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

Artificial Intelligence and Neuroscience: Transformative Synergies in Brain Research and Clinical Applications

The convergence of Artificial Intelligence (AI) and neuroscience is redefining our understanding of the brain, unlocking new possibilities in research, diagnosis, and therapy. This review explores how AI's cutting-edge algorithms-ranging from deep learning to neuromorphic computing-are revolutionizing neuroscience by enabling the analysis of complex neural datasets, from neuroimaging and electrophysiology to genomic profiling. These advancements are transforming the early detection of neurological disorders, enhancing brain-computer interfaces, and driving personalized medicine, paving the way for more precise and adaptive treatments. Beyond applications, neuroscience itself has inspired AI innovations, with neural architectures and brain-like processes shaping advances in learning algorithms and explainable models. This bidirectional exchange has fueled breakthroughs such as dynamic connectivity mapping, real-time neural decoding, and closed-loop brain-computer systems that adaptively respond to neural states. However, challenges persist, including issues of data integration, ethical considerations, and the "black-box" nature of many AI systems, underscoring the need for transparent, equitable, and interdisciplinary approaches. By synthesizing the latest breakthroughs and identifying future opportunities, this review charts a path forward for the integration of AI and neuroscience. From harnessing multimodal data to enabling cognitive augmentation, the fusion of these fields is not just transforming brain science, it is reimagining human potential. This partnership promises a future where the mysteries of the brain are unlocked, offering unprecedented advancements in healthcare, technology, and beyond.

Local field potential-based brain-machine interface to inhibit epileptic seizures by spinal cord electrical stimulation

Objective.This study proposes a closed-loop brain-machine interface (BMI) based on spinal cord stimulation to inhibit epileptic seizures, applying a semi-supervised machine learning approach that learns from Local Field Potential (LFP) patterns acquired on the pre-ictal (preceding the seizure) condition.Approach.LFP epochs from the hippocampus and motor cortex are band-pass filtered from 1 to 13 Hz, to obtain the time-frequency representation using the continuous Wavelet transform, and successively calculate the phase lock values (PLV). As a novelty, theZ-score-based PLV normalization using both modifiedk-means and Davies-Bouldin's measure for clustering is proposed here. Consequently, a generic seizure's detector is calibrated for detecting seizures on the normalized PLV, and enables the spinal cord stimulation for periods of 30 s in a closed-loop, while the BMI system detects seizure events. To calibrate the proposed BMI, a dataset with LFP signals recorded on five Wistar rats during basal state and epileptic crisis was used. The epileptic crisis was induced by injecting pentylenetetrazol (PTZ). Afterwards, two experiments without/with our BMI were carried out, inducing epileptic crisis by PTZ in Wistar rats.Main results.Stronger seizure events of high LFP amplitudes and long time periods were observed in the rat, when the BMI system was not used. In contrast, short-time seizure events of relative low intensity were observed in the rat, using the proposed BMI. The proposed system detected on unseen data the synchronized seizure activity in the hippocampus and motor cortex, provided stimulation appropriately, and consequently decreased seizure symptoms.Significance.Low-frequency LFP signals from the hippocampus and motor cortex, and cord spinal stimulation can be used to develop accurate closed-loop BMIs for early epileptic seizures inhibition, as an alternative treatment.

Neural decoding and feature selection methods for closed-loop control of avoidance behavior

Objective.Many psychiatric disorders involve excessive avoidant or defensive behavior, such as avoidance in anxiety and trauma disorders or defensive rituals in obsessive-compulsive disorders. Developing algorithms to predict these behaviors from local field potentials (LFPs) could serve as the foundational technology for closed-loop control of such disorders. A significant challenge is identifying the LFP features that encode these defensive behaviors.Approach.We analyzed LFP signals from the infralimbic cortex and basolateral amygdala of rats undergoing tone-shock conditioning and extinction, standard for investigating defensive behaviors. We utilized a comprehensive set of neuro-markers across spectral, temporal, and connectivity domains, employing SHapley Additive exPlanations for feature importance evaluation within Light Gradient-Boosting Machine models. Our goal was to decode three commonly studied avoidance/defensive behaviors: freezing, bar-press suppression, and motion (accelerometry), examining the impact of different features on decoding performance.Main results.Band power and band power ratio between channels emerged as optimal features across sessions. High-gamma (80-150 Hz) power, power ratios, and inter-regional correlations were more informative than other bands that are more classically linked to defensive behaviors. Focusing on highly informative features enhanced performance. Across 4 recording sessions with 16 subjects, we achieved an average coefficient of determination of 0.5357 and 0.3476, and Pearson correlation coefficients of 0.7579 and 0.6092 for accelerometry jerk and bar press rate, respectively. Utilizing only the most informative features revealed differential encoding between accelerometry and bar press rate, with the former primarily through local spectral power and the latter via inter-regional connectivity. Our methodology demonstrated remarkably low training/inference time and memory usage, requiring<310 ms for training,<0.051 ms for inference, and 16.6 kB of memory, using a single core of AMD Ryzen Threadripper PRO 5995WX CPU.Significance.Our results demonstrate the feasibility of accurately decoding defensive behaviors with minimal latency, using LFP features from neural circuits strongly linked to these behaviors. This methodology holds promise for real-time decoding to identify physiological targets in closed-loop psychiatric neuromodulation.

Active Neural Interface Circuits and Systems for Selective Control of Peripheral Nerves: A Review

Interfaces with peripheral nerves have been widely developed to enable bioelectronic control of neural activity. Peripheral nerve neuromodulation shows great potential in addressing motor dysfunctions, neurological disorders, and psychiatric conditions. The integration of high-density neural electrodes with stimulation and recording circuits poses a challenge in the design of neural interfaces. Recent advances in active electrode strategies have achieved improved reliability and performance by implementing in-situ control, stimulation, and recording of neural fibers. This paper presents an overview of state-of-the-art neural interface systems that comprise a range of neural electrodes, neurostimulators, and bio-amplifier circuits, with a special focus on interfaces for the peripheral nerves. A discussion on the efficacy of active electrode systems and recommendations for future directions conclude this paper.

A Reconfigurable, Nonlinear, Low-Power, VCO-Based ADC for Neural Recording Applications

Neural recording systems play a crucial role in comprehending the intricacies of the brain and advancing treatments for neurological disorders. Within these systems, the analog-to-digital converter (ADC) serves as a fundamental component, converting the electrical signals from the brain into digital data that can be further processed and analyzed by computing units. This research introduces a novel nonlinear ADC designed specifically for spike sorting in biomedical applications. Employing MOSFET varactors and voltage-controlled oscillators (VCOs), this ADC exploits the nonlinear capacitance properties of MOSFET varactors, achieving a parabolic quantization function that digitizes the noise with low resolution and the spikes with high resolution, effectively suppressing the background noise present in biomedical signals. This research aims to develop a reconfigurable, nonlinear voltage-controlled oscillator (VCO)-based ADC, specifically designed for implantable neural recording systems used in neuroprosthetics and brain-machine interfaces. The proposed design enhances the signal-to-noise ratio and reduces power consumption, making it more efficient for real-time neural data processing. By improving the performance and energy efficiency of these devices, the research contributes to the development of more reliable medical technologies for monitoring and treating neurological disorders. The quantization step of the ADC spans from 44.8 mV in the low-amplitude range to 1.4 mV in the high-amplitude range. The circuit was designed and simulated utilizing a 180 nm CMOS process; however, no physical prototype has been fabricated at this stage. Post-layout simulations confirm the expected performance. Occupying a silicon area is 0.09 mm2. Operating at a sampling frequency of 16 kS/s and a supply voltage of 1 volt, this ADC consumes 62.4 µW.

Neural Decoding and Feature Selection Techniques for Closed-Loop Control of Defensive Behavior

Objective

Many psychiatric disorders involve excessive avoidant or defensive behavior, such as avoidance in anxiety and trauma disorders or defensive rituals in obsessive-compulsive disorders. Developing algorithms to predict these behaviors from local field potentials (LFPs) could serve as foundational technology for closed-loop control of such disorders. A significant challenge is identifying the LFP features that encode these defensive behaviors.

Approach

We analyzed LFP signals from the infralimbic cortex and basolateral amygdala of rats undergoing tone-shock conditioning and extinction, standard for investigating defensive behaviors. We utilized a comprehensive set of neuro-markers across spectral, temporal, and connectivity domains, employing SHapley Additive exPlanations for feature importance evaluation within Light Gradient-Boosting Machine models. Our goal was to decode three commonly studied avoidance/defensive behaviors: freezing, bar-press suppression, and motion (accelerometry), examining the impact of different features on decoding performance.

Main results

Band power and band power ratio between channels emerged as optimal features across sessions. High-gamma (80-150 Hz) power, power ratios, and inter-regional correlations were more informative than other bands that are more classically linked to defensive behaviors. Focusing on highly informative features enhanced performance. Across 4 recording sessions with 16 subjects, we achieved an average coefficient of determination of 0.5357 and 0.3476, and Pearson correlation coefficients of 0.7579 and 0.6092 for accelerometry jerk and bar press rate, respectively. Utilizing only the most informative features revealed differential encoding between accelerometry and bar press rate, with the former primarily through local spectral power and the latter via inter-regional connectivity. Our methodology demonstrated remarkably low time complexity, requiring <110 ms for training and <1 ms for inference.

Significance

Our results demonstrate the feasibility of accurately decoding defensive behaviors with minimal latency, using LFP features from neural circuits strongly linked to these behaviors. This methodology holds promise for real-time decoding to identify physiological targets in closed-loop psychiatric neuromodulation.

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.

Innovative technologies for the fabrication of 3D/4D smart hydrogels and its biomedical applications - A comprehensive review

Repairing and regenerating damaged tissues or organs, and restoring their functioning has been the ultimate aim of medical innovations. 'Reviving healthcare' blends tissue engineering with alternative techniques such as hydrogels, which have emerged as vital tools in modern medicine. Additive manufacturing (AM) is a practical manufacturing revolution that uses building strategies like molding as a viable solution for precise hydrogel manufacturing. Recent advances in this technology have led to the successful manufacturing of hydrogels with enhanced reproducibility, accuracy, precision, and ease of fabrication. Hydrogels continue to metamorphose as the vital compatible bio-ink matrix for AM. AM hydrogels have paved the way for complex 3D/4D hydrogels that can be loaded with drugs or cells. Bio-mimicking 3D cell cultures designed via hydrogel-based AM is a groundbreaking in-vivo assessment tool in biomedical trials. This brief review focuses on preparations and applications of additively manufactured hydrogels in the biomedical spectrum, such as targeted drug delivery, 3D-cell culture, numerous regenerative strategies, biosensing, bioprinting, and cancer therapies. Prevalent AM techniques like extrusion, inkjet, digital light processing, and stereo-lithography have been explored with their setup and methodology to yield functional hydrogels. The perspectives, limitations, and the possible prospects of AM hydrogels have been critically examined in this study.

Advancements in non-invasive microwave brain stimulation: A comprehensive survey

This survey provides a comprehensive insight into the world of non-invasive brain stimulation and focuses on the evolving landscape of deep brain stimulation through microwave research. Non-invasive brain stimulation techniques provide new prospects for comprehending and treating neurological disorders. We investigate the methods shaping the future of deep brain stimulation, emphasizing the role of microwave technology in this transformative journey. Specifically, we explore antenna structures and optimization strategies to enhance the efficiency of high-frequency microwave stimulation. These advancements can potentially revolutionize the field by providing a safer and more precise means of modulating neural activity. Furthermore, we address the challenges that researchers currently face in the realm of microwave brain stimulation. From safety concerns to methodological intricacies, this survey outlines the barriers that must be overcome to fully unlock the potential of this technology. This survey serves as a roadmap for advancing research in microwave brain stimulation, pointing out potential directions and innovations that promise to reshape the field.

From basic sciences and engineering to epileptology: A translational approach

Collaborative efforts between basic scientists, engineers, and clinicians are enabling translational epileptology. In this article, we summarize the recent advances presented at the International Conference for Technology and Analysis of Seizures (ICTALS 2022): (1) novel developments of structural magnetic resonance imaging; (2) latest electroencephalography signal-processing applications; (3) big data for the development of clinical tools; (4) the emerging field of hyperdimensional computing; (5) the new generation of artificial intelligence (AI)-enabled neuroprostheses; and (6) the use of collaborative platforms to facilitate epilepsy research translation. We highlight the promise of AI reported in recent investigations and the need for multicenter data-sharing initiatives.

Microfabrication and lab-on-a-chip devices promotein vitromodeling of neural interfaces for neuroscience researches and preclinical applications

Neural tissues react to injuries through the orchestration of cellular reprogramming, generating specialized cells and activating gene expression that helps with tissue remodeling and homeostasis. Simplified biomimetic models are encouraged to amplify the physiological and morphological changes during neural regeneration at cellular and molecular levels. Recent years have witnessed growing interest in lab-on-a-chip technologies for the fabrication of neural interfaces. Neural system-on-a-chip devices are promisingin vitromicrophysiological platforms that replicate the key structural and functional characteristics of neural tissues. Microfluidics and microelectrode arrays are two fundamental techniques that are leveraged to address the need for microfabricated neural devices. In this review, we explore the innovative fabrication, mechano-physiological parameters, spatiotemporal control of neural cell cultures and chip-based neurogenesis. Although the high variability in different constructs, and the restriction in experimental and analytical access limit the real-life applications of microphysiological models, neural system-on-a-chip devices have gained considerable translatability for modeling neuropathies, drug screening and personalized therapy.

NET-TEN: a silicon neuromorphic network for low-latency detection of seizures in local field potentials

Objective. Therapeutic intervention in neurological disorders still relies heavily on pharmacological solutions, while the treatment of patients with drug resistance remains an unresolved issue. This is particularly true for patients with epilepsy, 30% of whom are refractory to medications. Implantable devices for chronic recording and electrical modulation of brain activity have proved a viable alternative in such cases. To operate, the device should detect the relevant electrographic biomarkers from local field potentials (LFPs) and determine the right time for stimulation. To enable timely interventions, the ideal device should attain biomarker detection with low latency while operating under low power consumption to prolong battery life.Approach. Here we introduce a fully-analog neuromorphic device implemented in CMOS technology for analyzing LFP signals in anin vitromodel of acute ictogenesis. Neuromorphic networks have progressively gained a reputation as low-latency low-power computing systems, which makes them a promising candidate as processing core of next-generation implantable neural interfaces.Main results. The developed system can detect ictal and interictal events with ms-latency and with high precision, consuming on average 3.50 nW during the task.Significance. The work presented in this paper paves the way to a new generation of brain implantable devices for personalized closed-loop stimulation for epilepsy treatment.

Trends in Volumetric-Energy Efficiency of Implantable Neurostimulators: A Review From a Circuits and Systems Perspective

This paper presents a comprehensive review of state-of-the-art, commercially available neurostimulators. We analyse key design parameters and performance metrics of 45 implantable medical devices across six neural target categories: deep brain, vagus nerve, spinal cord, phrenic nerve, sacral nerve and hypoglossal nerve. We then benchmark these alongside modern cardiac pacemaker devices that represent a more established market. This work studies trends in device size, electrode number, battery technology (i.e., primary and secondary use and chemistry), power consumption and longevity. This information is analysed to show the course of design decisions adopted by industry and identifying opportunity for further innovation. We identify fundamental limits in power consumption, longevity and size as well as the interdependencies and trade-offs. We propose a figure of merit to quantify volumetric efficiency within specific therapeutic targets, battery technologies/capacities, charging capabilities and electrode count. Finally, we compare commercially available implantable medical devices with recently developed systems in the research community. We envisage this analysis to aid circuit and system designers in system optimisation and identifying innovation opportunities, particularly those related to low power circuit design techniques.

NeuralTree: A 256-Channel 0.227-μJ/Class Versatile Neural Activity Classification and Closed-Loop Neuromodulation SoC

Closed-loop neural interfaces with on-chip machine learning can detect and suppress disease symptoms in neurological disorders or restore lost functions in paralyzed patients. While high-density neural recording can provide rich neural activity information for accurate disease-state detection, existing systems have low channel counts and poor scalability, which could limit their therapeutic efficacy. This work presents a highly scalable and versatile closed-loop neural interface SoC that can overcome these limitations. A 256-channel time-division multiplexed (TDM) front-end with a two-step fast-settling mixed-signal DC servo loop (DSL) is proposed to record high-spatial-resolution neural activity and perform channel-selective brain-state inference. A tree-structured neural network (NeuralTree) classification processor extracts a rich set of neural biomarkers in a patient- and disease-specific manner. Trained with an energy-aware learning algorithm, the NeuralTree classifier detects the symptoms of underlying disorders (e.g., epilepsy and movement disorders) at an optimal energy-accuracy trade-off. A 16-channel high-voltage (HV) compliant neurostimulator closes the therapeutic loop by delivering charge-balanced biphasic current pulses to the brain. The proposed SoC was fabricated in 65nm CMOS and achieved a 0.227μJ/class energy efficiency in a compact area of 0.014mm2/channel. The SoC was extensively verified on human electroencephalography (EEG) and intracranial EEG (iEEG) epilepsy datasets, obtaining 95.6%/94% sensitivity and 96.8%/96.9% specificity, respectively. In-vivo neural recordings using soft μECoG arrays and multi-domain biomarker extraction were further performed on a rat model of epilepsy. In addition, for the first time in literature, on-chip classification of rest-state tremor in Parkinson's disease (PD) from human local field potentials (LFPs) was demonstrated.

A lightweight learning-based decoding algorithm for intraneural vagus nerve activity classification in pigs

OBJECTIVE

Bioelectronic medicine is an emerging field that aims at developing closed-loop neuromodulation protocols for the autonomic nervous system (ANS) to treat a wide range of disorders. When designing a closed-loop protocol for real time modulation of the ANS, the computational execution time and the memory and power demands of the decoding step are important factors to consider. In the context of cardiovascular and respiratory diseases, these requirements may partially explain why closed-loop clinical neuromodulation protocols that adapt stimulation parameters on patient's clinical characteristics are currently missing.

APPROACH

Here, we developed a lightweight learning-based decoder for the classification of cardiovascular and respiratory functional challenges from neural signals acquired through intraneural electrodes implanted in the cervical vagus nerve (VN) of 5 anaesthetized pigs. Our algorithm is based on signal temporal windowing, 9 handcrafted features, and Random Forest (RF) model for classification. Temporal windowing ranging from 50 ms to 1 sec, compatible in duration with cardio-respiratory dynamics, was applied to the data in order to mimic a pseudo real-time scenario.

MAIN RESULTS

We were able to achieve high balanced accuracy (BA) values over the whole range of temporal windowing duration. We identified 500 ms as the optimal temporal windowing duration for both BA values and computational execution time processing, achieving more than 86% for BA and a computational execution time of only ∼6.8 ms. Our algorithm outperformed in terms of balanced accuracy and computational execution time a state of the art decoding algorithm tested on the same dataset [1]. We found that RF outperformed other machine learning models such as Support Vector Machines, K-Nearest Neighbors, and Multi-Layer Perceptrons.

SIGNIFICANCE

Our approach could represent an important step towards the implementation of a closed-loop neuromodulation protocol relying on a single intraneural interface able to perform real-time decoding tasks and selective modulation of the VN.

A fast and accurate learning-based decoding algorithm for the classification of cardiovascular and respiratory challenges using intraneural electrodes in the pig vagus nerve

Bioelectronic medicine is a new approach for developing closed-loop neuromodulation protocols on the peripheral nervous system (PNS) to treat a wide range of disorders currently treated with pharmacological approaches. Algorithms need to have low computational cost in order to acquire, process and model data for the modulation of the PNS in real time. Here, we present a fast learning-based decoding algorithm for the classification of cardiovascular and respiratory functional alterations (i.e., challenges) by using neural signals recorded from intraneural electrodes implanted in the vagus nerve of 5 pigs. Our algorithm relies on 9 handcrafted features, extracted following signal temporal windowing, and a multi-layer perceptron (MLP) for feature classification. We achieved fast and accurate classification of the challenges, with a computational time for feature extraction and prediction lower than 1.5 ms. The MLP achieved a balanced accuracy higher than 80 % for all recordings. Our algorithm could represent a step towards the development of a closed-loop system based on a single intraneural interface with both the potential of real time classification and selective modulation of the PNS.

A Power-Efficient Brain-Machine Interface System With a Sub-mw Feature Extraction and Decoding ASIC Demonstrated in Nonhuman Primates

Intracortical brain-machine interfaces have shown promise for restoring function to people with paralysis, but their translation to portable and implantable devices is hindered by their high power consumption. Recent devices have drastically reduced power consumption compared to standard experimental brain-machine interfaces, but still require wired or wireless connections to computing hardware for feature extraction and inference. Here, we introduce a Neural Recording And Decoding (NeuRAD) application specific integrated circuit (ASIC) in 180 nm CMOS that can extract neural spiking features and predict two-dimensional behaviors in real-time. To reduce amplifier and feature extraction power consumption, the NeuRAD has a hardware accelerator for extracting spiking band power (SBP) from intracortical spiking signals and includes an M0 processor with a fixed-point Matrix Acceleration Unit (MAU) for efficient and flexible decoding. We validated device functionality by recording SBP from a nonhuman primate implanted with a Utah microelectrode array and predicting the one- and two-dimensional finger movements the monkey was attempting to execute in closed-loop using a steady-state Kalman filter (SSKF). Using the NeuRAD's real-time predictions, the monkey achieved 100% success rate and 0.82 s mean target acquisition time to control one-dimensional finger movements using just 581 μW. To predict two-dimensional finger movements, the NeuRAD consumed 588 μW to enable the monkey to achieve a 96% success rate and 2.4 s mean acquisition time. By employing SBP, ASIC brain-machine interfaces can close the gap to enable fully implantable therapies for people with paralysis.