underlying consciousness during anesthesia is critical for advancing anesthesiology and neuroscience. However, given the high variability in individual sensitivity to anesthetic agents, accurately elucidating the relationship between individual characteristics and drug responses is also crucial for ensuring clinical anesthesia safety.&#xD;Approach: This study utilized high-density EEG data from 20 participants under various propofol-induced sedation states. We stratified participants into low- and high-sensitivity cohorts based on their behavioral responsiveness to standardized auditory stimuli during sedation. Then the metrics such as permutation entropy (PE), phase-lag entropy (PLE), and permutation cross mutual information (PCMI) were analyzed to evaluate neural complexity, the diversity of connectivity, and information integration. Machine learning models, including support vector machines (SVM), were applied to classify individual sensitivity to propofol, with SHapley Additive exPlanations (SHAP) analysis providing feature interpretability.&#xD;Main results: Subjects were divided into high-performance (low-sensitivity) group and low-performance (high-sensitivity) group based on the accuracy of their responses to auditory stimuli. In the moderate sedation, the high-performance group exhibited elevated PE, increased PLE in alpha band and the decreased PLE in beta band, and decreased PCMI in alpha band. In the resting-state, we extracted 18 metrics that were significantly different between the two groups. Using these resting-state metrics as features, the SVM model achieved an accuracy of 87.5%±0.06% in classifying individuals into high- or low-sensitivity groups. SHAP analysis results indicated that the features, including the PLE value of temporal in alpha band (α-PLET) and the PCMI value of frontal-parietal in beta band (β-PCMIFP), were identified as robust predictors of propofol sensitivity, with high weights across various models.&#xD;Significance: This study highlights the differential neural dynamics induced by propofol across performance groups. This study highlights that resting-state metrics can predict individual sensitivity to propofol. Our findings provide preliminary insights into the potential utility of pre-anesthesia brain state assessments in predicting individual propofol sensitivity, which may contribute to the development of more precise personalized anesthesia plans.&#xD;

Objective: Motor-evoked potentials (MEPs) in response to brain stimulation, such as transcranial magnetic stimulation (TMS), allow quantification of corticospinal excitability and have served in the design of almost all available neuromodulatory interventions. So-called thresholding of MEPs at a point not too far above the noise floor establishes the reference point for dosage and safety. Despite the fundamental importance of distinguishing true MEPs from background noise, the statistical properties of the noise floor are hardly known or characterised. Furthermore, detecting pre-activation of the motor system by endogenous signals before a stimulus---which substantially distorts the subsequent stimulation response---practically involves distinguishing spontaneous activity from the background noise. However, current methods for this detection are largely ad hoc. This study aims to determine the probability distribution of the noise floor.&#xD;&#xD;Approach: We tested four probability distribution models (log-normal, gamma, generalized extreme value (GEV), and normal) in experimental data from 19 healthy subjects. Additionally, we employed a mixture model of Gaussian and Laplacian distributions to simulate background electromyography signals and tested these models on the resulting distributions.&#xD;&#xD;Main results: The distribution of the background noise floor was highly skewed, which contradicted the common assumption of normality or even log-normality. The GEV distribution model consistently outperformed other models in describing both experimental and simulated data. The gamma distribution model performed similarly to the GEV model in simulations but emerged as the second-best option with experimental data.&#xD;&#xD;Significance: The GEV and gamma distribution models enable more accurate characterisation of the background noise floor. Improvements by using these models could enhance the precision and reliability of MEP detection criteria, pre-activation identification, and variability analysis in research and motor thresholding in clinical TMS for a more accurate interpretation of neurophysiological data.

Objective.&#xD;Decoding neural activity from ventral (speech) motor cortex is known to enable high-performance speech brain-computer interface (BCI) control. It was previously unknown whether this brain area could also enable computer control via neural cursor and click, as is typically associated with dorsal (arm and hand) motor cortex.&#xD;Approach.&#xD;We recruited a clinical trial participant with ALS and implanted intracortical microelectrode arrays in ventral precentral gyrus (vPCG), which the participant used to operate a speech BCI in a prior study. We developed a cursor BCI driven by the participant's vPCG neural activity, and evaluated performance on a series of target selection tasks.&#xD;Main results.&#xD;The reported vPCG cursor BCI enabled rapidly-calibrating (40 seconds), accurate (2.90 bits per second) cursor control and click. The participant also used the BCI to control his own personal computer independently.&#xD;Significance.&#xD;These results suggest that placing electrodes in vPCG to optimize for speech decoding may also be a viable strategy for building a multi-modal BCI which enables both speech-based communication and computer control via cursor and click.&#xD;(BrainGate2 ClinicalTrials.gov ID NCT00912041)

Objective: In this study, we introduce an end-to-end single microphone deep learning system for source separation and auditory attention decoding (AAD) in a competing speech and music setup. Deep source separation is applied directly on the envelope of the observed mixed audio signal. The resulting separated envelopes are compared to the envelope obtained from the EEG signals via deep stimulus reconstruction, where Pearson correlation is used as a loss function for training and evaluation.&#xD;Approach: Deep learning models for source envelope separation and AAD are trained on target/distractor pairs from speech and music, covering four cases: speech vs. speech, speech vs. music, music vs. speech, and music vs. music. We convolve 10 different HRTFs with our audio signals to simulate the effects of head, torso and outer ear, and evaluate our model's ability to generalize. The models are trained (and evaluated) on 20 second time windows extracted from 60 second electroencephalography (EEG) trials.&#xD;Main results: We achieve a target Pearson correlation and accuracy of 0.122 and 82.4\% on the original dataset and an average target Pearson correlation and accuracy of 0.106 and 75.4\% across the 10 HRTF variants. For the distractor, we achieve an average Pearson correlation of 0.004. Additionally, our model gives an accuracy of 82.8\%, 85.8\%, 79.7\% and 81.5\% across the four aforementioned cases for speech and music. With perfectly separated envelopes, we can achieve an accuracy of 83.0\%, which is comparable to the case of source separated envelopes.&#xD;Conclusion: We conclude that the deep learning models for source envelope separation and AAD generalize well across the set of speech and music signals and HRTFs tested in this study. We notice that source separation performs worse for a mixed music and speech signal, but the resulting AAD performance is not impacted.

Objective: Extracting universal, task-independent semantic features from electroencephalography (EEG) signals remains an open challenge. Traditional approaches are often task-specific, limiting their generalization across different EEG paradigms. This study aims to develop a robust, unsupervised framework for learning high-level, task-independent neural representations.&#xD;Approach: We propose a novel framework integrating convolutional neural networks (CNNs), AutoEncoders, and Transformers to extract both low-level spatiotemporal patterns and high-level semantic features from EEG signals. The model is trained in an unsupervised manner to ensure adaptability across diverse EEG paradigms, including motor imagery (MI), steady-state visually evoked potentials&#xD;(SSVEP), and event-related potentials (ERP, specifically P300). Extensive analyses, including clustering, correlation, and ablation studies, are conducted to validate the quality and interpretability of the extracted features.&#xD;Main Results: Our method achieves state-of-the-art performance, with average classification accuracies of 83.50% and 84.84% on MI datasets (BCICIV 2a and BCICIV 2b), 98.41% and 99.66% on SSVEP datasets (Lee2019-SSVEP and Nakanishi2015), and an average AUC of 91.80% across eight ERP datasets. t-SNE and clustering analyses reveal that the extracted features exhibit enhanced separability and structure compared to raw EEG data. Correlation studies confirm the framework's ability to balance universal and subject-specific features, while ablation results highlight the near-optimality of the&#xD;selected model configuration.&#xD;Significance: This work establishes a universal framework for task-independent semantic feature extraction from EEG signals, bridging the gap between conventional feature engineering and modern deep learning (DL) methods. By providing robust, generalizable representations across diverse EEG paradigms, this approach lays the foundation for advanced brain-computer interface (BCI) appli-&#xD;cations, cross-task EEG analysis, and future developments in semantic EEG processing.

Objective&#xD;Our objective was to perform a complete analysis of in-vitro impedance data for sputtered iridium oxide film (SIROF) micro-electrodes. The analysis included quantification of the stochastic and bias error structure and development of a process model that accounted for the chemistry and physics of the electrode-electrolyte interface.&#xD;Approach&#xD;The measurement model program was used to analyze electrochemical impedance spectroscopy (EIS) data for sputtered iridium oxide film (SIROF) micro-electrodes at potentials ranging from −0.4 to +0.6 V(Ag|AgCl). The frequency range used for the analysis was that determined to be consistent with the Kramers–Kronig relations. Interpretation of the data was enabled by truncating frequencies at which the ohmic impedance influenced the impedance.&#xD;Main Results&#xD;An interpretation model was developed that considered the impedance of the bare surface and the contribution of a porous component, based on the de Levie model of porous electrodes. The influence of iridium oxidation state on impedance was included. The proposed model fit all 36 EIS spectra well. The effective capacitance of the SIROF system ranged from 32 mF/cm2 at −0.4 V(Ag|AgCl) to a maximum of 93 mF/cm2 at 0.2 and 0.4 V(Ag|AgCl).&#xD;Significance&#xD;The model developed to interpret the impedance response of neural stimulation electrodes in vitro guides model development for in-vivo studies.

Objective. Cochlear implants are among the few clinical interventions for people with severe or profound hearing loss. However, current spread during monopolar electrical stimulation results in poor spectral resolution, prompting the exploration of optical stimulation as an alternative approach. Enabled by introducing light-sensitive ion channels into auditory neurons (optogenetics), optical stimulation has been shown to activate a more discrete neural area with minimal overlap between each frequency channel during simultaneous stimulation. However, the utility of optogenetic approaches is uncertain due to the low fidelity of responses to light and high-power requirements compared to electrical stimulation. Approach. Hybrid stimulation, combining sub-threshold electrical and optical pulses, has been shown to improve fidelity and use less light, but the impact on spread of activation and channel summation using a translatable, multi-channel hybrid implant is unknown. This study examined these factors during single channel and simultaneous multi-channel hybrid stimulation in transgenic mice expressing the ChR2/H134R opsin. Acutely deafened mice were implanted with a hybrid cochlear array containing alternating light emitting diodes and platinum electrode rings. Spiking activity in the inferior colliculus was recorded during electrical-only or hybrid stimulation in which optical and electrical stimuli were both at sub-threshold intensities. Thresholds, spread of activation, and threshold shifts during simultaneous hybrid stimulation were compared to electrical-only stimulation. Main results. The electrical current required to reach activation threshold during hybrid stimulation was reduced by 7.3 dB compared to electrical-only stimulation (p<0.001). The activation width measured at two levels of discrimination above threshold and channel summation during simultaneous hybrid stimulation were significantly lower compared to electrical-only stimulation (p<0.05), but there was no spatial advantage of hybrid stimulation at higher electrical stimulation levels. Significance. Reduced channel interaction would facilitate multi-channel simultaneous stimulation, thereby enhancing the perception of temporal fine structure which is crucial for music and speech in noise.

Objective. Visual prostheses can provide partial visual function in patients with retinal degenerative diseases. However, in clinical trials, patients implanted with retinal prostheses have reported perceptual fading, which is thought to be related to response desensitization. Additionally, natural stimuli consist of aperiodic events across a short temporal span, whereas periodic stimulation (fixed inter-pulse intervals (IPIs)) is the standard approach in retinal prosthesis research. In this study, we investigated how aperiodic stimulation of the epiretinal surface affects electrically evoked responses in the primary visual cortex (V1) compared with periodic stimulation. Approach. In vivo experiments were conducted in healthy and retinal-degenerated rats. Periodic stimulation consisted of constant IPIs, whereas aperiodic stimulation was provided by mixed IPIs. We calculated the spike time tiling coefficient to assess response consistency across trials, the significant response ratio, and the spike rate to analyze response desensitization. Main results. The results showed a significantly lower consistency of cortical responses in retinal degenerated rats than in healthy rats at 5 Hz. The consistency of the response to periodic stimulation decreased considerably as the frequency was increased to 10 Hz and 20 Hz in both groups and was greatly improved by applying aperiodic stimulation. In addition, aperiodic stimulation evoked a significantly higher spike rate in response to continuous stimulation at high frequencies (e.g. 10 and 20 Hz). Significance. By applying electrical stimulation with varying IPIs directly on the epiretinal surface, we observed promising results in terms of enhancing cortical response consistency and reducing desensitization. This finding presents a potential approach to enhance the effectiveness of retinal prostheses.

Objective. Ipsilateral motor evoked potentials (iMEPs) are believed to represent cortically evoked excitability of uncrossed brainstem-mediated pathways. In the event of extensive injury to (crossed) corticospinal pathways, which can occur following a stroke, uncrossed ipsilateral pathways may serve as an alternate resource to support the recovery of the paretic limb. However, iMEPs, even in neurally intact people, can be small, infrequent, and noisy, so discerning them in stroke survivors is very challenging. This study aimed to investigate the inter-rater reliability of iMEP features (presence/absence, amplitude, area, onset, and offset) to evaluate the reliability of existing methods for objectively analyzing iMEPs in stroke survivors with chronic upper extremity (UE) motor impairment. Approach. Two investigators subjectively measured iMEP features from thirty-two stroke participants with chronic UE motor impairment. Six objective methods based on standard deviation (SD) and mean consecutive differences (MCD) were used to measure the iMEP features from the same 32 participants. IMEP analysis used both trial-by-trial (individual signal) and average-signal analysis approaches. Inter-rater reliability of iMEP features and agreement between the subjective and objective methods were analyzed (percent agreement-PA and intraclass correlation coefficient-ICC). Main results. Inter-rater reliability was excellent for iMEP detection (PA > 85%), amplitude, and area (ICC > 0.9). Of the six objective methods we tested, the 1SD method was most appropriate for identifying and analyzing iMEP amplitude and area (ICC > 0.9) in both trial-by-trial and average signal analysis approaches. None of the objective methods were reliable for analyzing iMEP onset and offset. Results also support using the average-signal analysis approach over the trial-by-trial analysis approach, as it offers excellent reliability for iMEP analysis in stroke survivors with chronic UE motor impairment. Significance. Findings from our study have relevance for understanding the role of ipsilateral pathways that typically survive unilateral severe white matter injury in people with stroke.

Objective. Non-invasive brain stimulation (NIBS) offers therapeutic benefits for various brain disorders. Personalization may enhance these benefits by optimizing stimulation parameters for individual subjects. Approach. We present a computational pipeline for simulating and assessing the effects of NIBS using personalized, large-scale brain network activity models. Using structural MRI and diffusion-weighted imaging data, the pipeline leverages a convolutional neural network-based segmentation algorithm to generate subject-specific head models with up to 40 tissue types and personalized dielectric properties. We integrate electromagnetic simulations of NIBS exposure with whole-brain network models to predict NIBS-dependent perturbations in brain dynamics, simulate the resulting EEG traces, and quantify metrics of brain dynamics. Main results. The pipeline is implemented on o²S²PARC, an open, cloud-based infrastructure designed for collaborative and reproducible computational life science. Furthermore, a dedicated planning tool provides guidance for optimizing electrode placements for transcranial temporal interference stimulation. In two proof-of-concept applications, we demonstrate that: (i) transcranial alternating current stimulation produces expected shifts in the EEG spectral response, and (ii) simulated baseline network activity exhibits physiologically plausible fluctuations in inter-hemispheric synchronization. Significance. This pipeline facilitates a shift from exposure-based to response-driven optimization of NIBS, supporting new stimulation paradigms that steer brain dynamics towards desired activity patterns in a controlled manner.

Objective: The study of neurovascular coupling (NVC), the relationship between neuronal activity and cerebral blood flow, is essential for understanding brain physiology in both healthy and pathological states. Current methods to study NVC include neuroimaging techniques with limited temporal resolution and indirect neuronal activity measures. Methods including electroencephalographic (EEG) data are predominantly linear and display limitations that nonlinear methods address. Transfer Entropy (TE) explores linear and nonlinear relationships simultaneously. This study hypothesizes that complex NVC interactions in stroke patients, both linear and&#xD;nonlinear, can be detected using TE. Approach: TE between simultaneously recorded EEG and cerebral blood flow velocity (CBFV) signals was computed and analyzed in three settings: ipsilateral (EEG and CBFV from same hemisphere) stroke and nonstroke, and contralateral (EEG from stroke hemisphere, CBFV from nonstroke hemisphere). A surrogate analysis was performed to evaluate the significance of TE values and to identify the nature of the interactions. Main results: The results showed that EEG generally influenced CBFV. There were more linear+nonlinear interactions in the ipsilateral nonstroke setting and in the delta band in ipsilateral&#xD;stroke and contralateral settings. Interactions between EEG and CBFV were stronger on the nonstroke side for linear+nonlinear dynamics. The strength and nature of the interactions were weakly correlated with clinical outcomes (e.g., delta band (p<0.05): infarct growth linear = -0.448, linear+nonlinear = -0.339; NHISS linear = -0.473, linear+nonlinear = -0.457). Significance: This study exemplifies the benefits of using TE in linear and nonlinear NVC analysis to better understand the implications of these dynamics in stroke severity.