Table of contents

Objective. Electrical stimulation can induce sensation in the phantom limb of individuals with amputation. It is difficult to generalize existing findings as there are many approaches to delivering stimulation and to assessing the characteristics and benefits of sensation. Therefore, the goal of this systematic review was to explore the stimulation parameters that effectively elicited referred sensation, the qualities of elicited sensation, and how the utility of referred sensation was assessed. Approach. We searched PubMed, Web of Science, and Engineering Village through January of 2022 to identify relevant papers. We included papers which electrically induced referred sensation in individuals with limb loss and excluded papers that did not contain stimulation parameters or outcome measures pertaining to stimulation. We extracted information on participant demographics, stimulation approaches, and participant outcomes. Main results. After applying exclusion criteria, 49 papers were included covering nine stimulation methods. Amplitude was the most commonly adjusted parameter (n = 25), followed by frequency (n = 22), and pulse width (n = 15). Of the 63 reports of sensation quality, most reported feelings of pressure (n = 52), paresthesia (n = 48), or vibration (n = 40) while less than half (n = 29) reported a sense of position or movement. Most papers evaluated the functional benefits of sensation (n = 33) using force matching or object identification tasks, while fewer papers quantified subjective measures (n = 16) such as pain or embodiment. Only 15 studies (36%) observed percept intensity, quality, or location over multiple sessions. Significance. Most studies that measured functional performance demonstrated some benefit to providing participants with sensory feedback. However, few studies could experimentally manipulate sensation location or quality. Direct comparisons between studies were limited by variability in methodologies and outcome measures. As such, we offer recommendations to aid in more standardized reporting for future research.

Automated interictal epileptiform discharge (IED) detection has been widely studied, with machine learning methods at the forefront in recent years. As computational resources become more accessible, researchers have applied deep learning (DL) to IED detection with promising results. This systematic review aims to provide an overview of the current DL approaches to automated IED detection from scalp electroencephalography (EEG) and establish recommendations for the clinical research community. We conduct a systematic review according to the PRISMA guidelines. We searched for studies published between 2012 and 2022 implementing DL for automating IED detection from scalp EEG in major medical and engineering databases. We highlight trends and formulate recommendations for the research community by analyzing various aspects: data properties, preprocessing methods, DL architectures, evaluation metrics and results, and reproducibility. The search yielded 66 studies, and 23 met our inclusion criteria. There were two main DL networks, convolutional neural networks in 14 studies and long short-term memory networks in three studies. A hybrid approach combining a hidden Markov model with an autoencoder was employed in one study. Graph convolutional network was seen in one study, which considered a montage as a graph. All DL models involved supervised learning. The median number of layers was 9 (IQR: 5–21). The median number of IEDs was 11 631 (IQR: 2663–16 402). Only six studies acquired data from multiple clinical centers. AUC was the most reported metric (median: 0.94; IQR: 0.94–0.96). The application of DL to IED detection is still limited and lacks standardization in data collection, multi-center testing, and reporting of clinically relevant metrics (i.e. F1, AUCPR, and false-positive/minute). However, the performance is promising, suggesting that DL might be a helpful approach. Further testing on multiple datasets from different clinical centers is required to confirm the generalizability of these methods.

Objective. Vagus nerve stimulation (VNS) is Food and Drug Administration-approved for epilepsy, depression, and obesity, and stroke rehabilitation; however, the morphological anatomy of the vagus nerve targeted by stimulatation is poorly understood. Here, we used microCT to quantify the fascicular structure and neuroanatomy of human cervical vagus nerves (cVNs). Approach. We collected eight mid-cVN specimens from five fixed cadavers (three left nerves, five right nerves). Analysis focused on the 'surgical window': 5 cm of length, centered around the VNS implant location. Tissue was stained with osmium tetroxide, embedded in paraffin, and imaged on a microCT scanner. We visualized and quantified the merging and splitting of fascicles, and report a morphometric analysis of fascicles: count, diameter, and area. Main results. In our sample of human cVNs, a fascicle split or merge event was observed every ∼560 µm (17.8 ± 6.1 events cm⁻¹). Mean morphological outcomes included: fascicle count (6.6 ± 2.8 fascicles; range 1–15), fascicle diameter (514 ± 142 µm; range 147–1360 µm), and total cross-sectional fascicular area (1.32 ± 0.41 mm²; range 0.58–2.27 mm). Significance. The high degree of fascicular splitting and merging, along with wide range in key fascicular morphological parameters across humans may help to explain the clinical heterogeneity in patient responses to VNS. These data will enable modeling and experimental efforts to determine the clinical effect size of such variation. These data will also enable efforts to design improved VNS electrodes.

Objective. Retinal electrical stimulation using multi-electrode arrays (MEAs) aims to restore visual object perception in blind patients. However, the rate and duration of the artificial visual sensations are limited due to the rapid response decay of the stimulated neurons. Hence, we investigated a novel nature-inspired saccadic-like stimulation paradigm (biomimetic) to evoke sustained retinal responses. For implementation, the macroelectrode was replaced by several contiguous microelectrodes and activated non-simultaneously but alternating topologically. Approach. MEAs with hexagonally arranged electrodes were utilized to simulate and record mouse retinal ganglion cells (RGCs). Two shapes were presented electrically using MEAs: a 6e-hexagon (six hexagonally arranged 10 µm electrodes; 6e-hexagon diameter: 80 µm) and a double-bar (180 µm spaced, 320 µm in length). Electrodes of each shape were activated in three different modes (simultaneous, circular, and biomimetic ('zig-zag')), stimulating at different frequencies (1–20 Hz). Main results. The biomimetic stimulation generated enhanced RGC responses increasing the activity rate by 87.78%. In the spatiotemporal context, the electrical representation of the 6e-hexagon produced sustained and local RGC responses (∼130 µm corresponding to ∼2.5° of the human visual angle) for up to 90 s at 10 Hz stimulation and resolved the electrically presented double-bar. In contrast, during conventional simultaneous stimulation, the responses were poor and declined within seconds. Similarly, the applicability of the biomimetic mode for retinal implants (7 × 8 pixels) was successfully demonstrated. An object shape impersonating a smile was presented electrically, and the recorded data were used to emulate the implant's performance. The spatiotemporal pixel mapping of the activity produced a complete retinal image of the smile. Significance. The application of electrical stimulation in the biomimetic mode produced locally enhanced RGC responses with significantly reduced fading effects and yielded advanced spatiotemporal performance reflecting the presented electrode shapes in the mapped activity imprint. Therefore, it is likely that the RGC responses persist long enough to evoke visual perception and generate a seamless image, taking advantage of the flicker fusion. Hence, replacing the implant's macroelectrodes with microelectrodes and their activation in a topologically alternating biomimetic fashion may overcome the patient's perceptual image fading, thereby enhancing the spatiotemporal characteristics of artificial vision.

Work in the last two decades has shown that neural mass models (NMM) can realistically reproduce and explain epileptic seizure transitions as recorded by electrophysiological methods (EEG, SEEG). In previous work, advances were achieved by increasing excitation and heuristically varying network inhibitory coupling parameters in the models. Based on these early studies, we provide a laminar NMM capable of realistically reproducing the electrical activity recorded by SEEG in the epileptogenic zone during interictal to ictal states. With the exception of the external noise input into the pyramidal cell population, the model dynamics are autonomous. By setting the system at a point close to bifurcation, seizure-like transitions are generated, including pre-ictal spikes, low voltage fast activity, and ictal rhythmic activity. A novel element in the model is a physiologically motivated algorithm for chloride dynamics: the gain of GABAergic post-synaptic potentials is modulated by the pathological accumulation of chloride in pyramidal cells due to high inhibitory input and/or dysfunctional chloride transport. In addition, in order to simulate SEEG signals for comparison with real seizure recordings, the NMM is embedded first in a layered model of the neocortex and then in a realistic physical model. We compare modeling results with data from four epilepsy patient cases. By including key pathophysiological mechanisms, the proposed framework captures succinctly the electrophysiological phenomenology observed in ictal states, paving the way for robust personalization methods based on NMMs.

Objective. Retinal prostheses aim at restoring sight in patients with retinal degeneration by electrically stimulating the inner retinal neurons. Clinical trials with patients blinded by atrophic age-related macular degeneration using the PRIMA subretinal implant, a 2 × 2 mm array of 100 µm-wide photovoltaic pixels, have demonstrated a prosthetic visual acuity closely matching the pixel size. Further improvement in resolution requires smaller pixels, which, with the current bipolar design, necessitates more intense stimulation. Approach. We examine the lower limit of the pixel size for PRIMA implants by modeling the electric field, leveraging the clinical benchmarks, and using animal data to assess the stimulation strength and contrast of various patterns. Visually evoked potentials measured in Royal College of Surgeons rats with photovoltaic implants composed of 100 µm and 75 µm pixels were compared to clinical thresholds with 100 µm pixels. Electrical stimulation model calibrated by the clinical and rodent data was used to predict the performance of the implant with smaller pixels. Main results. PRIMA implants with 75 µm bipolar pixels under the maximum safe near-infrared (880 nm) illumination of 8 mW mm⁻² with 30% duty cycle (10 ms pulses at 30 Hz) should provide a similar perceptual brightness as with 100 µm pixels under 3 mW mm⁻² irradiance, used in the current clinical trials. Contrast of the Landolt C pattern scaled down to 75 µm pixels is also similar under such illumination to that with 100 µm pixels, increasing the maximum acuity from 20/420 to 20/315. Significance. Computational modeling defines the minimum pixel size of the PRIMA implants as 75 µm. Increasing the implant width from 2 to 3 mm and reducing the pixel size from 100 to 75 µm will nearly quadrupole the number of pixels, which should be very beneficial for patients. Smaller pixels of the same bipolar flat geometry would require excessively intense illumination, and therefore a different pixel design should be considered for further improvement in resolution.

Objective. Trans-corneal electrical stimulation (TcES) produces therapeutic effects on many ophthalmic diseases non-invasively. Existing clinical TcES devices use largely variable design of electrode distribution and stimulation parameters. Better understanding of how electrode configuration paradigms and stimulation parameters influence the electric field distribution on the retina, will be beneficial to the design of next-generation TcES devices. Approach. In this study, we constructed a realistic finite element human head model with fine eyeball structure. Commonly used DTL-Plus and ERG-Jet electrodes were simulated. We then conducted in silico investigations of retina observation surface (ROS) electric field distributions induced by different return electrode configuration paradigms and different stimulus intensities. Main results. Our results suggested that the ROS electric field distribution could be modulated by re-designing TcES electrode settings and stimulus parameters. Under far return location paradigms, either DTL-Plus or ERG-Jet approach could induce almost identical ROS electric field distribution regardless where the far return was located. However, compared with the ERG-Jet mode, DTL-Plus stimulation induced stronger nasal lateralization. In contrast, ERG-Jet stimulation induced relatively stronger temporal lateralization. The ROS lateralization can be further tweaked by changing the DTL-Plus electrode length. Significance. These results may contribute to the understanding of the characteristics of DTL-Plus and ERG-Jet electrodes based electric field distribution on the retina, providing practical implications for the therapeutic application of TcES.

Objective. In partial epilepsies, interictal epileptiform discharges (IEDs) are paroxysmal events observed in epileptogenic zone (EZ) and non-epileptogenic zone (NEZ). IEDs' generation and recurrence are subject to different hypotheses: they appear through glutamatergic and gamma-aminobutyric acidergic (GABAergic) processes; they may trigger seizures or prevent seizure propagation. This paper focuses on a specific class of IEDs, spike-waves (SWs), characterized by a short-duration spike followed by a longer duration wave, both of the same polarity. Signal analysis and neurophysiological mathematical models are used to interpret puzzling IED generation. Approach. Interictal activity was recorded by intracranial stereo-electroencephalography (SEEG) electrodes in five different patients. SEEG experts identified the epileptic and non-epileptic zones in which IEDs were detected. After quantifying spatial and temporal features of the detected IEDs, the most significant features for classifying epileptic and non-epileptic zones were determined. A neurophysiologically-plausible mathematical model was then introduced to simulate the IEDs and understand the underlying differences observed in epileptic and non-epileptic zone IEDs. Main results. Two classes of SWs were identified according to subtle differences in morphology and timing of the spike and wave component. Results showed that type-1 SWs were generated in epileptogenic regions also involved at seizure onset, while type-2 SWs were produced in the propagation or non-involved areas. The modeling study indicated that synaptic kinetics, cortical organization, and network interactions determined the morphology of the simulated SEEG signals. Modeling results suggested that the IED morphologies were linked to the degree of preserved inhibition. Significance. This work contributes to the understanding of different mechanisms generating IEDs in epileptic networks. The combination of signal analysis and computational models provides an efficient framework for exploring IEDs in partial epilepsies and classifying EZ and NEZ.

Objective. Microelectronic retinal implant aims to restore functional vision with electric stimulation. Short pulses are generally known to directly activate retinal ganglion cells (RGCs) with a notion of one or two spike(s) per pulse. In the present work, we systematically explore network-mediated responses that arise from various short pulses in both normal and degenerate retinas. Approach. Cell-attached patch clamping was used to record spiking responses of RGCs in wild-type (C57BL/6J) and retinal degeneration (rd10) mice. Alpha RGCs of the mouse retinas were targeted by their large soma sizes and classified by their responses to spot flashes. Then, RGCs were electrically stimulated by various conditions such as duration (100–460 μs), count (1–10), amplitude (100–400 μA), and repeating frequency (10–40 Hz) of short pulses. Also, their responses were compared with each own response to a single 4 ms long pulse which is known to evoke strong indirect responses. Main results. Short pulses evoked strong network-mediated responses not only in both ON and OFF types of RGCs in the healthy retinas but also in RGCs of the severely degenerate retina. However, the spike timing consistency across repeats not decreased significantly in the rd10 RGCs compared to the healthy ON and OFF RGCs. Network-mediated responses of ON RGCs were highly dependent on the current amplitude of stimuli but much less on the pulse count and the repetition frequency. In contrast, responses of OFF RGCs were more influenced by the number of stimuli than the current amplitude. Significance. Our results demonstrate that short pulses also elicit indirect responses by activating presynaptic neurons. In the case of the commercial retinal prostheses using repeating short pulses, there is a possibility that the performance of clinical devices is highly related to the preserved retinal circuits. Therefore, examination of surviving retinal neurons in patients would be necessary to improve the efficacy of retinal prostheses.

Objective. Because of the lack of highly skilled experts, automated technologies that support electroencephalogram (EEG)-based in epilepsy diagnosis are advancing. Deep convolutional neural network-based models have been used successfully for detecting epileptic spikes, one of the biomarkers, from EEG. However, a sizeable number of supervised EEG records are required for training. Approach. This study introduces the Satelight model, which uses the self-attention (SA) mechanism. The model was trained using a clinical EEG dataset labeled by five specialists, including 16 008 epileptic spikes and 15 478 artifacts from 50 children. The SA mechanism is expected to reduce the number of parameters and efficiently extract features from a small amount of EEG data. To validate the effectiveness, we compared various spike detection approaches with the clinical EEG data. Main results. The experimental results showed that the proposed method detected epileptic spikes more effectively than other models (accuracy = 0.876 and false positive rate = 0.133). Significance. The proposed model had only one-tenth the number of parameters as the other effective model, despite having such a high detection performance. Further exploration of the hidden parameters revealed that the model automatically attended to the EEG's characteristic waveform locations of interest.

Objective. PRIMA, the photovoltaic subretinal prosthesis, restores central vision in patients blinded by atrophic age-related macular degeneration (AMD), with a resolution closely matching the 100 µm pixel size of the implant. Improvement in resolution requires smaller pixels, but the resultant electric field may not provide sufficient stimulation strength in the inner nuclear layer (INL) or may lead to excessive crosstalk between neighboring electrodes, resulting in low contrast stimulation patterns. We study the approaches to electric field shaping in the retina for prosthetic vision with higher resolution and improved contrast. Approach. We present a new computational framework, Retinal Prosthesis Simulator (RPSim), that efficiently computes the electric field in the retina generated by a photovoltaic implant with thousands of electrodes. Leveraging the PRIMA clinical results as a benchmark, we use RPSim to predict the stimulus strength and contrast of the electric field in the retina with various pixel designs and stimulation patterns. Main results. We demonstrate that by utilizing monopolar pixels as both anodes and cathodes to suppress crosstalk, most patients may achieve resolution no worse than 48 µm. Closer proximity between the electrodes and the INL, achieved with pillar electrodes, enhances the stimulus strength and contrast and may enable 24 µm resolution with 20 µm pixels, at least in some patients. Significance. A resolution of 24 µm on the retina corresponds to a visual acuity of 20/100, which is over 4 times higher than the current best prosthetic acuity of 20/438, promising a significant improvement of central vision for many AMD patients.

Objective. High-frequency oscillations (HFOs) are promising biomarkers for localizing epileptogenic brain tissue. Previous studies have revealed that HFOs that present concurrence with interictal epileptic discharges (IEDs) better delineate epileptogenic brain tissue, particularly for epilepsy patients with multitype interictal discharges. However, the analysis of noninvasively recorded epileptic HFOs involves many complex procedures, such as data preprocessing, detection and source localization, impeding the translation of this approach to clinical practice. Approach. To address these problems, we developed a graphical user interface (GUI)-based pipeline called EMHapp, which can be used for the automatic detection, source localization and visualization of HFO events concurring with IEDs in magnetoencephalography (MEG) signals by using a beamformer-based virtual sensor (VS) technique. An improved VS reconstruction method was developed to enhance the amplitudes of both HFO and IED VS signals. To test the capability of our pipeline, we collected MEG data from 11 complex focal epilepsy patients with surgical resections or seizure onset zones (SOZs) that were identified by intracranial electroencephalography. Main results. Our results showed that the HFO sources of eight patients were concordant with their resection margins or SOZs. Our proposed VS signal reconstruction approach achieved an 83.2% improvement regarding the number of detected HFO events and a 17.3% improvement in terms of the spatial overlaps between the HFO sources and the resection margins or SOZs in comparison with conventional VS reconstruction approaches. Significance. EMHapp is the first GUI-based pipeline for the analysis of epileptic magnetoencephalographic HFOs, which conveniently obtains HFO source locations using clinical data and enables direct translation to clinical applications.

Objective. Electrochemically safe and efficient charge injection for neural stimulation necessitates monitoring of polarization and enhanced charge injection capacity of the stimulating electrodes. In this work, we present improved microstimulation capability by developing a custom-designed multichannel portable neurostimulator with a fully programmable anodic bias circuitry and voltage transient monitoring feature. Approach. We developed a 16-channel multichannel neurostimulator system, compared charge injection capacities as a function of anodic bias potentials, and demonstrated convenient control of the system by a custom-designed user interface allowing bidirectional wireless data transmission of stimulation parameters and recorded voltage transients. Charge injections were conducted in phosphate-buffered saline with silicon-based iridium oxide microelectrodes. Main results. Under charge-balanced 200 µs cathodic first pulsing, the charge injection capacities increased proportionally to the level of anodic bias applied, reaching a maximum of ten-fold increase in current intensity from 10 µA (100 µC cm⁻²) to 100 µA (1000 µC cm⁻²) with a 600 mV anodic bias. Our custom-designed and completely portable 16-channel neurostimulator enabled a significant increase in charge injection capacity in vitro. Significance. Limited charge injection capacity has been a bottleneck in neural stimulation applications, and our system may enable efficacious behavioral animal study involving chronic microstimulation while ensuring electrochemical safety.

Objective. Subretinal prostheses electrically stimulate the residual inner retinal neurons to partially restore vision. We investigated the changes in neurosensory macular structures and it is thickness associated with subretinal implantation in geographic atrophy (GA) secondary to age-related macular degeneration (AMD). Approach. Using optical coherence tomography, changes in distance between electrodes and retinal inner nuclear layer (INL) as well as alterations in thickness of retinal layers were measured over time above and near the subretinal chip implanted within the atrophic area. Retinal thickness (RT) was quantified across the implant surface and edges as well as outside the implant zone to compare with the natural macular changes following subretinal surgery, and the natural course of dry AMD. Main results. GA was defined based on complete retinal pigment epithelium and outer retinal atrophy (cRORA). Based on the analysis of three patients with subretinal implantation, we found that the distance between the implant and the target cells was stable over the long-term follow-up. Total RT above the implant decreased on average, by 39 ± 12 µm during 3 months post-implantation, but no significant changes were observed after that, up to 36 months of the follow-up. RT also changed near the temporal entry point areas outside the implantation zone following the surgical trauma of retinal detachment. There was no change in the macula cRORA nasal to the implanted zone, where there was no surgical trauma or manipulation. Significance. The surgical delivery of the photovoltaic subretinal implant causes minor RT changes that settle after 3 months, and then remain stable over long-term with no adverse structural or functional effects. Distance between the implant and the INL remains stable up to 36 months of the follow-up.

Objective. Current autism clinical detection relies on doctor observation and filling of clinical scales, which is subjective and prone to misdetection. Existing autism research of functional magnetic resonance imaging (fMRI) over-compresses the time-scale information and has poor generalization ability. This study extracts multiple time scale brain features of fMRI, providing objective detection. Approach. We first use least absolute shrinkage and selection operator to build a sparse network and extract features with a time scale of 1. Then, we use hidden markov model to extract features that describe the dynamic changes of the brain, with a time scale of 2. Additionally, to analyze the features of the potential network activity of autism from a higher time scale, we use long short-term memory to construct an auto-encoder to re-encode the original data and extract the features at a higher time scale, with a time scale of T, and T is the time length of fMRI. We use recursive feature elimination for feature selection for three different time scale features, merge them into multiple time scale features, and finally use one-dimensional convolution neural network for classification. Main results. Compared with well-established models, our method has achieved better results. The accuracy of our method is 76.0%, and the area under the roc curve is 0.83, tested on completely independent data, so our method has better generalization ability. Significance. This research analyzes fMRI sequences from multiple time scale to detect autism, and it also provides a new framework and research ideas for subsequent fMRI analysis.

Objective. Transcranial focused ultrasound (tFUS) is a neuromodulation technique which has been the focus of increasing interest for noninvasive brain stimulation with high spatial specificity. Its ability to excite and inhibit neural circuits as well as to modulate perception and behavior has been demonstrated, however, we currently lack understanding of how tFUS modulates the ways neurons interact with each other. This understanding would help elucidate tFUS's mechanism of systemic neuromodulation and allow future development of therapies for treating neurological disorders. Approach. In this study, we investigate how tFUS modulates neural interaction and response to peripheral electrical limb stimulation through intracranial multi-electrode recordings in the rat somatosensory cortex. We deliver ultrasound in a pulsed pattern to induce frequency dependent plasticity in a manner similar to what is found following electrical stimulation. Main Results. We show that neural firing in response to peripheral electrical stimulation is increased after ultrasound stimulation at all frequencies, showing tFUS induced changes in excitability of individual neurons in vivo. We demonstrate tFUS sonication repetition frequency dependent pairwise correlation changes between neurons, with both increases and decreases observed at different frequencies. Significance. These results extend previous research showing tFUS to be capable of inducing synaptic depression and demonstrate its ability to modulate network dynamics as a whole.

Objective. Temporal interference stimulation (TIS) has shown the potential as a new method for selective stimulation of deep brain structures in small animal experiments. However, it is challenging to deliver a sufficient temporal interference (TI) current to directly induce an action potential in the deep area of the human brain when electrodes are attached to the scalp because the amount of injection current is generally limited due to safety issues. Thus, we propose a novel method called epidural TIS (eTIS) to address this issue; in this method, the electrodes are attached to the epidural surface under the skull. Approach. We employed finite element method (FEM)-based electric field simulations to demonstrate the feasibility of eTIS. We first optimized the electrode conditions to deliver maximum TI currents to each of the three different targets (anterior hippocampus, subthalamic nucleus, and ventral intermediate nucleus) based on FEM, and compared the stimulation focality between eTIS and transcranial TIS (tTIS). Moreover, we conducted realistic skull-phantom experiments for validating the accuracy of the computational simulation for eTIS. Main results. Our simulation results showed that eTIS has the advantage of avoiding the delivery of TI currents over unwanted neocortical regions compared with tTIS for all three targets. It was shown that the optimized eTIS could induce neural action potentials at each of the three targets when a sufficiently large current equivalent to that for epidural cortical stimulation is injected. Additionally, the simulated results and measured results via the phantom experiments were in good agreement. Significance. We demonstrated the feasibility of eTIS, facilitating more focalized and stronger electrical stimulation of deep brain regions than tTIS, with the relatively less invasive placement of electrodes than conventional deep brain stimulation via computational simulation and realistic skull phantom experiments.

Objective. Research on the decoding of brain signals to control external devices is rapidly emerging due to its versatile potential applications, including neuroprosthetic control and neurorehabilitation. Electroencephalogram (EEG)-based non-invasive brain–computer interface (BCI) systems decode brain signals to establish an augmented communication and control pathway between the brain and the computer. The development of an efficient BCI system requires accurate decoding of neural activity underlying the user's intentions. This study investigates the directional tuning of EEG characteristics from the posterior parietal region, associated with bidirectional hand movement imagination or motor imagery (MI) in left and right directions. Approach. The imagined movement directions of the chosen hand were decoded using a combination of envelope and phase features derived from parietal EEGs of both hemispheres. The proposed algorithm uses wavelets for spectral decomposition, and discriminative subject-specific subband levels are identified based on Fisher analysis of envelope and phase features. The selected features from the discriminative subband levels are used to classify left and right MI directions of the hand using a support vector machine classifier. Furthermore, the performance of the proposed algorithm is evaluated by incorporating a maximum-variance-based EEG time bin selection algorithm. Main results. With the time bin selection approach using subject-specific features, the proposed algorithm yielded an average left vs right MI direction decoding accuracy of 73.33% across 15 healthy subjects. In addition, the decoding accuracy offered by the phase features was higher than that of the envelope features, indicating the importance of phase features in MI kinematics decoding. Significance. The results reveal the significance of the parietal EEG in decoding of imagined kinematics and open new possibilities for future BCI research.

Objective. The study of human neuromechanical control at the motor unit (MU) level has predominantly focussed on electrical activity and force generation, whilst the link between these, i.e. the muscle deformation, has not been widely studied. To address this gap, we analysed the kinematics of muscle units in natural contractions. Approach. We combined high-density surface electromyography (HDsEMG) and ultrafast ultrasound (US) recordings, at 1000 frames per second, from the tibialis anterior muscle to measure the motion of the muscular tissue caused by individual MU contractions. The MU discharge times were identified online by decomposition of the HDsEMG and provided as biofeedback to 12 subjects who were instructed to keep the MU active at the minimum discharge rate (9.8 ± 4.7 pulses per second; force less than 10% of the maximum). The series of discharge times were used to identify the velocity maps associated with 51 single muscle unit movements with high spatio-temporal precision, by a novel processing method on the concurrently recorded US images. From the individual MU velocity maps, we estimated the region of movement, the duration of the motion, the contraction time, and the excitation–contraction (E–C) coupling delay. Main results. Individual muscle unit motions could be reliably identified from the velocity maps in 10 out of 12 subjects. The duration of the motion, total contraction time, and E–C coupling were 17.9 $\pm$ 5.3 ms, 56.6 $\pm$ 8.4 ms, and 3.8 $\pm$ 3.0 ms (n = 390 across ten participants). The experimental measures also provided the first evidence of muscle unit twisting during voluntary contractions and MU territories with distinct split regions. Significance. The proposed method allows for the study of kinematics of individual MU twitches during natural contractions. The described measurements and characterisations open new avenues for the study of neuromechanics in healthy and pathological conditions.

Objective. To spatio-temporally resolve cardiac signals in functional magnetic resonance imaging (fMRI) time-series of the human brain using neither external physiological measurements nor ad hoc modelling assumptions. Approach. Cardiac pulsation is a physiological confound of fMRI time-series that introduces spurious signal fluctuations in proximity to blood vessels. fMRI alone is not sufficiently fast to resolve cardiac pulsation. Depending on the ratio between the instantaneous heart-rate and the acquisition sampling frequency ($1/{\text{TR}}$, with TR being the repetition time), the cardiac signal may alias into the frequency band of neural activation so that its removal through spectral filtering techniques is generally not possible. In this paper, we show that it is feasible to temporally and spatially resolve cardiac signals throughout the brain even when cardiac aliasing occurs by combining fMRI hyper-sampling with simultaneous multislice (SMS) imaging. The technique, which we name WHOle-brain CArdiac signal REgression from highly accelerated simultaneous multi-Slice fMRI acquisitions (WHOCARES), was developed on 695 healthy subjects selected from the Human Connectome Project and its performance validated against the RETROICOR, HAPPY and the pulse oxymeter signal regression methods. Main results. WHOCARES is capable of retrieving voxel-wise cardiac signal regressors. This is achieved without employing external physiological recordings nor through ad hoc modelling assumptions. The performance of WHOCARES was, on average, superior to RETROICOR, HAPPY and the pulse oxymeter regression methods. Significance. WHOCARES holds basis for the reliable mapping of cardiac activity in fMRI time-series. WHOCARES can be employed for the retrospective removal of cardiac noise in publicly available fMRI datasets where physiological recordings are not available. WHOCARES is freely available at https://github.com/gferrazzi/WHOCARES.

Objective. By means of electrical stimulation of the visual system, visual prostheses provide promising solution for blind patients through partial restoration of their vision. Despite the great success achieved so far in this field, the limited resolution of the perceived vision using these devices hinders the ability of visual prostheses users to correctly recognize viewed objects. Accordingly, we propose a deep learning approach based on generative adversarial networks (GANs), termed prosthetic vision GAN (PVGAN), to enhance object recognition for the implanted patients by representing objects in the field of view based on a corresponding simplified clip art version. Approach. To assess the performance, an axon map model was used to simulate prosthetic vision in experiments involving normally-sighted participants. In these experiments, four types of image representation were examined. The first and second types comprised presenting phosphene simulation of real images containing the actual high-resolution object, and presenting phosphene simulation of the real image followed by the clip art image, respectively. The other two types were utilized to evaluate the performance in the case of electrode dropout, where the third type comprised presenting phosphene simulation of only clip art images without electrode dropout, while the fourth type involved clip art images with electrode dropout. Main results. The performance was measured through three evaluation metrics which are the accuracy of the participants in recognizing the objects, the time taken by the participants to correctly recognize the object, and the confidence level of the participants in the recognition process. Results demonstrate that representing the objects using clip art images generated by the PVGAN model results in a significant enhancement in the speed and confidence of the subjects in recognizing the objects. Significance. These results demonstrate the utility of using GANs in enhancing the quality of images perceived using prosthetic vision.

Objective. Brain computer interface (BCI) technology is an innovative way of information exchange, which can effectively convert physiological signals into control instructions of machines. Due to its spontaneity and device independence, the motor imagery (MI) electroencephalography (EEG) signal is used as a common BCI signal source to achieve direct control of external devices. Several online MI EEG-based systems have shown potential for rehabilitation. However, the generalization ability of the current classification model of MI tasks is still limited and the real-time prototype is far from widespread in practice. Approach. To solve these problems, this paper proposes an optimized neural network architecture based on our previous work. Firstly, the artifact components in the MI-EEG signal are removed by using the threshold and threshold function related to the artifact removal evaluation index, and then the data is augmented by the empirical mode decomposition (EMD) algorithm. Furthermore, the ensemble learning (EL) method and fine-tuning strategy in transfer learning (TL) are used to optimize the classification model. Finally, combined with the flexible binary encoding strategy, the EEG signal recognition results are mapped to the control commands of the robotic arm, which realizes multiple degrees of freedom control of the robotic arm. Main results. The results show that EMD has an obvious data amount enhancement effect on a small dataset, and the EL and TL can improve intra-subject and inter-subject model evaluation performance, respectively. The use of a binary coding method realizes the expansion of control instructions, i.e. four kinds of MI-EEG signals are used to complete the control of 7 degrees of freedom of the robotic arm. Significance. Our work not only improves the classification accuracy of the subject and the generality of the classification model while also extending the BCI control instruction set.

Objective. Mixing/dissociation of sleep stages in narcolepsy adds to the difficulty in automatic sleep staging. Moreover, automatic analytical studies for narcolepsy and multiple sleep latency test (MSLT) have only done automatic sleep staging without leveraging the sleep stage profile for further patient identification. This study aims to establish an automatic narcolepsy detection method for MSLT. Approach. We construct a two-phase model on MSLT recordings, where ambiguous sleep staging and sleep transition dynamics make joint efforts to address this issue. In phase 1, we extract representative features from electroencephalogram (EEG) and electrooculogram (EOG) signals. Then, the features are input to an EasyEnsemble classifier for automatic sleep staging. In phase 2, we investigate sleep transition dynamics, including sleep stage transitions and sleep stages, and output likelihood of narcolepsy by virtue of principal component analysis (PCA) and a logistic regression classifier. To demonstrate the proposed framework in clinical application, we conduct experiments on 24 participants from the Children's Hospital of Fudan University, considering ten patients with narcolepsy and fourteen patients with MSLT negative. Main results. Applying the two-phase leave-one-subject-out testing scheme, the model reaches an accuracy, sensitivity, and specificity of 87.5%, 80.0%, and 92.9% for narcolepsy detection. Influenced by disease pathology, accuracy of automatic sleep staging in narcolepsy appears to decrease compared to that in the non-narcoleptic population. Significance. This method can automatically and efficiently distinguish patients with narcolepsy based on MSLT. It probes into the amalgamation of automatic sleep staging and sleep transition dynamics for narcolepsy detection, which would assist clinic and neuroelectrophysiology specialists in visual interpretation and diagnosis.

Objective. Although neural-enabled prostheses have been used to restore some lost functionality in clinical trials, they have faced difficulty in achieving high degree of freedom, natural use compared to healthy limbs. This study investigated the in vivo functionality of a flexible and scalable regenerative peripheral-nerve interface suspended within a microchannel-embedded, tissue-engineered hydrogel (the magnetically aligned regenerative tissue-engineered electronic nerve interface (MARTEENI)) as a potential approach to improving current issues in peripheral nerve interfaces. Approach. Assembled MARTEENI devices were implanted in the gaps of severed sciatic nerves in Lewis rats. Both acute and chronic electrophysiology were recorded, and channel-isolated activity was examined. In terminal experiments, evoked activity during paw compression and stimulus response curves generated from proximal nerve stimulation were examined. Electrochemical impedance spectroscopy was performed to assess the complex impedance of recording sites during chronic data collection. Features of the foreign-body response (FBR) in non-functional implants were examined using immunohistological methods. Main results. Channel-isolated activity was observed in acute, chronic, and terminal experiments and showed a typically biphasic morphology with peak-to-peak amplitudes varying between 50 and 500 µV. For chronic experiments, electrophysiology was observed for 77 days post-implant. Within the templated hydrogel, regenerating axons formed minifascicles that varied in both size and axon count and were also found to surround device threads. No axons were found to penetrate the FBR. Together these results suggest the MARTEENI is a promising approach for interfacing with peripheral nerves. Significance. Findings demonstrate a high likelihood that observed electrophysiological activity recorded from implanted MARTEENIs originated from neural tissue. The variation in minifascicle size seen histologically suggests that amplitude distributions observed in functional MARTEENIs may be due to a combination of individual axon and mini-compound action potentials. This study provided an assessment of a functional MARTEENI in an in vivo animal model for the first time.

Objective. Reaching hand movement is an important motor skill actively examined in the brain–computer interface (BCI). Among the various components of movement analyzed is the hand's trajectory, which describes the hand's continuous positions in three-dimensional space. While a large body of studies have investigated the decoding of real movements and the reconstruction of real hand movement trajectories from neural signals, fewer studies have attempted to decode the trajectory of the imagined hand movement. To develop BCI systems for patients with hand motor dysfunctions, the systems essentially have to achieve movement-free control of external devices, which is only possible through successful decoding of purely imagined hand movement. Approach. To achieve this goal, this study used a machine learning technique (i.e. the variational Bayesian least square) to analyze the electrocorticogram (ECoG) of 18 epilepsy patients obtained from when they performed movement execution (ME) and kinesthetic movement imagination (KMI) of the reach-and-grasp hand action. Main results. The variational Bayesian decoding model was able to successfully predict the imagined trajectories of the hand movement significantly above the chance level. The Pearson's correlation coefficient between the imagined and predicted trajectories was 0.3393 and 0.4936 for the KMI (KMI trials only) and MEKMI paradigm (alternating trials of ME and KMI), respectively. Significance. This study demonstrated a high accuracy of prediction for the trajectories of imagined hand movement, and more importantly, a higher decoding accuracy of the imagined trajectories in the MEKMI paradigm compared to the KMI paradigm solely.

Objective. Closed-loop prosthesis interfaces, which combine electromyography (EMG)-based control with supplementary feedback, represent a promising direction for developing the next generation of bionic limbs. However, we still lack an understanding of how users utilize these interfaces and how to evaluate competing solutions. In this study, we used the framework of speed–accuracy trade-off functions (SAF) to understand, evaluate, and compare the performance of two closed-loop user-prosthesis interfaces. Approach. Ten able-bodied participants and an amputee performed a force-matching task in a functional box-and-block setup at three different speeds. All participants were subjected to both interfaces in a crossover study design with a 1 week washout period. Importantly, both interfaces used direct proportional control but differed in the feedback provided to the participant (EMG feedback vs. Force feedback). We estimated the SAFs afforded by the two interfaces and sought to understand how the participants planned and executed the task under the various conditions. Main results. We found that execution speed significantly influenced performance, and that EMG feedback afforded better overall performance, especially at medium speeds. Notably, we found that there was a difference in the SAF between the two interfaces, with EMG feedback enabling participants to attain higher accuracies faster than Force feedback. Furthermore, both interfaces enabled participants to develop flexible control policies, while EMG feedback also afforded participants the ability to generate smoother, more repeatable EMG commands. Significance. Overall, the results indicate that the performance of closed-loop prosthesis interfaces depends critically on the feedback approach and execution speed. This study showed that the SAF framework could be used to reveal the differences between feedback approaches, which might not have been detected if the assessment was performed at a single speed. Therefore, we argue that it is important to consider the speed–accuracy trade-offs to rigorously evaluate and compare user-prosthesis interfaces.

Objective. Acoustoelectric brain imaging (ABI) is a potential noninvasive electrophysiological neuroimaging method with high spatiotemporal resolution. At the focal spot of the focused ultrasound, with the couple of acoustic and electric fields, high-frequency acoustoelectric (HF AE) signal is generated. Because the brain is a volume conductor, HF AE signal can be detected in other brain cortex. The processing of HF AE signal is critical for improving decoding precision, further improving the spatial resolution performance of ABI. This study investigates the processing network of HF AE signal in the living rat brain. Approach. When HF AE generated on the left primary visual cortex (V1-L), low-frequency (LF) electroencephalography and HF AE signals on different cortex were recorded at the same time. Firstly, AE signal on different sides of the brain cortex were compared, including prefrontal cortex (FrA) and primary somatosensory cortex (S1FL). Then, we constructed and analyzed functional networks of two signals. Main results. In the same cortex, HF AE signal on the right side had stronger intensity. And compared with LF networks, HF AE network had larger global efficiency and shorter characteristic path length, denoting the stronger processing and transmission of AE signal. Additionally, in HF AE network, the node had significantly increased local properties and the connection were concentrated in the occipital lobe, reflecting the occipital lobe plays an important role in the processing. Significance. Experiment results demonstrate that, compared with LF network, HF AE network is more efficient and had stronger transmission capabilities. And the connection of HF AE network is concentrated in the occipital lobe. This work preliminarily reveals the HF AE signal processing, which is significant for improving the ABI quality and provides a new insight for understanding the brain HF signal.

Objective. Steady-state visual evoked potentials (SSVEPs) based brain–computer interface (BCI) has received great interests owing to the high information transfer rate and available large number of targets. However, the performance of frequency recognition methods heavily depends on the amount of the calibration data for intra-subject classification. Some research adopted the deep learning (DL) algorithm to conduct the inter-subject classification, which could reduce the calculation procedure, but the performance still has large room to improve compared with the intra-subject classification. Approach. To address these issues, we proposed an efficient SSVEP DL NETwork (termed SSVEPNET) based on one-dimensional convolution and long short-term memory (LSTM) module. To enhance the performance of SSVEPNET, we adopted the spectral normalization and label smoothing technologies during implementing the network architecture. We evaluated the SSVEPNET and compared it with other methods for the intra- and inter-subject classification under different conditions, i.e. two datasets, two time-window lengths (1 s and 0.5 s), three sizes of training data. Main results. Under all the experimental settings, the proposed SSVEPNET achieved the highest average accuracy for the intra- and inter-subject classification on the two SSVEP datasets, when compared with other traditional and DL baseline methods. Significance. The extensive experimental results demonstrate that the proposed DL model holds promise to enhance frequency recognition performance in SSVEP-based BCIs. Besides, the mixed network structures with convolutional neural network and LSTM, and the spectral normalization and label smoothing could be useful optimization strategies to design efficient models for electroencephalography data.

Objective. To develop and in vivo demonstrate threadlike wireless implantable neuromuscular microstimulators that are digitally addressable. Approach. These devices perform, through its two electrodes, electronic rectification of innocuous high frequency current bursts delivered by volume conduction via epidermal textile electrodes. By avoiding the need of large components to obtain electrical energy, this approach allows the development of thin devices that can be intramuscularly implanted by minimally invasive procedures such as injection. For compliance with electrical safety standards, this approach requires a minimum distance, in the order of millimeters or a very few centimeters, between the implant electrodes. Additionally, the devices must cause minimal mechanical damage to tissues, avoid dislocation and be adequate for long-term implantation. Considering these requirements, the implants were conceived as tubular and flexible devices with two electrodes at opposite ends and, at the middle section, a hermetic metallic capsule housing the electronics. Main results. The developed implants have a submillimetric diameter (0.97 mm diameter, 35 mm length) and consist of a microcircuit, which contains a single custom-developed integrated circuit, housed within a titanium capsule (0.7 mm diameter, 6.5 mm length), and two platinum–iridium coils that form two electrodes (3 mm length) located at opposite ends of a silicone body. These neuromuscular stimulators are addressable, allowing to establish a network of microstimulators that can be controlled independently. Their operation was demonstrated in an acute study by injecting a few of them in the hind limb of anesthetized rabbits and inducing controlled and independent contractions. Significance. These results show the feasibility of manufacturing threadlike wireless addressable neuromuscular stimulators by using fabrication techniques and materials well established for chronic electronic implants. Although long-term operation still must be demonstrated, the obtained results pave the way to the clinical development of advanced motor neuroprostheses formed by dense networks of such wireless devices.

A growing number of studies have revealed significant abnormalities in electroencephalography (EEG) microstate in patients with depression, but these findings may be affected by medication. Therefore, how the EEG microstates abnormally change in patients with depression in the early stage and without the influence of medication has not been investigated so far. Resting-state EEG data and Hamilton Depression Rating Scale (HDRS) were collected from 34 first-episode drug-naïve adolescent with depression and 34 matched healthy controls. EEG microstate analysis was applied and nonlinear characteristics of EEG microstate sequences were studied by sample entropy and Lempel–Ziv complexity (LZC). The microstate temporal parameters and complexity were tried to train an SVM for classification of patients with depression. Four typical EEG microstate topographies were obtained in both groups, but microstate C topography was significantly abnormal in depression patients. The duration of microstate B, C, D and the occurrence and coverage of microstate B significantly increased, the occurrence and coverage of microstate A, C reduced significantly in depression group. Sample entropy and LZC in the depression group were abnormally increased and were negatively correlated with HDRS. When the combination of EEG microstate temporal parameters and complexity of microstate sequence was used to classify patients with depression from healthy controls, a classification accuracy of 90.9% was obtained. Abnormal EEG microstate has appeared in early depression, reflecting an underlying abnormality in configuring neural resources and transitions between distinct brain network states. EEG microstate can be used as a neurophysiological biomarker for early auxiliary diagnosis of depression.

Objective. To obtain a formalism for real-time concurrent sequential estimation of neural membrane time constant and input–output (IO) curve with transcranial magnetic stimulation (TMS). Approach. First, the neural membrane response and depolarization factor, which leads to motor evoked potentials with TMS are analytically computed and discussed. Then, an integrated model is developed which combines the neural membrane time constant and IO curve. Identifiability of the proposed integrated model is discussed. A condition is derived, which assures estimation of the proposed integrated model. Finally, sequential parameter estimation (SPE) of the neural membrane time constant and IO curve is described through closed-loop optimal sampling and open-loop uniform sampling TMS. Without loss of generality, this paper focuses on a specific case of commercialized TMS pulse shapes. The proposed formalism and SPE method are directly applicable to other pulse shapes. Main results. The results confirm satisfactory estimation of the membrane time constant and IO curve parameters. By defining a stopping rule based on five times consecutive convergence of the estimation parameters with a tolerances of 0.01, the membrane time constant and IO curve parameters are estimated with 82 TMS pulses with absolute relative estimation errors (AREs) of less than 4% with the optimal sampling SPE method. At this point, the uniform sampling SPE method leads to AREs up to 16%. The uniform sampling method does not satisfy the stopping rule due to the large estimation variations. Significance. This paper provides a tool for real-time closed-loop SPE of the neural time constant and IO curve, which can contribute novel insights in TMS studies. SPE of the membrane time constant enables selective stimulation, which can be used for advanced brain research, precision medicine and personalized medicine.

Objective. Since the introduction of transcranial temporal interference stimulation, there has been an ever-growing interest in this novel method, as it theoretically allows non-invasive stimulation of deep brain target regions. To date, attempts have been made to optimize the electrode montages and injected current to achieve personalized area targeting using two electrode pairs. Most of these methods use exhaustive search to find the best match, but faster and, at the same time, reliable solutions are required. In this study, the electrode combinations as well as the injected current for a two-electrode pair stimulation were optimized using a genetic algorithm, considering the right hippocampus as the region of interest (ROI). Approach. Simulations were performed on head models from the Population Head Model repository. First, each model was fitted with an electrode array based on the 10–10 international EEG electrode placement system. Following electrode placement, the models were meshed and solved for all single-pair electrode combinations, using an electrode on the left mastoid as a reference (ground). At the optimization stage, different electrode pairs and injection currents were tested using a genetic algorithm to obtain the optimal combination for each model, by setting three different maximum electric field thresholds (0.2, 0.5, and 0.8 V m⁻¹) in the ROI. The combinations below the set threshold were given a high penalty. Main results. Greater focality was achieved with our optimization, specifically in the ROI, with a significant decrease in the surrounding electric field intensity. In the non-optimized case, the mean brain volumes stimulated above 0.2 V m⁻¹ were 99.9% in the ROI, and 76.4% in the rest of the gray matter. In contrast, the stimulated mean volumes were 91.4% and 29.6%, respectively, for the best optimization case with a threshold of 0.8 V m⁻¹. Additionally, the maximum electric field intensity inside the ROI was consistently higher than that outside of the ROI for all optimized cases. Objective. Given that the accomplishment of a globally optimal solution requires a brute-force approach, the use of a genetic algorithm can significantly decrease the optimization time, while achieving personalized deep brain stimulation. The results of this work can be used to facilitate further studies that are more clinically oriented; thus, enabling faster and at the same time accurate treatment planning for the stimulation sessions.

Objective. To determine the effect of epilepsy on intracranial electroencephalography (EEG) functional connectivity, and the ability of functional connectivity to localize the seizure onset zone (SOZ), controlling for spatial biases. Approach. We analyzed intracranial EEG data from patients with drug-resistant epilepsy admitted for pre-surgical planning. We calculated intracranial EEG functional networks and determined whether changes in functional connectivity lateralized the SOZ using a spatial subsampling method to control for spatial bias. We developed a 'spatial null model' to localize the SOZ electrode using only spatial sampling information, ignoring EEG data. We compared the performance of this spatial null model against models incorporating EEG functional connectivity and interictal spike rates. Main results. About 110 patients were included in the study, although the number of patients differed across analyses. Controlling for spatial sampling, the average connectivity was lower in the SOZ region relative to the same anatomic region in the contralateral hemisphere. A model using intra-hemispheric connectivity accurately lateralized the SOZ (average accuracy 75.5%). A spatial null model incorporating spatial sampling information alone achieved moderate accuracy in classifying SOZ electrodes (mean AUC = 0.70, 95% CI 0.63–0.77). A model incorporating intracranial EEG functional connectivity and spike rate data further outperformed this spatial null model (AUC 0.78, p = 0.002 compared to spatial null model). However, a model incorporating functional connectivity without spike rate data did not significantly outperform the null model (AUC 0.72, p = 0.38). Significance. Intracranial EEG functional connectivity is reduced in the SOZ region, and interictal data predict SOZ electrode localization and laterality, however a predictive model incorporating functional connectivity without interictal spike rates did not significantly outperform a spatial null model. We propose constructing a spatial null model to provide an estimate of the pre-implant hypothesis of the SOZ, and to serve as a benchmark for further machine learning algorithms in order to avoid overestimating model performance because of electrode sampling alone.

Objective. Recently developed magnetoelectric nanoparticles (MENPs) provide a potential tool to enable different biomedical applications. They could be used to overcome the intrinsic constraints posed by traditional neurostimulation techniques, namely the invasiveness of electrodes-based techniques, the limited spatial resolution, and the scarce efficiency of magnetic stimulation. Approach. By using computational electromagnetic techniques, we modelled the behaviour of recently designed biocompatible MENPs injected, in the shape of clusters, in specific cortical targets of a highly detailed anatomical head model. The distributions and the tissue penetration of the electric fields induced by MENPs clusters in each tissue will be compared to the distributions induced by traditional transcranial magnetic stimulation (TMS) coils for non-invasive brain stimulation positioned on the left prefrontal cortex (PFC) of a highly detailed anatomical head model. Main results. MENPs clusters can induce highly focused electric fields with amplitude close to the neural activation threshold in all the brain tissues of interest for the treatment of most neuropsychiatric disorders. Conversely, TMS coils can induce electric fields of several tens of V m⁻¹ over a broad volume of the PFC, but they are unlikely able to efficiently stimulate even small volumes of subcortical and deep tissues. Significance. Our numerical results suggest that the use of MENPs for brain stimulation may potentially led to a future pinpoint treatment of neuropshychiatric disorders, in which an impairment of electric activity of specific cortical and subcortical tissues and networks has been assumed to play a crucial role.

Objective. Neural decoding is an important tool in neural engineering and neural data analysis. Of various machine learning algorithms adopted for neural decoding, the recently introduced deep learning is promising to excel. Therefore, we sought to apply deep learning to decode movement trajectories from the activity of motor cortical neurons. Approach. In this paper, we assessed the performance of deep learning methods in three different decoding schemes, concurrent, time-delay, and spatiotemporal. In the concurrent decoding scheme where the input to the network is the neural activity coincidental to the movement, deep learning networks including artificial neural network (ANN) and long-short term memory (LSTM) were applied to decode movement and compared with traditional machine learning algorithms. Both ANN and LSTM were further evaluated in the time-delay decoding scheme in which temporal delays are allowed between neural signals and movements. Lastly, in the spatiotemporal decoding scheme, we trained convolutional neural network (CNN) to extract movement information from images representing the spatial arrangement of neurons, their activity, and connectomes (i.e. the relative strengths of connectivity between neurons) and combined CNN and ANN to develop a hybrid spatiotemporal network. To reveal the input features of the CNN in the hybrid network that deep learning discovered for movement decoding, we performed a sensitivity analysis and identified specific regions in the spatial domain. Main results. Deep learning networks (ANN and LSTM) outperformed traditional machine learning algorithms in the concurrent decoding scheme. The results of ANN and LSTM in the time-delay decoding scheme showed that including neural data from time points preceding movement enabled decoders to perform more robustly when the temporal relationship between the neural activity and movement dynamically changes over time. In the spatiotemporal decoding scheme, the hybrid spatiotemporal network containing the concurrent ANN decoder outperformed single-network concurrent decoders. Significance. Taken together, our study demonstrates that deep learning could become a robust and effective method for the neural decoding of behavior.

Objective. The diagnosis of nerve disorders in humans has relied heavily on the measurement of electrical signals from nerves or muscles in response to electrical stimuli applied at appropriate locations on the body surface. The present study investigated the demyelinating subtype of Guillain–Barré syndrome using multiscale computational model simulations to verify how demyelination of peripheral axons may affect plantar flexion torque as well as the ongoing electromyogram (EMG) during voluntary isometric or isotonic contractions. Approach. Changes in axonal conduction velocities, mimicking those found in patients with the disease at different stages, were imposed on a multiscale computational neuromusculoskeletal model to simulate subjects performing unipodal plantar flexion force and position tasks. Main results. The simulated results indicated changes in the torque signal during the early phase of the disease while performing isotonic tasks, as well as in torque variability after partial conduction block while performing both isometric and isotonic tasks. Our results also indicated changes in the root mean square values and in the power spectrum of the soleus EMG signal as well as changes in the synchronization index computed from the firing times of the active motor units. All these quantitative changes in functional indicators suggest that the adoption of such additional measurements, such as torques and ongoing EMG, could be used with advantage in the diagnosis and be relevant in providing extra information for the neurologist about the level of the disease. Significance. Our findings enrich the knowledge of the possible ways demyelination affects force generation and position control during plantarflexion. Moreover, this work extends computational neuroscience to computational neurology and shows the potential of biologically compatible neuromuscular computational models in providing relevant quantitative signs that may be useful for diagnosis in the clinic, complementing the tools traditionally used in neurological electrodiagnosis.

Objective. Understanding neurophysiological changes that accompany transitions between anaesthetized and conscious states is a key objective of anesthesiology and consciousness science. This study aimed to characterize the dynamics of auditory-evoked potential morphology in mice along a continuum of consciousness. Approach. Epidural field potentials were recorded from above the primary auditory cortices of two groups of laboratory mice: urethane-anaesthetized (A, n = 14) and conscious (C, n = 17). Both groups received auditory stimulation in the form of a repeated pure-tone stimulus, before and after receiving 10 mg kg⁻¹ i.p. ketamine (AK and CK). Evoked responses were then ordered by ascending sample entropy into AK, A, CK, and C, considered to reflect physiological correlates of awareness. These data were used to train a recurrent neural network (RNN) with an input parameter encoding state. Model outputs were compared with grand-average event-related potential (ERP) waveforms. Subsequently, the state parameter was varied to simulate changes in the ERP that occur during transitions between states, and relationships with dominant peak amplitudes were quantified. Main results. The RNN synthesized output waveforms that were in close agreement with grand-average ERPs for each group (r² > 0.9, p < 0.0001). Varying the input state parameter generated model outputs reflecting changes in ERP morphology predicted to occur between states. Positive peak amplitudes within 25–50 ms, and negative peak amplitudes within 50–75 ms post-stimulus-onset, were found to display a sigmoidal characteristic during the transition from anaesthetized to conscious states. In contrast, negative peak amplitudes within 0–25 ms displayed greater linearity. Significance. This study demonstrates a method for modelling changes in ERP morphology that accompany transitions between states of consciousness using an RNN. In future studies, this approach may be applied to human data to support the clinical use of ERPs to predict transition to consciousness.

Objective. Flexible implantable electrodes enable months-long stable recording of single-unit signals from rat brains. Despite extensive efforts in the development of flexible probes for brain recording, thus far there are no conclusions on their application in long-term single neuronal recording from the spinal cord which is more mechanically active. To this end, we realized the chronic recording of single-unit signals from the spinal cord of freely-moving rats using flexible carbon nanotube fiber (CNTF) electrodes. Approach. We developed flexible CNTF electrodes for intraspinal recording. Continuous in vivo impedance monitoring and histology studies were conducted to explore the critical factors determining the longevity of the recording, as well as to illustrate the evolution of the electrode–tissue interface. Gait analysis were performed to evaluate the biosafety of the chronic intraspinal implantation of the CNTF electrodes. Main results. By increasing the insulation thickness of the CNTF electrodes, single-unit signals were continuously recorded from the spinal cord of freely-moving rats without electrode repositioning for 3–4 months. Single neuronal and local field potential activities in response to somatic mechanical stimulation were successfully recorded from the spinal dorsal horns. Histological data demonstrated the ability of the CNTF microelectrodes to form an improved intraspinal interfaces with greatly reduced gliosis compared to their stiff metal counterparts. Continuous impedance monitoring suggested that the longevity of the intraspinal recording with CNTF electrodes was determined by the insulation durability. Gait analysis showed that the chronic presence of the CNTF electrodes caused no noticeable locomotor deficits in rats. Significance. It was found that the chronic recording from the spinal cord faces more stringent requirements on the electrode structural durability than recording from the brain. The stable, long-term intraspinal recording provides unique capabilities for studying the physiological functions of the spinal cord relating to motor, sensation, and autonomic control in both health and disease.

Objective. Motor imagery-based brain computer interfaces (MI-BCIs) have been widely researched because they do not demand external stimuli and have a high degree of maneuverability. In most scenarios, superabundant selected channels, fixed time windows, and frequency bands would certainly affect the performance of MI-BCIs due to the neurophysiological diversities among different individuals. In this study, we attempt to effectively use the Riemannian geometry of spatial covariance matrix to extract more robust features and thus enhance the decoding efficiency. Approach. First, we utilize a Riemannian distance-based electroencephalography (EEG) channel selection method, which preliminarily reduces the information redundancy in the first stage. Second, we extract discriminative Riemannian tangent space features of EEG signals of selected channels from the most discriminant time-frequency bands to further enhance decoding accuracy for MI-BCIs. Finally, we train a support vector machine model with a linear kernel to classify our extracted discriminative Riemannian features, and evaluate our proposed method using publicly available BCI Competition IV dataset Ⅰ (DS1) and Competition Ⅲ dataset Ⅲa (DS2). Main results. The experimental results show that the average classification accuracy with the selected 16-channel EEG signals of our method is 90.0% and 89.4% in DS1 and DS2 respectively. The average improvements are 20.0% and 21.2% on DS1, 9.4% and 7.2% on DS2 for 8 and 16 selected channels, respectively. Significance. These results show that our proposed method is a promising candidate for the performance improvement of MI-BCIs.

Objective. The challenge for motor imagery (MI) in brain-computer interface (BCI) systems is finding a reliable classification model that has high classification accuracy and excellent robustness. Currently, one of the main problems leading to degraded classification performance is the inaccuracy caused by nonstationarities and low signal-to-noise ratio in electroencephalogram (EEG) signals. Approach. This study proposes a novel attention-based 3D densely connected cross-stage-partial network (DCSPNet) model to achieve efficient EEG-based MI classification. This is an end-to-end classification model framework based on the convolutional neural network (CNN) architecture. In this framework, to fully utilize the complementary features in each dimension, the optimal features are extracted adaptively from the EEG signals through the spatial-spectral-temporal (SST) attention mechanism. The 3D DCSPNet is introduced to reduce the gradient loss by segmenting the extracted feature maps to strengthen the network learning capability. Additionally, the design of the densely connected structure increases the robustness of the network. Main results. The performance of the proposed method was evaluated using the BCI competition IV 2a and the high gamma dataset, achieving an average accuracy of 84.45% and 97.88%, respectively. Our method outperformed most state-of-the-art classification algorithms, demonstrating its effectiveness and strong generalization ability. Significance. The experimental results show that our method is promising for improving the performance of MI-BCI. As a general framework based on time-series classification, it can be applied to BCI-related fields.

Objective. Transcranial magnetic stimulation-electroencephalogram (TMS-EEG) technology has played an increasingly important role in the field of neuroscience, and closed loop TMS has also been gradually concerned. However, the characteristics of closed-loop TMS-EEG were few discussed. To study the dependence of EEG reactivity on cortical oscillation phase under TMS stimulation, we explored in detail the TMS-EEG characteristics induced by closed-loop TMS contingent on occipital alpha phase. Approach. By collecting 30 healthy volunteers' closed-loop TMS-EEG data, we verified the real-time accuracy of our closed-loop system and analyzed the inter-trial phase coherence (ITPC) value, the TMS-induced natural frequency, the N100 TMS-evoked potential and the spatial characteristics of TMS-EEG data. Main results. The ITPC value of closed-loop TMS-EEG was higher than that of open loop TMS-EEG, suggesting that our research improves the repeatability of TMS-EEG experiments; the alpha power induced by 0° TMS was higher than that induced by 180° stimulation in the central region and parietal/occipital lobe; the N100 amplitude of 90° (3.85 μV) stimulation was significantly higher than that of 270° (1.87 μV) stimulation, and the latency of the N100 of the 90° stimulation (mean 95.01 ms) was significantly less than that of the 270° stimulation (mean 113.94 ms); the topographical distributions of the N45-P70-N100 potential were significantly affected by the O1 alpha phase at the moment of TMS. Significance. Our experimental results provided support for the dependence of EEG reactivity on cortical oscillation phase under TMS stimulation.

Objective. The electroencephalogram (EEG) is one of the most important brain-imaging tools. The few-channel EEG is more suitable and affordable for practical use as a wearable device. Removing artifacts from collected EEGs is a prerequisite for accurately interpreting brain function and state. Previous studies proposed methods combining signal decomposition with the blind source separation (BSS) algorithms, but most of them used threshold-based criteria for artifact rejection, resulting in a lack of effectiveness in removing specific artifacts and the excessive suppression of brain activities. In this study, we proposed an outlier detection-based method for artifact removal under the few-channel condition. Approach. The underlying components (sources) were extracted using the decomposition-BSS schema. Based on our assumptions that in the feature space, the artifact-related components are dispersed, while the components related to brain activities are closely distributed, the artifact-related components were identified and rejected using one-class support vector machine. The assumptions were validated by visualizing the distribution of clusters of components. Main results. In quantitative analyses with semisimulated data, the proposed method outperformed the threshold-based methods for various artifacts, including muscle artifact, ocular artifact, and power line noise. With a real dataset and an event-related potential dataset, the proposed method demonstrated good performance in real-life situations. Significance. This study provided a fully data-driven and adaptive method for removing various artifacts in a single process without excessive suppression of brain activities.

All electric and magnetic stimulation of the brain deposits thermal energy in the brain. This occurs through either Joule heating of the conductors carrying current through electrodes and magnetic coils, or through dissipation of energy in the conductive brain. Objective. Although electrical interaction with brain tissue is inseparable from thermal effects when electrodes are used, magnetic induction enables us to separate Joule heating from induction effects by contrasting AC and DC driving of magnetic coils using the same energy deposition within the conductors. Since mammalian cortical neurons have no known sensitivity to static magnetic fields, and if there is no evidence of effect on spike timing to oscillating magnetic fields, we can presume that the induced electrical currents within the brain are below the molecular shot noise where any interaction with tissue is purely thermal. Approach. In this study, we examined a range of frequencies produced from micromagnetic coils operating below the molecular shot noise threshold for electrical interaction with single neurons. Main results. We found that small temperature increases and decreases of 1 ^∘C caused consistent transient suppression and excitation of neurons during temperature change. Numerical modeling of the biophysics demonstrated that the Na-K pump, and to a lesser extent the Nernst potential, could account for these transient effects. Such effects are dependent upon compartmental ion fluxes and the rate of temperature change. Significance. A new bifurcation is described in the model dynamics that accounts for the transient suppression and excitation; in addition, we note the remarkable similarity of this bifurcation's rate dependency with other thermal rate-dependent tipping points in planetary warming dynamics. These experimental and theoretical findings demonstrate that stimulation of the brain must take into account small thermal effects that are ubiquitously present in electrical and magnetic stimulation. More sophisticated models of electrical current interaction with neurons combined with thermal effects will lead to more accurate modulation of neuronal activity.

Objective. Inferring the optimized and sparse network structure from the fully connected matrix is a key step in functional connectivity (FC) analysis. However, it is still an urgent problem to be solved, how to exclude the weak and spurious connections contained in functional networks objectively. Most existing binarization methods assume that the network has some certain constraint structures, which lead to changes in the original topology of the network. Approach. To solve this problem, we develop a Trade-off Model between Cost and Topology under Role Division (MCT), which consists of three crucial strategies, including modularity detection, definition of node role, and E-cost optimization algorithm. This algorithm weighs the physical cost and adaptive value of the network while preserving the network structure. Reliability and validity of MCT were evaluated by comparing different binarization methods (efficiency cost optimization, cluster-span threshold, threshold method, and MCT) on synthetic and real data sets. Main results. Experiment results demonstrated that the recovery rate of MCT for networks under noise interference is superior to other methods. In addition, brain networks filtered with MCT had higher network efficiency and shorter characteristic path length, which is more in line with the small world characteristics. Finally, applying MCT to resting-state electroencephalography data from patients with major depression reveals abnormal topology of the patients' connectivity networks, manifested as lower clustering coefficient (CC) and higher global efficiency (GE). Significance. This study provides an objective method for complex network analysis, which may contribute to the future of FC research.

Objective. Biomedical instrumentation and clinical systems for electrophysiology rely on electrodes and wires for sensing and transmission of bioelectric signals. However, this electronic approach constrains bandwidth, signal conditioning circuit designs, and the number of channels in invasive or miniature devices. This paper demonstrates an alternative approach using light to sense and transmit the electrophysiological signals. Approach. We develop a sensing, passive, fluorophore-free optrode based on the birefringence property of liquid crystals (LCs) operating at the microscale. Main results. We show that these optrodes can have the appropriate linearity (µ ± s.d.: 99.4 ± 0.5%, n = 11 devices), relative responsivity (µ ± s.d.: 57 ± 12%V⁻¹, n = 5 devices), and bandwidth (µ ± s.d.: 11.1 ± 0.7 kHz, n = 7 devices) for transducing electrophysiology signals into the optical domain. We report capture of rabbit cardiac sinoatrial electrograms and stimulus-evoked compound action potentials from the rabbit sciatic nerve. We also demonstrate miniaturisation potential by fabricating multi-optrode arrays, by developing a process that automatically matches each transducer element area with that of its corresponding biological interface. Significance. Our method of employing LCs to convert bioelectric signals into the optical domain will pave the way for the deployment of high-bandwidth optical telecommunications techniques in ultra-miniature clinical diagnostic and research laboratory neural and cardiac interfaces.

Objective. Electro/Magnetoencephalography (EEG/MEG) source-space network analysis is increasingly recognized as a powerful tool for tracking fast electrophysiological brain dynamics. However, an objective and quantitative evaluation of pipeline steps is challenging due to the lack of realistic 'controlled' data. Here, our aim is two-folded: (a) provide a quantitative assessment of the advantages and limitations of the analyzed techniques and (b) introduce (and share) a complete framework that can be used to optimize the entire pipeline of EEG/MEG source connectivity. Approach. We used a human brain computational model containing both physiologically based cellular GABAergic and Glutamatergic circuits coupled through Diffusion Tensor Imaging, to generate high-density EEG recordings. We designed a scenario of successive gamma-band oscillations in distinct cortical areas to emulate a virtual picture-naming task. We identified fast time-varying network states and quantified the performance of the key steps involved in the pipeline: (a) inverse models to reconstruct cortical-level sources, (b) functional connectivity measures to compute statistical interdependency between regional signals, and (c) dimensionality reduction methods to derive dominant brain network states (BNS). Main results. Using a systematic evaluation of the different decomposition techniques, results show significant variability among tested algorithms in terms of spatial and temporal accuracy. We outlined the spatial precision, the temporal sensitivity, and the global accuracy of the extracted BNS relative to each method. Our findings suggest a good performance of weighted minimum norm estimate/ Phase Locking Value combination to elucidate the appropriate functional networks and ICA techniques to derive relevant dynamic BNS. Significance. We suggest using such brain models to go further in the evaluation of the different steps and parameters involved in the EEG/MEG source-space network analysis. This can reduce the empirical selection of inverse model, connectivity measure, and dimensionality reduction method as some of the methods can have a considerable impact on the results and interpretation.

Objective. Early diagnosis of mild traumatic brain injury (mTBI) is challenging, yet crucial for providing patients with timely treatments and minimizing the risks of developing injury-related disorders. To tackle this problem, this paper presents a framework based on measures of frequency-specified brain functional networks identifying mTBI. Approach. Cortical activity of 15 control and 15 injury Thy1-GCaMP6s mice are recorded, using widefield calcium imaging, prior to and 20 minutes after inducing injury. Power spectral distribution (PSD) of the recorded cortical activities is examined, and the frequency bands with significant difference in PSD between the injury and control groups are identified. Frequency-specified functional networks are then constructed. Employing graph theoretical analysis, various network measures from the constructed frequency-specified functional networks are extracted and used as features. Several classifiers are utilized to evaluate the performance of the computed network measures, either individually or collectively as features, to classify mTBI from control. Main results. Spectral analysis reveals the presence of two dominant frequency bands (low: ${\lt}1$ Hz) and high: [1–8] Hz) in the cortical activities recorded via calcium imaging. Comparison of the brain networks of control and injury groups shows significant reduction (p < 0.05) in global functional connectivity following injury, specially for the high frequency band network. Interestingly, graph measures of the high frequency band network provided higher classification accuracy results, compared to those computed from the low frequency band network, suggesting that mTBI network-based features are frequency dependent. Using all network measures collectively as a multi-measure feature vector and a convolutional neural networks classifier, a model for identifying mTBI is developed, offering an average classification accuracy of 97.28%. Significance. Results signifies the importance of considering frequency-specific analysis in functional networks for mTBI identification, and demonstrate the possibility of using network measures for early mTBI diagnosis.

Objective. Autism spectrum disorder (ASD) is a prevalent neurodevelopmental disorder with the main symptoms of social communication disabilities. ASD is more than four times more common among males than females. The diagnosis of ASD is currently a subjective process by experts the same for males and females. Various studies have suggested the use of brain connectivity features for the diagnosis of ASD. Also, sex-related biological factors have been shown to play a role in ASD etiology and influence the brain connectivity. Therefore, proposing an accurate computer-aided diagnosis system (CADS) for ASD which considers the sex of subjects seems necessary. In this study, we present a sex-dependent connectivity-based CADS for ASD using resting-state functional magnetic resonance imaging. The proposed CADS classifies ASD males from normal males, and ASD females from normal females. Approach. After data preprocessing, group independent component analysis (GICA) was applied to obtain the resting-state networks (RSNs) followed by applying dual-regression to obtain the time course of each RSN for each subject. Afterwards, functional connectivity measures of full correlation and partial correlation and the effective connectivity measure of bivariate Granger causality were computed between time series of RSNs. To consider the role of sex differences in the classification process, male, female, and mixed groups were taken into account, and feature selection and classification were designed for each sex group separately. At the end, the classification accuracy was computed for each sex group. Main results. In the female group, a classification accuracy of 93.3% was obtained using full correlation while in the male group, a classification accuracy of 86.7% was achieved using both full correlation and bivariate Granger causality. Also, in the mixed group, a classification accuracy of 83.3% was obtained using full correlation. Significance. This supports the importance of considering sex in diagnosing ASD patients from normal controls.

Objective. Auditory attention decoding (AAD) determines which speaker the listener is focusing on by analyzing his/her EEG. Convolutional neural network (CNN) was adopted to extract spectro-spatial-feature (SSF) from short-time-interval of EEG to detect auditory spatial attention without stimuli. However, the following factors are not considered in SSF-CNN scheme. (a) Single-band frequency analysis cannot represent the EEG pattern precisely. (b) The power cannot represent the EEG feature related to the dynamic patterns of the attended auditory stimulus. (c) The temporal feature of EEG representing the relationship between EEG and attended stimulus is not extracted. To solve these problems, SSF-CNN scheme was modified. Approach. (a) Multiple-frequency bands, but not a single alpha frequency band, of EEG, were analyzed to represent the EEG pattern more precisely. (b) Differential entropy, but not power, was extracted from each frequency band to represent the disorder degree of EEG, which was related to the dynamic patterns of the attended auditory stimulus. (c) CNN and convolutional-long-short-term-memory (ConvLSTM) were combined to extract spectro-spatial-temporal features from the 3D descriptor sequence constructed based on the topographical activity maps of multiple-frequency bands. Main results. Experimental results on KUL, DTU, and PKU with 0.1 s, 1 s, 2 s, and 5 s decision windows demonstrated that: (a) The proposed model outperformed SSF-CNN and state-of-the-art AAD models. Specifically, when the auditory stimulus was unavailable, AAD accuracy could be enhanced by at least $3.25\%$, $3.96\%$ and $5.08\%$ on KUL, DTU, and PKU, respectively, compared with the baselines. And, on KUL, the longer decision window corresponded to lower enhancement, while on both DTU and PKU, the longer decision window corresponded to higher enhancement, except for two cases when decision window length was 2 s on PKU or 5 s on DTU. (b) Each modification contributed to the performance enhancement. Significance. DE feature, multi-band frequency analysis, and ConvLSTM-based temporal analysis help to enhance AAD accuracy.

Objective. Critical decisions are made by effective teams that are characterized by individuals who trust each other and know how to best integrate their opinions. Here, we introduce a multimodal brain-computer interface (BCI) to help collaborative teams of humans and an artificial agent achieve more accurate decisions in assessing danger zones during a pandemic scenario. Approach. Using high-resolution simultaneous electroencephalography/functional MRI (EEG/fMRI), we first disentangled the neural markers of decision-making confidence and trust and then employed machine-learning to decode these neural signatures for BCI-augmented team decision-making. We assessed the benefits of BCI on the team's decision-making process compared to the performance of teams of different sizes using the standard majority or weighing individual decisions. Main results. We showed that BCI-assisted teams are significantly more accurate in their decisions than traditional teams, as the BCI is capable of capturing distinct neural correlates of confidence on a trial-by-trial basis. Accuracy and subjective confidence in the context of collaborative BCI engaged parallel, spatially distributed, and temporally distinct neural circuits, with the former being focused on incorporating perceptual information processing and the latter involving action planning and executive operations during decision making. Among these, the superior parietal lobule emerged as a pivotal region that flexibly modulated its activity and engaged premotor, prefrontal, visual, and subcortical areas for shared spatial-temporal control of confidence and trust during decision-making. Significance. Multimodal, collaborative BCIs that assist human-artificial agent teams may be utilized in critical settings for augmented and optimized decision-making strategies.

Objective. Epidural electrical stimulation (EES) has emerged as an approach to restore motor function following spinal cord injury (SCI). However, identifying optimal EES parameters presents a significant challenge due to the complex and stochastic nature of muscle control and the combinatorial explosion of possible parameter configurations. Here, we describe a machine-learning approach that leverages modern deep neural networks to learn bidirectional mappings between the space of permissible EES parameters and target motor outputs. Approach. We collected data from four sheep implanted with two 24-contact EES electrode arrays on the lumbosacral spinal cord. Muscle activity was recorded from four bilateral hindlimb electromyography (EMG) sensors. We introduce a general learning framework to identify EES parameters capable of generating desired patterns of EMG activity. Specifically, we first amortize spinal sensorimotor computations in a forward neural network model that learns to predict motor outputs based on EES parameters. Then, we employ a second neural network as an inverse model, which reuses the amortized knowledge learned by the forward model to guide the selection of EES parameters. Main results. We found that neural networks can functionally approximate spinal sensorimotor computations by accurately predicting EMG outputs based on EES parameters. The generalization capability of the forward model critically benefited our inverse model. We successfully identified novel EES parameters, in under 20 min, capable of producing desired target EMG recruitment during in vivo testing. Furthermore, we discovered potential functional redundancies within the spinal sensorimotor networks by identifying unique EES parameters that result in similar motor outcomes. Together, these results suggest that our framework is well-suited to probe spinal circuitry and control muscle recruitment in a completely data-driven manner. Significance. We successfully identify novel EES parameters within minutes, capable of producing desired EMG recruitment. Our approach is data-driven, subject-agnostic, automated, and orders of magnitude faster than manual approaches.

Objective. Despite the tremendous promise of invasive brain-computer interfaces (iBCIs), the associated study costs, risks, and ethical considerations limit the opportunity to develop and test the algorithms that decode neural activity into a user's intentions. Our goal was to address this challenge by designing an iBCI model capable of testing many human subjects in closed-loop. Approach. We developed an iBCI model that uses artificial neural networks (ANNs) to translate human finger movements into realistic motor cortex firing patterns, which can then be decoded in real time. We call the model the joint angle BCI, or jaBCI. jaBCI allows readily recruited, healthy subjects to perform closed-loop iBCI tasks using any neural decoder, preserving subjects' control-relevant short-latency error correction and learning dynamics. Main results. We validated jaBCI offline through emulated neuron firing statistics, confirming that emulated neural signals have firing rates, low-dimensional PCA geometry, and rotational jPCA dynamics that are quite similar to the actual neurons (recorded in monkey M1) on which we trained the ANN. We also tested jaBCI in closed-loop experiments, our single study examining roughly as many subjects as have been tested world-wide with iBCIs (n = 25). Performance was consistent with that of the paralyzed, human iBCI users with implanted intracortical electrodes. jaBCI allowed us to imitate the experimental protocols (e.g. the same velocity Kalman filter decoder and center-out task) and compute the same seven behavioral measures used in three critical studies. Significance. These encouraging results suggest the jaBCI's real-time firing rate emulation is a useful means to provide statistically robust sample sizes for rapid prototyping and optimization of decoding algorithms, the study of bi-directional learning in iBCIs, and improving iBCI control.

Objective. Accurate decoding of surface electromyography (sEMG) is pivotal for muscle-to-machine-interfaces and their application e.g. rehabilitation therapy. sEMG signals have high inter-subject variability, due to various factors, including skin thickness, body fat percentage, and electrode placement. Deep learning algorithms require long training time and tend to overfit if only few samples are available. In this study, we aim to investigate methods to calibrate deep learning models to a new user when only a limited amount of training data is available. Approach. Two methods are commonly used in the literature, subject-specific modeling and transfer learning. In this study, we investigate the effectiveness of transfer learning using weight initialization for recalibration of two different pretrained deep learning models on new subjects data and compare their performance to subject-specific models. We evaluate two models on three publicly available databases (non invasive adaptive prosthetics database 2–4) and compare the performance of both calibration schemes in terms of accuracy, required training data, and calibration time. Main results. On average over all settings, our transfer learning approach improves 5%-points on the pretrained models without fine-tuning, and 12%-points on the subject-specific models, while being trained for 22% fewer epochs on average. Our results indicate that transfer learning enables faster learning on fewer training samples than user-specific models. Significance. To the best of our knowledge, this is the first comparison of subject-specific modeling and transfer learning. These approaches are ubiquitously used in the field of sEMG decoding. But the lack of comparative studies until now made it difficult for scientists to assess appropriate calibration schemes. Our results guide engineers evaluating similar use cases.

Objective. Motor-evoked potentials (MEPs) are among the most prominent responses to brain stimulation, such as supra-threshold transcranial magnetic stimulation and electrical stimulation. Understanding of the neurophysiology and the determination of the lowest stimulation strength that evokes responses requires the detection of even smaller responses, e.g. from single motor units. However, available detection and quantization methods suffer from a large noise floor. This paper develops a detection method that extracts MEPs hidden below the noise floor. With this method, we aim to estimate excitatory activations of the corticospinal pathways well below the conventional detection level. Approach. The presented MEP detection method presents a self-learning matched-filter approach for improved robustness against noise. The filter is adaptively generated per subject through iterative learning. For responses that are reliably detected by conventional detection, the new approach is fully compatible with established peak-to-peak readings and provides the same results but extends the dynamic range below the conventional noise floor. Main results. In contrast to the conventional peak-to-peak measure, the proposed method increases the signal-to-noise ratio by more than a factor of 5. The first detectable responses appear to be substantially lower than the conventional threshold definition of 50 µV median peak-to-peak amplitude. Significance. The proposed method shows that stimuli well below the conventional 50 µV threshold definition can consistently and repeatably evoke muscular responses and thus activate excitable neuron populations in the brain. As a consequence, the input–output (IO) curve is extended at the lower end, and the noise cut-off is shifted. Importantly, the IO curve extends so far that the 50 µV point turns out to be closer to the center of the logarithmic sigmoid curve rather than close to the first detectable responses. The underlying method is applicable to a wide range of evoked potentials and other biosignals, such as in electroencephalography.

Objective. With practice, the control of brain-computer interfaces (BCI) would improve over time; the neural correlate for such learning had not been well studied. We demonstrated here that monkeys controlling a motor BCI using a linear discriminant analysis (LDA) decoder could learn to make the firing patterns of the recorded neurons more distinct over a short period of time for different output classes to improve task performance. Approach. Using an LDA decoder, we studied two Macaque monkeys implanted with microelectrode arrays as they controlled the movement of a mobile robotic platform. The LDA decoder mapped high-dimensional neuronal firing patterns linearly onto a lower-dimensional linear discriminant (LD) space, and we studied the changes in the spatial coordinates of these neural signals in the LD space over time, and their correspondence to trial performance. Direction selectivity was quantified with permutation feature importance (FI). Main results. We observed that, within individual sessions, there was a tendency for the points in the LD space encoding different directions to diverge, leading to fewer misclassification errors, and, hence, improvement in task accuracy. Accuracy was correlated with the presence of channels with strong directional preference (i.e. high FI), as well as a varied population code (i.e. high variance in FI distribution). Significance. We emphasized the importance of studying the short-term/intra-sessional variations in neural representations during the use of BCI. Over the course of individual sessions, both monkeys could modulate their neural activities to create increasingly distinct neural representations.

Objective. The aim of the present study was to elucidate the brain dynamics underlying the maintenance of a constant force level exerted during a visually guided isometric contraction task by optimizing a predictive multivariate model based on global and spectral brain dynamics features. Approach. Electroencephalography (EEG) was acquired in 18 subjects who were asked to press a bulb and maintain a constant force level, indicated by a bar on a screen. For intervals of 500 ms, we calculated an index of force stability as well as indices of brain dynamics: microstate metrics (duration, occurrence, global explained variance, directional predominance) and EEG spectral amplitudes in the theta, low alpha, high alpha and beta bands. We optimized a multivariate regression model (partial least square (PLS)) where the microstate features and the spectral amplitudes were the input variables and the indexes of force stability were the output variables. The issues related to the collinearity among the input variables and to the generalizability of the model were addressed using PLS in a nested cross-validation approach. Main results. The optimized PLS regression model reached a good generalizability and succeeded to show the predictive value of microstates and spectral features in inferring the stability of the exerted force. Longer duration and higher occurrence of microstates, associated with visual and executive control networks, corresponded to better contraction performances, in agreement with the role played by the visual system and executive control network for visuo-motor integration. Significance. A combination of microstate metrics and brain rhythm amplitudes could be considered as biomarkers of a stable visually guided motor output not only at a group level, but also at an individual level. Our results may play an important role for a better understanding of the motor control in single trials or in real-time applications as well as in the study of motor control.

Objective. The aim of this work was to assess vascular remodeling after the placement of an endovascular neural interface (ENI) in the superior sagittal sinus (SSS) of sheep. We also assessed the efficacy of neural recording using an ENI. Approach. The study used histological analysis to assess the composition of the foreign body response. Micro-CT images were analyzed to assess the profiles of the foreign body response and create a model of a blood vessel. Computational fluid dynamic modeling was performed on a reconstructed blood vessel to evaluate the blood flow within the vessel. Recording of brain activity in sheep was used to evaluate efficacy of neural recordings. Main results. Histological analysis showed accumulated extracellular matrix material in and around the implanted ENI. The extracellular matrix contained numerous macrophages, foreign body giant cells, and new vascular channels lined by endothelium. Image analysis of CT slices demonstrated an uneven narrowing of the SSS lumen proportional to the stent material within the blood vessel. However, the foreign body response did not occlude blood flow. The ENI was able to record epileptiform spiking activity with distinct spike morphologies. Significance. This is the first study to show high-resolution tissue profiles, the histological response to an implanted ENI and blood flow dynamic modeling based on blood vessels implanted with an ENI. The results from this study can be used to guide surgical planning and future ENI designs; stent oversizing parameters to blood vessel diameter should be considered to minimize detrimental vascular remodeling.

Objective. Repetitive transcranial magnetic stimulation (rTMS) emerges as a useful therapy for autism spectrum disorder (ASD) clinically. Whereas the mechanisms of action of rTMS on ASD are not fully understood, and no biomarkers until now are available to reliably predict the follow-up rTMS efficacy in clinical practice. Approach. In the current work, the temporal variability was investigated in resting-state electroencephalogram of ASD patients, and the nonlinear complexity of related time-varying networks was accordingly evaluated by fuzzy entropy. Main results. The results showed the hyper-variability in the resting-state networks of ASD patients, while three week rTMS treatment alleviates the hyper fluctuations occurring in the frontal-parietal and frontal-occipital connectivity and further contributes to the ameliorative ASD symptoms. In addition, the changes in variability network properties are closely correlated with clinical scores, which further serve as potential predictors to reliably track the long-term rTMS efficacy for ASD. Significance. The findings consistently demonstrated that the temporal variability of time-varying networks of ASD patients could be modulated by rTMS, and related variability properties also help predict follow-up rTMS efficacy, which provides the potential for formulating individualized treatment strategies for ASD (ChiCTR2000033586).

Objective. Transcranial electrical stimulation (tES) is a promising method for modulating brain activity and excitability with variable results to date. To minimize electric (E-)field strength variability, we introduce the 2-sample prospective E-field dosing (2-SPED) approach, which uses E-field strengths induced by tES in a first population to individualize stimulation intensity in a second population. Approach. We performed E-field modeling of three common tES montages in 300 healthy younger adults. First, permutation analyses identified the sample size required to obtain a stable group average E-field in the primary motor cortex (M1), with stability being defined as the number of participants where all group-average E-field strengths ± standard deviation did not leave the population's 5–95 percentile range. Second, this stable group average was used to individualize tES intensity in a second independent population (n = 100). The impact of individualized versus fixed intensity tES on E-field strength variability was analyzed. Main results. In the first population, stable group average E-field strengths (V/m) in M1 were achieved at 74–85 participants, depending on the tES montage. Individualizing the stimulation intensity (mA) in the second population resulted in uniform M1 E-field strength (all p < 0.001) and significantly diminished peak cortical E-field strength variability (all p < 0.01), across all montages. Significance. 2-SPED is a feasible way to prospectively induce more uniform E-field strengths in a region of interest. Future studies might apply 2-SPED to investigate whether decreased E-field strength variability also results in decreased physiological and behavioral variability in response to tES.

Objective. Spinal cord injury (SCI) often results in debilitating movement impairments and neuropathic pain. Electrical stimulation of spinal neurons holds considerable promise both for enhancing neural transmission in weakened motor pathways and for reducing neural transmission in overactive nociceptive pathways. However, spinal stimulation paradigms currently under development for individuals living with SCI continue overwhelmingly to be developed in the context of motor rehabilitation alone. The objective of this study is to test the hypothesis that motor-targeted spinal stimulation simultaneously modulates spinal nociceptive transmission. Approach. We characterized the neuromodulatory actions of motor-targeted intraspinal microstimulation (ISMS) on the firing dynamics of large populations of discrete nociceptive specific and wide dynamic range (WDR) neurons. Neurons were accessed via dense microelectrode arrays implanted in vivo into lumbar enlargement of rats. Nociceptive and non-nociceptive cutaneous transmission was induced before, during, and after ISMS by mechanically probing the L5 dermatome. Main results. Our primary findings are that (a) sub-motor threshold ISMS delivered to spinal motor pools immediately modulates concurrent nociceptive transmission; (b) the magnitude of anti-nociceptive effects increases with longer durations of ISMS, including robust carryover effects; (c) the majority of all identified nociceptive-specific and WDR neurons exhibit firing rate reductions after only 10 min of ISMS; and (d) ISMS does not increase spinal responsiveness to non-nociceptive cutaneous transmission. These results lead to the conclusion that ISMS parameterized to enhance motor output results in an overall net decrease n spinal nociceptive transmission. Significance. These results suggest that ISMS may hold translational potential for neuropathic pain-related applications and that it may be uniquely suited to delivering multi-modal therapeutic benefits for individuals living with SCI.

Objective. Connectors for implantable neural prosthetic systems provide several advantages such as simplification of surgery, safe replacement of implanted devices, and modular design of the implant systems. With the rapid advancement of technologies for neural implants, miniaturized multichannel implantable connectors are also required. In this study, we propose a reconnectable and area-efficient multichannel implantable connector. Approach. A female-to-female adapter was fabricated using the thermal-press bonding of micropatterned liquid crystal polymer films. A bump inside the adapter enabled a reliable electrical connection by increasing the contact pressure between the contact pads of the adapter and the inserted cable. After connection, the adapter is enclosed in a metal case sealed with silicone elastomer packing. With different sizes of the packings, leakage current tests were performed under accelerated conditions to determine the optimal design for long-term reliability. Repeated connection tests were performed to verify the durability and reconnectability of the fabricated connector. The connector was implanted in rats, and the leakage currents were monitored to evaluate the stability of the connector in vivo. Main results. The fabricated four- and eight-channel implantable connectors, assembled with the metal cases, had a diameter and length of 6 and 17 mm, respectively. Further, the contact resistances of the four- and eight-channel connectors were 53.2 and 75.2 mΩ, respectively. The electrical contact remained stable during repeated connection tests (50 times). The fabricated connectors with packings having 125%, 137%, and 150% volume ratios to the internal space of the metal case failed after 14, 88, and 14 d, respectively, in a 75 °C saline environment. In animal tests with rats, the connector maintained low leakage current levels for up to 92 d. Significance. An implantable and reconnectable multichannel connector was developed and evaluated. The feasibility of the proposed connector was evaluated in terms of electrical and mechanical characteristics as well as sealing performance. The proposed connector is expected to have potential applications in implantable neural prosthetic systems.

Objective. Among the existing active brain–computer interfaces (BCI), the motor imagination (MI) is widely used. To operate the MI BCI effectively, subjects need to carry out trainings on corresponding imagining tasks. Here, we studied how to reduce the discomfort and fatigue of active BCI imaginary tasks and the inability to concentrate on them while improving the accuracy. Approach. This paper proposes a hybrid BCI composed of MI and pronunciation imagination (PI). The electroencephalogram signals of ten subjects are recognized by the adaptive Riemannian distance classification and the improved frequency selective filter-bank Common Spatial Pattern recognition. Main results. The results show that under the new paradigm with the combination of MI and PI, the recognition accuracy is higher than the MI alone. The highest recognition rate of the proposed hybrid system can reach more than 90%. Furthermore, through the subjects' scoring results of the operation difficulty, it is concluded that the designed hybrid paradigm is more operable than the traditional BCI paradigm. Significance. The separable tasks in the active BCI are limited and the accuracy needs to be improved. The new hybrid paradigm proposed by us improves the accuracy and operability of the active BCI system, providing a new possibility for the research direction of the active BCI.

Objective. Decomposition methods are efficient to decode steady-state visual evoked potentials (SSVEPs). In recent years, the brain–computer interface community has also been developing deep learning networks for decoding SSVEPs. However, there is no clear evidence that current deep learning models outperform decomposition methods on the SSVEP decoding tasks. Many studies lacked the comparison with state-of-the-art decomposition methods in a fair environment. Approach. This study proposed a novel network design motivated by the works of decomposition methods. Fixed template network (FTN) and dynamic template network (DTN) are two novel networks combining the advantages of fixed templates and subject-specific templates. This study also proposed a data augmentation method for SSVEPs. This study compared the intra-subject classification performance of DTN and FTN with that of state-of-the-art decomposition methods on three public SSVEP datasets. Main results. The results show that both FTN and DTN achieved the suboptimal classification performance compared with state-of-the-art decomposition methods. Significance. Both network designs could enhance the decoding performance of SSVEPs, making them promising networks for improving the practicality of SSVEP-based applications.

Objective. Critical slowing features (variance and autocorrelation) of long-term continuous electroencephalography (EEG) and electrocardiography (ECG) data have previously been used to forecast epileptic seizure onset. This study tested the feasibility of forecasting non-epileptic seizures using the same methods. In doing so, we examined if long-term cycles of brain and cardiac activity are present in clinical physiological recordings of psychogenic non-epileptic seizures (PNES). Approach. Retrospectively accessed ambulatory EEG and ECG data from 15 patients with non-epileptic seizures and no background of epilepsy were used for developing the forecasting system. The median period of recordings was 161 h, with a median of 7 non-epileptic seizures per patient. The phases of different cycles (5 min, 1 h, 6 h, 12 h, 24 h) of EEG and RR interval (RRI) critical slowing features were investigated. Forecasters were generated using combinations of the variance and autocorrelation of both EEG and the RRI of the ECG at each of the aforementioned cycle lengths. Optimal forecasters were selected as those with the highest area under the receiver-operator curve (AUC). Main results. It was found that PNES events occurred in the rising phases of EEG feature cycles of 12 and 24 h in duration at a rate significantly above chance. We demonstrated that the proposed forecasters achieved performance significantly better than chance in 8/15 of patients, and the mean AUC of the best forecaster across patients was 0.79. Significance. To our knowledge, this is the first study to retrospectively forecast non-epileptic seizures using both EEG and ECG data. The significance of EEG in the forecasting models suggests that cyclic EEG features of non-epileptic seizures exist. This study opens the potential of seizure forecasting beyond epilepsy, into other disorders of episodic loss of consciousness or dissociation.