Optimization of closed-loop electrical stimulation enables robust cerebellar-directed seizure control

Reinforcement learning for closed-loop regulation of cardiovascular system with vagus nerve stimulation: a computational study

Objective. Vagus nerve stimulation (VNS) is being investigated as a potential therapy for cardiovascular diseases including heart failure, cardiac arrhythmia, and hypertension. The lack of a systematic approach for controlling and tuning the VNS parameters poses a significant challenge. Closed-loop VNS strategies combined with artificial intelligence (AI) approaches offer a framework for systematically learning and adapting the optimal stimulation parameters. In this study, we presented an interactive AI framework using reinforcement learning (RL) for automated data-driven design of closed-loop VNS control systems in a computational study.Approach.Multiple simulation environments with a standard application programming interface were developed to facilitate the design and evaluation of the automated data-driven closed-loop VNS control systems. These environments simulate the hemodynamic response to multi-location VNS using biophysics-based computational models of healthy and hypertensive rat cardiovascular systems in resting and exercise states. We designed and implemented the RL-based closed-loop VNS control frameworks in the context of controlling the heart rate and the mean arterial pressure for a set point tracking task. Our experimental design included two approaches; a general policy using deep RL algorithms and a sample-efficient adaptive policy using probabilistic inference for learning and control.Main results.Our simulation results demonstrated the capabilities of the closed-loop RL-based approaches to learn optimal VNS control policies and to adapt to variations in the target set points and the underlying dynamics of the cardiovascular system. Our findings highlighted the trade-off between sample-efficiency and generalizability, providing insights for proper algorithm selection. Finally, we demonstrated that transfer learning improves the sample efficiency of deep RL algorithms allowing the development of more efficient and personalized closed-loop VNS systems.Significance.We demonstrated the capability of RL-based closed-loop VNS systems. Our approach provided a systematic adaptable framework for learning control strategies without requiring prior knowledge about the underlying dynamics.

Online Bayesian optimization of vagus nerve stimulation

Objective.In bioelectronic medicine, neuromodulation therapies induce neural signals to the brain or organs, modifying their function. Stimulation devices capable of triggering exogenous neural signals using electrical waveforms require a complex and multi-dimensional parameter space to control such waveforms. Determining the best combination of parameters (waveform optimization or dosing) for treating a particular patient's illness is therefore challenging. Comprehensive parameter searching for an optimal stimulation effect is often infeasible in a clinical setting due to the size of the parameter space. Restricting this space, however, may lead to suboptimal therapeutic results, reduced responder rates, and adverse effects.Approach. As an alternative to a full parameter search, we present a flexible machine learning, data acquisition, and processing framework for optimizing neural stimulation parameters, requiring as few steps as possible using Bayesian optimization. This optimization builds a model of the neural and physiological responses to stimulations, enabling it to optimize stimulation parameters and provide estimates of the accuracy of the response model. The vagus nerve (VN) innervates, among other thoracic and visceral organs, the heart, thus controlling heart rate (HR), making it an ideal candidate for demonstrating the effectiveness of our approach.Main results.The efficacy of our optimization approach was first evaluated on simulated neural responses, then applied to VN stimulation intraoperatively in porcine subjects. Optimization converged quickly on parameters achieving target HRs and optimizing neural B-fiber activations despite high intersubject variability.Significance.An optimized stimulation waveform was achieved in real time with far fewer stimulations than required by alternative optimization strategies, thus minimizing exposure to side effects. Uncertainty estimates helped avoiding stimulations outside a safe range. Our approach shows that a complex set of neural stimulation parameters can be optimized in real-time for a patient to achieve a personalized precision dosing.

Thalamic stimulation induced changes in effective connectivity

Deep brain stimulation (DBS) is a viable treatment for a variety of neurological conditions, however, the mechanisms through which DBS modulates large-scale brain networks are unresolved. Clinical effects of DBS are observed over multiple timescales. In some conditions, such as Parkinson's disease and essential tremor, clinical improvement is observed within seconds. In many other conditions, such as epilepsy, central pain, dystonia, neuropsychiatric conditions or Tourette syndrome, the DBS related effects are believed to require neuroplasticity or reorganization and often take hours to months to observe. To optimize DBS parameters, it is therefore essential to develop electrophysiological biomarkers that characterize whether DBS settings are successfully engaging and modulating the network involved in the disease of interest. In this study, 10 individuals with drug resistant epilepsy undergoing intracranial stereotactic EEG including a thalamus electrode underwent a trial of repetitive thalamic stimulation. We evaluated thalamocortical effective connectivity using single pulse electrical stimulation, both at baseline and following a 145 Hz stimulation treatment trial. We found that when high frequency stimulation was delivered for >1.5 hours, the evoked potentials measured from remote regions were significantly reduced in amplitude and the degree of modulation was proportional to the strength of baseline connectivity. When stimulation was delivered for shorter time periods, results were more variable. These findings suggest that changes in effective connectivity in the network targeted with DBS accumulate over hours of DBS. Stimulation evoked potentials provide an electrophysiological biomarker that allows for efficient data-driven characterization of neuromodulation effects, which could enable new objective approaches for individualized DBS optimization.

Intravital photoacoustic brain stimulation with high-precision

Significance

Neural regulation at high precision vitally contributes to propelling fundamental understanding in the field of neuroscience and providing innovative clinical treatment options. Recently, photoacoustic brain stimulation has emerged as a cutting-edge method for precise neuromodulation and shows great potential for clinical application.

Aim

The goal of this perspective is to outline the advancements in photoacoustic brain stimulation in recent years. And, we also provide an outlook delineating several prospective paths through which this burgeoning approach may be substantively refined for augmented capability and wider implementations.

Approach

First, the mechanisms of photoacoustic generation as well as the potential mechanisms of photoacoustic brain stimulation are provided and discussed. Then, the state-of-the-art achievements corresponding to this technology are reviewed. Finally, future directions for photoacoustic technology in neuromodulation are provided.

Results

Intensive research endeavors have prompted substantial advancements in photoacoustic brain stimulation, illuminating the unique advantages of this modality for noninvasive and high-precision neuromodulation via a nongenetic way. It is envisaged that further technology optimization and randomized prospective clinical trials will enable a wide acceptance of photoacoustic brain stimulation in clinical practice.

Conclusions

The innovative practice of photoacoustic technology serves as a multifaceted neuromodulation approach, possessing noninvasive, high-accuracy, and nongenetic characteristics. It has a great potential that could considerably enhance not only the fundamental underpinnings of neuroscience research but also its practical implementations in a clinical setting.

The bidirectional relationship between the cerebellum and seizure networks: a double-edged sword

Epilepsy is highly prevalent and notoriously pharmacoresistant. New therapeutic interventions are urgently needed, both for preventing the seizures themselves as well as negative outcomes and comorbidities associated with chronic epilepsy. While the cerebellum is not traditionally associated with epilepsy or seizures, research over the past decade has outlined the cerebellum as a brain region that is uniquely suited for both therapeutic needs. This review discusses our current understanding of the cerebellum as a key node within seizure networks, capable of both attenuating seizures in several animal models, and conversely, prone to altered structure and function in chronic epilepsy. Critical next steps are to advance therapeutic modulation of the cerebellum more towards translation, and to provide a more comprehensive characterization of how the cerebellum is impacted by chronic epilepsy, in order to subvert negative outcomes.

Optogenetic stimulation of the superior colliculus suppresses genetic absence seizures

While anti-seizure medications are effective for many patients, nearly one-third of individuals have seizures that are refractory to pharmacotherapy. Prior studies using evoked preclinical seizure models have shown that pharmacological activation or excitatory optogenetic stimulation of the deep and intermediate layers of the superior colliculus (DLSC) display multi-potent anti-seizure effects. Here we monitored and modulated DLSC activity to suppress spontaneous seizures in the WAG/Rij genetic model of absence epilepsy. Female and male WAG/Rij adult rats were employed as study subjects. For electrophysiology studies, we recorded single unit activity from microwire arrays placed within the DLSC. For optogenetic experiments, animals were injected with virus coding for channelrhodopsin-2 or a control vector, and we compared the efficacy of continuous neuromodulation to that of closed-loop neuromodulation paradigms. For each, we compared three stimulation frequencies on a within-subject basis (5, 20, 100 Hz). For closed-loop stimulation, we detected seizures in real time based on the EEG power within the characteristic frequency band of spike-and-wave discharges (SWDs). We quantified the number and duration of each SWD during each 2 h-observation period. Following completion of the experiment, virus expression and fibre-optic placement was confirmed. We found that single-unit activity within the DLSC decreased seconds prior to SWD onset and increased during and after seizures. Nearly 40% of neurons displayed suppression of firing in response to the start of SWDs. Continuous optogenetic stimulation of the DLSC (at each of the three frequencies) resulted in a significant reduction of SWDs in males and was without effect in females. In contrast, closed-loop neuromodulation was effective in both females and males at all three frequencies. These data demonstrate that activity within the DLSC is suppressed prior to SWD onset, increases at SWD onset, and that excitatory optogenetic stimulation of the DLSC exerts anti-seizure effects against absence seizures. The striking difference between open- and closed-loop neuromodulation approaches underscores the importance of the stimulation paradigm in determining therapeutic effects.

The cerebellum contributes to generalized seizures by altering activity in the ventral posteromedial nucleus

Thalamo-cortical networks are central to seizures, yet it is unclear how these circuits initiate seizures. We test whether a facial region of the thalamus, the ventral posteromedial nucleus (VPM), is a source of generalized, convulsive motor seizures and if convergent VPM input drives the behavior. To address this question, we devise an in vivo optogenetic mouse model to elicit convulsive motor seizures by driving these inputs and perform single-unit recordings during awake, convulsive seizures to define the local activity of thalamic neurons before, during, and after seizure onset. We find dynamic activity with biphasic properties, raising the possibility that heterogenous activity promotes seizures. Virus tracing identifies cerebellar and cerebral cortical afferents as robust contributors to the seizures. Of these inputs, only microinfusion of lidocaine into the cerebellar nuclei blocks seizure initiation. Our data reveal the VPM as a source of generalized convulsive seizures, with cerebellar input providing critical signals.

Hidden behavioral fingerprints in epilepsy

Epilepsy is a major disorder affecting millions of people. Although modern electrophysiological and imaging approaches provide high-resolution access to the multi-scale brain circuit malfunctions in epilepsy, our understanding of how behavior changes with epilepsy has remained rudimentary. As a result, screening for new therapies for children and adults with devastating epilepsies still relies on the inherently subjective, semi-quantitative assessment of a handful of pre-selected behavioral signs of epilepsy in animal models. Here, we use machine learning-assisted 3D video analysis to reveal hidden behavioral phenotypes in mice with acquired and genetic epilepsies and track their alterations during post-insult epileptogenesis and in response to anti-epileptic drugs. These results show the persistent reconfiguration of behavioral fingerprints in epilepsy and indicate that they can be employed for rapid, automated anti-epileptic drug testing at scale.

Autonomous optimization of neuroprosthetic stimulation parameters that drive the motor cortex and spinal cord outputs in rats and monkeys

Neural stimulation can alleviate paralysis and sensory deficits. Novel high-density neural interfaces can enable refined and multipronged neurostimulation interventions. To achieve this, it is essential to develop algorithmic frameworks capable of handling optimization in large parameter spaces. Here, we leveraged an algorithmic class, Gaussian-process (GP)-based Bayesian optimization (BO), to solve this problem. We show that GP-BO efficiently explores the neurostimulation space, outperforming other search strategies after testing only a fraction of the possible combinations. Through a series of real-time multi-dimensional neurostimulation experiments, we demonstrate optimization across diverse biological targets (brain, spinal cord), animal models (rats, non-human primates), in healthy subjects, and in neuroprosthetic intervention after injury, for both immediate and continual learning over multiple sessions. GP-BO can embed and improve "prior" expert/clinical knowledge to dramatically enhance its performance. These results advocate for broader establishment of learning agents as structural elements of neuroprosthetic design, enabling personalization and maximization of therapeutic effectiveness.

LINCs Are Vulnerable to Epileptic Insult and Fail to Provide Seizure Control via On-Demand Activation

Temporal lobe epilepsy (TLE) is notoriously pharmacoresistant, and identifying novel therapeutic targets for controlling seizures is crucial. Long-range inhibitory neuronal nitric oxide synthase-expressing cells (LINCs), a population of hippocampal neurons, were recently identified as a unique source of widespread inhibition in CA1, able to elicit both GABAA-mediated and GABAB-mediated postsynaptic inhibition. We therefore hypothesized that LINCs could be an effective target for seizure control. LINCs were optogenetically activated for on-demand seizure intervention in the intrahippocampal kainate (KA) mouse model of chronic TLE. Unexpectedly, LINC activation at 1 month post-KA did not substantially reduce seizure duration in either male or female mice. We tested two different sets of stimulation parameters, both previously found to be effective with on-demand optogenetic approaches, but neither was successful. Quantification of LINCs following intervention revealed a substantial reduction of LINC numbers compared with saline-injected controls. We also observed a decreased number of LINCs when the site of initial insult (i.e., KA injection) was moved to the amygdala [basolateral amygdala (BLA)-KA], and correspondingly, no effect of light delivery on BLA-KA seizures. This indicates that LINCs may be a vulnerable population in TLE, regardless of the site of initial insult. To determine whether long-term circuitry changes could influence outcomes, we continued testing once a month for up to 6 months post-KA. However, at no time point did LINC activation provide meaningful seizure suppression. Altogether, our results suggest that LINCs are not a promising target for seizure inhibition in TLE.

Optogenetic activation of the superior colliculus attenuates spontaneous seizures in the pilocarpine model of temporal lobe epilepsy

OBJECTIVE

Decades of studies have indicated that activation of the deep and intermediate layers of the superior colliculus can suppress seizures in a wide range of experimental models of epilepsy. However, prior studies have not examined efficacy against spontaneous limbic seizures. The present study aimed to address this gap through chronic optogenetic activation of the superior colliculus in the pilocarpine model of temporal lobe epilepsy.

METHODS

Sprague Dawley rats underwent pilocarpine-induced status epilepticus and were maintained until the onset of spontaneous seizures. Virus coding for channelrhodopsin-2 was injected into the deep and intermediate layers of the superior colliculus, and animals were implanted with head-mounted light-emitting diodes at the same site. Rats were stimulated with either 5- or 100-Hz light delivery. Seizure number, seizure duration, 24-h seizure burden, and behavioral seizure severity were monitored.

RESULTS

Both 5- and 100-Hz optogenetic stimulation of the deep and intermediate layers of the superior colliculus reduced daily seizure number and total seizure burden in all animals in the active vector group. Stimulation did not affect either seizure duration or behavioral seizure severity. Stimulation was without effect in opsin-negative control animals.

SIGNIFICANCE

Activation of the deep and intermediate layers of the superior colliculus reduces both the number of seizures and total daily seizure burden in the pilocarpine model of temporal lobe epilepsy. These novel data demonstrating an effect against chronic experimental seizures complement a long history of studies documenting the antiseizure efficacy of superior colliculus activation in a range of acute seizure models.

The little brain and the seahorse: Cerebellar-hippocampal interactions

There is a growing appreciation for the cerebellum beyond its role in motor function and accumulating evidence that the cerebellum and hippocampus interact across a range of brain states and behaviors. Acute and chronic manipulations, simultaneous recordings, and imaging studies together indicate coordinated coactivation and a bidirectional functional connectivity relevant for various physiological functions, including spatiotemporal processing. This bidirectional functional connectivity is likely supported by multiple circuit paths. It is also important in temporal lobe epilepsy: the cerebellum is impacted by seizures and epilepsy, and modulation of cerebellar circuitry can be an effective strategy to inhibit hippocampal seizures. This review highlights some of the recent key hippobellum literature.

Regulating absence seizures by tri-phase delay stimulation applied to globus pallidus internal

In this paper, a reduced globus pallidus internal (GPI)-corticothalamic (GCT) model is developed, and a tri-phase delay stimulation (TPDS) with sequentially applying three pulses on the GPI representing the inputs from the striatal D ₁ neurons, subthalamic nucleus (STN), and globus pallidus external (GPE), respectively, is proposed. The GPI is evidenced to control absence seizures characterized by 2 Hz-4 Hz spike and wave discharge (SWD). Hence, based on the basal ganglia-thalamocortical (BGCT) model, we firstly explore the triple effects of D _l-GPI, GPE-GPI, and STN-GPI pathways on seizure patterns. Then, using the GCT model, we apply the TPDS on the GPI to potentially investigate the alternative and improved approach if these pathways to the GPI are blocked. The results show that the striatum D ₁, GPE, and STN can indeed jointly and significantly affect seizure patterns. In particular, the TPDS can effectively reproduce the seizure pattern if the D ₁-GPI, GPE-GPI, and STN-GPI pathways are cut off. In addition, the seizure abatement can be obtained by well tuning the TPDS stimulation parameters. This implies that the TPDS can play the surrogate role similar to the modulation of basal ganglia, which hopefully can be helpful for the development of the brain-computer interface in the clinical application of epilepsy.

Automated deep brain stimulation programming with safety constraints for tremor suppression in patients with Parkinson's Disease and essential tremor

OBJECTIVES

Deep brain stimulation programming for movement disorders requires systematic fine tuning of stimulation parameters to ameliorate tremor and other symptoms while avoiding side effects. DBS programming can be a time-consuming process and requires clinical expertise to assess response to DBS to optimize therapy for each patient. In this study, we describe and evaluate an automated, closed-loop, and patient-specific framework for DBS programming that measures tremor using a smartwatch and automatically changes DBS parameters based on the recommendations from a closed-loop optimization algorithm thus eliminating the need for an expert clinician.

APPROACH

Bayesian optimization which is a sample-efficient global optimization method was used as the core of this DBS programming framework to adaptively learn each patient's response to DBS and suggest the next best settings to be evaluated. Input from a clinician was used initially to define a maximum safe amplitude, but we also implemented 'safe Bayesian optimization' to automatically discover tolerable exploration boundaries.

RESULTS

We tested the system in 15 patients (9 with Parkinson's disease and 6 with essential tremor). Tremor suppression at best automated settings was statistically comparable to previously established clinical settings. The optimization algorithm converged after testing 15.1±0.7 settings when maximum safe exploration boundaries were predefined, and 17.7±4.9 when the algorithm itself determined safe exploration boundaries.

SIGNIFICANCE

We demonstrate that fully automated DBS programming framework for treatment of tremor is efficient and safe while providing outcomes comparable to that achieved by expert clinicians.