Closed-loop electrical neurostimulation: Challenges and opportunities

Predictive modeling of evoked intracranial EEG response to medial temporal lobe stimulation in patients with epilepsy

Despite promising advancements, closed-loop neurostimulation for drug-resistant epilepsy (DRE) still relies on manual tuning and produces variable outcomes, while automated predictable algorithms remain an aspiration. As a fundamental step towards addressing this gap, here we study predictive dynamical models of human intracranial EEG (iEEG) response under parametrically rich neurostimulation. Using data from n = 13 DRE patients, we find that stimulation-triggered switched-linear models with ~300 ms of causal historical dependence best explain evoked iEEG dynamics. These models are highly consistent across different stimulation amplitudes and frequencies, allowing for learning a generalizable model from abundant STIM OFF and limited STIM ON data. Further, evoked iEEG in nearly all subjects exhibited a distance-dependent pattern, whereby stimulation directly impacts the actuation site and nearby regions (≲ 20 mm), affects medium-distance regions (20 ~ 100 mm) through network interactions, and hardly reaches more distal areas (≳ 100 mm). Peak network interaction occurs at 60 ~ 80 mm from the stimulation site. Due to their predictive accuracy and mechanistic interpretability, these models hold significant potential for model-based seizure forecasting and closed-loop neurostimulation design.

Adaptive fuzzy iterative learning control based neurostimulation system and in-silico evaluation

Closed-loop neural stimulation has been an effective treatment for epilepsy patients. Currently, most closed-loop neural stimulation strategies are designed based on accurate neural models. However, the uncertainty and complexity of the neural system make it difficult to build an accurate neural model, which poses a significant challenge to the design of the controller. This paper proposes an Adaptive Fuzzy Iterative Learning Control (AFILC) framework for closed-loop neural stimulation, which can realize neuromodulation with no model or model uncertainty. Recognizing the periodic characteristics of neural stimulation and neuronal firing, Iterative Learning Control (ILC) is employed as the primary controller. Furthermore, a fuzzy optimization module is established to update the internal parameters of the ILC controller in real-time. This module enhances the anti-interference ability of the control system and reduces the influence of initial controller parameters on the control process. The efficacy of this strategy is evaluated using a neural computational model. The simulation results validate the capability of the AFILC strategy to suppress epileptic states. Compared with ILC-based closed-loop neurostimulation schemes, the AFILC-based neurostimulation strategy has faster convergence speed and stronger anti-interference ability. Moreover, the control algorithm is implemented based on a digital signal processor, and the hardware-in-the-loop experimental platform is implemented. The experimental results show that the control method has good control performance and computational efficiency, which provides the possibility for future application in clinical research.

Brain Feature Extraction With an Artifact-Tolerant Multiplexed Time-Encoding Neural Frontend for True Real-Time Closed-Loop Neuromodulation

Closed-loop neuromodulation is emerging as a more effective and targeted solution for the treatment of neurological symptoms compared to traditional open-loop stimulation. The majority of the present designs lack the ability to continuously record brain activity during electrical stimulation; hence they cannot fully monitor the treatment's effectiveness. This is due to the large stimulation artifacts that can saturate the sensitive readout circuits. To overcome this challenge, this work presents a rapid-artifact-recovery time-multiplexed neural readout frontend in combination with backend linear interpolation to reconstruct the artifact-corrupted local field potentials' (LFP) features. Our hybrid technique is an alternative approach to avoid power-hungry large-dynamic-range readout architectures or large and complex artifact template subtraction circuits. We discuss the design and measurements of a prototype implementation of the proposed readout frontend in 180-nm CMOS. It combines time multiplexing and time-domain conversion in a novel 13-bit incremental ADC, requiring only 0.0018 mm2/channel of readout area despite the large 180-nm CMOS process used, while consuming only 4.51 μW/channel. This is the smallest reported area for stimulation-voltage-compatible technologies (i.e. ≥ 65 nm). The frontend also yields a best-in-class peak total harmonic distortion of -72.6 dB @2.5-mVpp input, thanks to its implicit DAC mismatch-error shaping property. We employ the chip to measure brain LFP signals corrupted with artifacts, then perform linear interpolation and feature extraction on the measured signals and evaluate the reconstruction quality, using a set of sixteen commonly used features and three stimulation scenarios. The results show relative accuracies above 95% with respect to the situation without artifacts. This work is an ideal candidate for integration in high-channel-count true closed-loop neuromodulation systems.

Personalized strategies of neurostimulation: from static biomarkers to dynamic closed-loop assessment of neural function

Despite considerable advancement of first choice treatment (pharmacological, physical therapy, etc.) over many decades, neurological disorders still represent a major portion of the worldwide disease burden. Particularly concerning, the trend is that this scenario will worsen given an ever expanding and aging population. The many different methods of brain stimulation (electrical, magnetic, etc.) are, on the other hand, one of the most promising alternatives to mitigate the suffering of patients and families when conventional treatment fall short of delivering efficacious treatment. With applications in virtually all neurological conditions, neurostimulation has seen considerable success in providing relief of symptoms. On the other hand, a large variability of therapeutic outcomes has also been observed, particularly in the usage of non-invasive brain stimulation (NIBS) modalities. Borrowing inspiration and concepts from its pharmacological counterpart and empowered by unprecedented neurotechnological advancement, the neurostimulation field has seen in recent years a widespread of methods aimed at the personalization of its parameters, based on biomarkers of the individuals being treated. The rationale is that, by taking into account important factors influencing the outcome, personalized stimulation can yield a much-improved therapy. Here, we review the literature to delineate the state-of-the-art of personalized stimulation, while also considering the important aspects of the type of informing parameter (anatomy, function, hybrid), invasiveness, and level of development (pre-clinical experimentation versus clinical trials). Moreover, by reviewing relevant literature on closed loop neuroengineering solutions in general and on activity dependent stimulation method in particular, we put forward the idea that improved personalization may be achieved when the method is able to track in real time brain dynamics and adjust its stimulation parameters accordingly. We conclude that such approaches have great potential of promoting the recovery of lost functions and enhance the quality of life for patients.

Insights from a model based study on optimizing non invasive brain electrical stimulation for Parkinson's disease

Parkinson's Disease (PD) is a disorder in the central nervous system which includes symptoms such as tremor, rigidity, and Bradykinesia. Deep brain stimulation (DBS) is the most effective method to treat PD motor symptoms especially when the patient is not responsive to other treatments. However, its invasiveness and high risk, involving electrode implantation in the Basal Ganglia (BG), prompt recent research to emphasize non-invasive Transcranial Electrical Stimulation (TES). TES proves to be effective in treating some PD symptoms with inherent safety and no associated risks. This study explores the potential of using TES, to modify the firing pattern of cells in BG that are responsible for motor symptoms in PD. The research employs a mathematical model of the BG to examine the impact of applying TES to the brain. This is conducted using a realistic head model incorporating the Finite Element Method (FEM). According to our findings, the firing pattern associated with Parkinson's disease shifted towards a healthier firing pattern through the use of tACS. Employing an adaptive algorithm that continually monitored the behavior of BG cells (specifically, Globus Pallidus Pars externa (GPe)), we determined the optimal electrode number and placement to concentrate the current within the intended region. This resulted in a peak induced electric field of 1.9 v/m at the BG area. Our mathematical modeling together with precise finite element simulation of the brain and BG suggests that proposed method effectively mitigates Parkinsonian behavior in the BG cells. Furthermore, this approach ensures an improvement in the condition while adhering to all safety constraints associated with the current injection into the brain.

Resting-state EEG signatures of Alzheimer's disease are driven by periodic but not aperiodic changes

Electroencephalography (EEG) has shown potential for identifying early-stage biomarkers of neurocognitive dysfunction associated with dementia due to Alzheimer's disease (AD). A large body of evidence shows that, compared to healthy controls (HC), AD is associated with power increases in lower EEG frequencies (delta and theta) and decreases in higher frequencies (alpha and beta), together with slowing of the peak alpha frequency. However, the pathophysiological processes underlying these changes remain unclear. For instance, recent studies have shown that apparent shifts in EEG power from high to low frequencies can be driven either by frequency specific periodic power changes or rather by non-oscillatory (aperiodic) changes in the underlying 1/f slope of the power spectrum. Hence, to clarify the mechanism(s) underlying the EEG alterations associated with AD, it is necessary to account for both periodic and aperiodic characteristics of the EEG signal. Across two independent datasets, we examined whether resting-state EEG changes linked to AD reflect true oscillatory (periodic) changes, changes in the aperiodic (non-oscillatory) signal, or a combination of both. We found strong evidence that the alterations are purely periodic in nature, with decreases in oscillatory power at alpha and beta frequencies (AD < HC) leading to lower (alpha + beta) / (delta + theta) power ratios in AD. Aperiodic EEG features did not differ between AD and HC. By replicating the findings in two cohorts, we provide robust evidence for purely oscillatory pathophysiology in AD and against aperiodic EEG changes. We therefore clarify the alterations underlying the neural dynamics in AD and emphasize the robustness of oscillatory AD signatures, which may further be used as potential prognostic or interventional targets in future clinical investigations.

Challenges of neural interfaces for stroke motor rehabilitation

More than 85% of stroke survivors suffer from different degrees of disability for the rest of their lives. They will require support that can vary from occasional to full time assistance. These conditions are also associated to an enormous economic impact for their families and health care systems. Current rehabilitation treatments have limited efficacy and their long-term effect is controversial. Here we review different challenges related to the design and development of neural interfaces for rehabilitative purposes. We analyze current bibliographic evidence of the effect of neuro-feedback in functional motor rehabilitation of stroke patients. We highlight the potential of these systems to reconnect brain and muscles. We also describe all aspects that should be taken into account to restore motor control. Our aim with this work is to help researchers designing interfaces that demonstrate and validate neuromodulation strategies to enforce a contingent and functional neural linkage between the central and the peripheral nervous system. We thus give clues to design systems that can improve or/and re-activate neuroplastic mechanisms and open a new recovery window for stroke patients.

Delayed closed-loop neurostimulation for the treatment of pathological brain rhythms in mental disorders: a computational study

Mental disorders are among the top most demanding challenges in world-wide health. A large number of mental disorders exhibit pathological rhythms, which serve as the disorders characteristic biomarkers. These rhythms are the targets for neurostimulation techniques. Open-loop neurostimulation employs stimulation protocols, which are rather independent of the patients health and brain state in the moment of treatment. Most alternative closed-loop stimulation protocols consider real-time brain activity observations but appear as adaptive open-loop protocols, where e.g., pre-defined stimulation sets in if observations fulfil pre-defined criteria. The present theoretical work proposes a fully-adaptive closed-loop neurostimulation setup, that tunes the brain activities power spectral density (PSD) according to a user-defined PSD. The utilized brain model is non-parametric and estimated from the observations via magnitude fitting in a pre-stimulus setup phase. Moreover, the algorithm takes into account possible conduction delays in the feedback connection between observation and stimulation electrode. All involved features are illustrated on pathological α- and γ-rhythms known from psychosis. To this end, we simulate numerically a linear neural population brain model and a non-linear cortico-thalamic feedback loop model recently derived to explain brain activity in psychosis.

Resting-state EEG signatures of Alzheimer's disease are driven by periodic but not aperiodic changes

Electroencephalography (EEG) has shown potential for identifying early-stage biomarkers of neurocognitive dysfunction associated with dementia due to Alzheimer's disease (AD). A large body of evidence shows that, compared to healthy controls (HC), AD is associated with power increases in lower EEG frequencies (delta and theta) and decreases in higher frequencies (alpha and beta), together with slowing of the peak alpha frequency. However, the pathophysiological processes underlying these changes remain unclear. For instance, recent studies have shown that apparent shifts in EEG power from high to low frequencies can be driven either by frequency specific periodic power changes or rather by non-oscillatory (aperiodic) changes in the underlying 1/f slope of the power spectrum. Hence, to clarify the mechanism(s) underlying the EEG alterations associated with AD, it is necessary to account for both periodic and aperiodic characteristics of the EEG signal. Across two independent datasets, we examined whether resting-state EEG changes linked to AD reflect true oscillatory (periodic) changes, changes in the aperiodic (non-oscillatory) signal, or a combination of both. We found strong evidence that the alterations are purely periodic in nature, with decreases in oscillatory power at alpha and beta frequencies (AD < HC) leading to lower (alpha + beta) / (delta + theta) power ratios in AD. Aperiodic EEG features did not differ between AD and HC. By replicating the findings in two cohorts, we provide robust evidence for purely oscillatory pathophysiology in AD and against aperiodic EEG changes. We therefore clarify the alterations underlying the neural dynamics in AD and emphasise the robustness of oscillatory AD signatures, which may further be used as potential prognostic or interventional targets in future clinical investigations.

Optimization of phase prediction for brain-state dependent stimulation: a grid-search approach

Objective.Sources of heterogeneity in non-invasive brain stimulation literature can be numerous, with underlying brain states and protocol differences at the top of the list. Yet, incoherent results from brain-state-dependent stimulation experiments suggest that there are further factors adding to the variance. Hypothesizing that different signal processing pipelines might be partly responsible for heterogeneity; we investigated their effects on brain-state forecasting approaches.Approach.A grid-search was used to determine the fastest and most-accurate combination of preprocessing parameters and phase-forecasting algorithms. The grid-search was applied on a synthetic dataset and validated on electroencephalographic (EEG) data from a healthy (n= 18) and stroke (n= 31) cohort.Main results.Differences in processing pipelines led to different results; the grid-search chosen pipelines significantly increased the accuracy of published forecasting methods. The accuracy achieved in healthy was comparably high in stroke patients.Significance.This systematic offline analysis highlights the importance of the specific EEG processing and forecasting pipelines used for online state-dependent setups where precision in phase prediction is critical. Moreover, successful results in the stroke cohort pave the way to test state-dependent interventional treatment approaches.

Algorithm and hardware considerations for real-time neural signal on-implant processing

Objective Various on-workstation neural-spike-based brain machine interface(BMI) systems have reached the point of in-human trials, but on-node and on-implant BMI systems are still under exploration. Such systems are constrained by the area and battery. Researchers should consider the algorithm complexity, available resources, power budgets, CMOS technologies, and the choice of platforms when designing BMI systems. However, the effect of these factors is currently still unclear. Approaches. Here we have proposed a novel real-time 128 channel spike detection algorithm and optimised it on Microcontroller(MCU) and Field Programmable Gate Array(FPGA) platforms towards consuming minimal power and memory/resources. It is presented as a use case to explore the different considerations in system design. Main results. The proposed spike detection algorithm achieved over 97% sensitivity and a smaller than 3% false detection rate. The MCU implementation occupies less than 3KB RAM and consumes 31.5μW/ch. The FPGA platform only occupies 299 logic cells and 3KB RAM for 128 channels and consumes 0.04μW/ch. Significance. On the spike detection algorithm front, we have eliminated the processing bottleneck by reducing the dynamic power consumption to lower than the hardware static power, without sacrificing detection performance. More importantly, we have explored the considerations in algorithm and hardware design with respect to scalability, portability, and costs. These findings can facilitate and guide the future development of real-time on-implant neural signal processing platforms.

Soft Devices for High-Resolution Neuro-Stimulation: The Interplay Between Low-Rigidity and Resolution

The field of neurostimulation has evolved over the last few decades from a crude, low-resolution approach to a highly sophisticated methodology entailing the use of state-of-the-art technologies. Neurostimulation has been tested for a growing number of neurological applications, demonstrating great promise and attracting growing attention in both academia and industry. Despite tremendous progress, long-term stability of the implants, their large dimensions, their rigidity and the methods of their introduction and anchoring to sensitive neural tissue remain challenging. The purpose of this review is to provide a concise introduction to the field of high-resolution neurostimulation from a technological perspective and to focus on opportunities stemming from developments in materials sciences and engineering to reduce device rigidity while optimizing electrode small dimensions. We discuss how these factors may contribute to smaller, lighter, softer and higher electrode density devices.

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

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

A Mixed Filtering Approach for Real-Time Seizure State Tracking Using Multi-Channel Electroencephalography Data

Real-time continuous tracking of seizure state is necessary to develop feedback neuromodulation therapy that can prevent or terminate a seizure early. Due to its high temporal resolution, high scalp coverage, and non-invasive applicability, electroencephalography (EEG) is a good candidate for seizure tracking. In this research, we make multiple seizure state estimations using a mixed-filter and multiple channels found over the entire sensor space; then by applying a Kalman filter, we produce a single seizure state estimation made up of these individual estimations. Using a modified wrapper feature selection, we determine two optimal features of mixed data type, one continuous and one binary analyzing all available channels. These features are used in a state-space framework to model the continuous hidden seizure state. Expectation maximization is performed offline on the training and validation data sets to estimate unknown parameters. The seizure state estimation process is performed for multiple channels, and the seizure state estimation is derived using a square-root Kalman filter. A second expectation maximization step is utilized to estimate the unknown square-root Kalman filter parameters. This method is tested in a real-time applicable way for seizure state estimation. Applying this approach, we obtain a single seizure state estimation with quantitative information about the likelihood of a seizure occurring, which we call seizure probability. Our results on the experimental data (CHB-MIT EEG database) validate the proposed estimation method and we achieve an average accuracy, sensitivity, and specificity of 92.7%, 92.8%, and 93.4%, respectively. The potential applications of this seizure estimation model are for closed-loop neuromodulation and long-term quantitative analysis of seizure treatment efficacy.

Neurostimulation for Stroke Rehabilitation

Neurological injuries such as strokes can lead to important loss in motor function. Thanks to neuronal plasticity, some of the lost functionality may be recovered over time. However, the recovery process is often slow and incomplete, despite the most effective conventional rehabilitation therapies. As we improve our understanding of the rules governing activity-dependent plasticity, neuromodulation interventions are being developed to harness neural plasticity to achieve faster and more complete recovery. Here, we review the principles underlying stimulation-driven plasticity as well as the most commonly used stimulation techniques and approaches. We argue that increased spatiotemporal precision is an important factor to improve the efficacy of neurostimulation and drive a more useful neuronal reorganization. Consequently, closed-loop systems and optogenetic stimulation hold theoretical promise as interventions to promote brain repair after stroke.

Intensity and Dose of Neuromuscular Electrical Stimulation Influence Sensorimotor Cortical Excitability

Neuromuscular electrical stimulation (NMES) of the nervous system has been extensively used in neurorehabilitation due to its capacity to engage the muscle fibers, improving muscle tone, and the neural pathways, sending afferent volleys toward the brain. Although different neuroimaging tools suggested the capability of NMES to regulate the excitability of sensorimotor cortex and corticospinal circuits, how the intensity and dose of NMES can neuromodulate the brain oscillatory activity measured with electroencephalography (EEG) is still unknown to date. We quantified the effect of NMES parameters on brain oscillatory activity of 12 healthy participants who underwent stimulation of wrist extensors during rest. Three different NMES intensities were included, two below and one above the individual motor threshold, fixing the stimulation frequency to 35 Hz and the pulse width to 300 μs. Firstly, we efficiently removed stimulation artifacts from the EEG recordings. Secondly, we analyzed the effect of amplitude and dose on the sensorimotor oscillatory activity. On the one hand, we observed a significant NMES intensity-dependent modulation of brain activity, demonstrating the direct effect of afferent receptor recruitment. On the other hand, we described a significant NMES intensity-dependent dose-effect on sensorimotor activity modulation over time, with below-motor-threshold intensities causing cortical inhibition and above-motor-threshold intensities causing cortical facilitation. Our results highlight the relevance of intensity and dose of NMES, and show that these parameters can influence the recruitment of the sensorimotor pathways from the muscle to the brain, which should be carefully considered for the design of novel neuromodulation interventions based on NMES.

The effects of direct brain stimulation in humans depend on frequency, amplitude, and white-matter proximity

BACKGROUND

Researchers have used direct electrical brain stimulation to treat a range of neurological and psychiatric disorders. However, for brain stimulation to be maximally effective, clinicians and researchers should optimize stimulation parameters according to desired outcomes.

OBJECTIVE

The goal of our large-scale study was to comprehensively evaluate the effects of stimulation at different parameters and locations on neuronal activity across the human brain.

METHODS

To examine how different kinds of stimulation affect human brain activity, we compared the changes in neuronal activity that resulted from stimulation at a range of frequencies, amplitudes, and locations with direct human brain recordings. We recorded human brain activity directly with electrodes that were implanted in widespread regions across 106 neurosurgical epilepsy patients while systematically stimulating across a range of parameters and locations.

RESULTS

Overall, stimulation most often had an inhibitory effect on neuronal activity, consistent with earlier work. When stimulation excited neuronal activity, it most often occurred from high-frequency stimulation. These effects were modulated by the location of the stimulating electrode, with stimulation sites near white matter more likely to cause excitation and sites near gray matter more likely to inhibit neuronal activity.

CONCLUSION

By characterizing how different stimulation parameters produced specific neuronal activity patterns on a large scale, our results provide an electrophysiological framework that clinicians and researchers may consider when designing stimulation protocols to cause precisely targeted changes in human brain activity.

A Miniature Dual-Biomarker-Based Sensing and Conditioning Device for Closed-Loop DBS

In this paper, a dual-biomarker-based neural sensing and conditioning device is proposed for closing the feedback loop in deep brain stimulation devices. The device explores both local field potentials (LFPs) and action potentials (APs) as measured biomarkers. It includes two channels, each having four main parts: (1) a pre-amplifier with built-in low-pass filter, (2) a ground shifting circuit, (3) an amplifier with low-pass function, and (4) a high-pass filter. The design specifications include miniature-size, light-weight, and 100 dB gain in the LFP and AP channels. This device has been validated through bench and in-vitro tests. The bench tests have been performed using different sinusoidal signals and pre-recorded neural signals. The in-vitro tests have been conducted in the saline solution that mimics the brain environment. The total weight of the device including a 3 V coin battery, and battery holder is 1.2 g. The diameter of the device is 11.2 mm. The device can be used to concurrently sense LFPs and APs for closing the feedback loop in closed-loop deep brain stimulation systems. It provides a tetherless head-mountable platform suitable for pre-clinical trials.

Differential contributions of subthalamic beta rhythms and 1/f broadband activity to motor symptoms in Parkinson's disease

Excessive beta oscillatory activity in the subthalamic nucleus (STN) is linked to Parkinson's Disease (PD) motor symptoms. However, previous works have been inconsistent regarding the functional role of beta activity in untreated Parkinsonian states, questioning such role. We hypothesized that this inconsistency is due to the influence of electrophysiological broadband activity -a neurophysiological indicator of synaptic excitation/inhibition ratio- that could confound measurements of beta activity in STN recordings. Here we propose a data-driven, automatic and individualized mathematical model that disentangles beta activity and 1/f broadband activity in the STN power spectrum, and investigate the link between these individual components and motor symptoms in thirteen Parkinsonian patients. We show, using both modeled and actual data, how beta oscillatory activity significantly correlates with motor symptoms (bradykinesia and rigidity) only when broadband activity is not considered in the biomarker estimations, providing solid evidence that oscillatory beta activity does correlate with motor symptoms in untreated PD states as well as the significant impact of broadband activity. These findings emphasize the importance of data-driven models and the identification of better biomarkers for characterizing symptom severity and closed-loop applications.