System for controlling neurostimulation in a closed control loop based on real-time electroencephalographic pattern recognition

Abstract

A closed-loop system for regulating neurostimulation based on real-time electroencephalographic pattern recognition, wherein the system includes: one or more electroencephalographic signal acquisition units configured to be surgically coupled to a scalp or intracranial electrode array to continuously record bioelectrical brain signals of a subject; an analog input unit electrically coupled to the electroencephalographic signal acquisition units, wherein the analog input unit comprises an amplification circuit, a filter circuit and an analog-to-digital conversion circuit configured to convert acquired electroencephalographic signals into digitized signal streams; a signal processing processor that is operationally coupled with the analog input stage and is configured to perform real-time noise reduction, base correction, motion artifact removal, and band-specific filtering on the digitized signal streams; a feature extraction processor configured to receive processed electroencephalographic signals and compute temporal, spectral and spatial features, including waveform morphology descriptors, frequency band power distributions, phase relationships and interchannel coherence metrics; a pattern recognition processor configured to classify the extracted features into predefined or adaptive neurophysiological states using trained computational models and to generate state identification outputs corresponding to the detected neuronal states; a control processor operationally coupled to the pattern recognition processor, configured to determine the neurostimulation control parameters based on the identified neuronal states, wherein the control processor dynamically adjusts the stimulation amplitude, pulse width, frequency and waveform according to a control guideline; a neurostimulation unit that is electrically connected to one or more stimulation electrodes and is configured to deliver controlled electrical stimulation signals to the target tissue depending on the defined neurostimulation control parameters; and a feedback synchronization unit configured to synchronize the acquired electroencephalographic signals with the delivered stimulation signals and to update the control processor based on real-time feedback for a continuous closed control loop.

Description

Technical field of the invention

[0001] The present disclosure relates generally to biomedical engineering, neurotechnology, and intelligent therapy systems, but in particular to a device-centric system for controlling neurostimulation by means of real-time EEG pattern recognition. The invention integrates signal acquisition hardware, embedded processing circuits, adaptive control units, and stimulation structures configured to dynamically modulate neuronal activity in response to recognized electrophysiological patterns. Background of the invention

[0002] Neurostimulation systems are increasingly used to treat neurological disorders such as epilepsy, Parkinson's disease, depression, and chronic pain. Conventional neurostimulation devices predominantly operate in an open-loop system, where electrical stimulation is delivered according to predefined parameters without adapting to the patient's neuronal state in real time. Such systems do not account for dynamic fluctuations in brain activity, leading to suboptimal therapeutic efficacy, increased energy consumption, and potential side effects from overstimulation or faulty stimulation.

[0003] Electroencephalography (EEG) enables the non-invasive or minimally invasive acquisition of neuronal activity patterns with high temporal resolution. However, existing systems that incorporate EEG signals into neurostimulation control often rely on offline analysis or delayed processing, limiting their responsiveness. Furthermore, many systems lack robust pattern recognition capabilities capable of distinguishing pathological neuronal signatures from normal physiological fluctuations in real time. The absence of integrated architectures that combine high-resolution EEG data acquisition, real-time signal processing, adaptive decision-making, and precise stimulation delivery remains a significant limitation.

[0004] Furthermore, state-of-the-art systems often employ rigid control techniques that do not adapt to patient-specific neural profiles or changing disease states. These systems also exhibit latency problems, inadequate noise suppression, and inefficient data fusion strategies when processing multichannel EEG signals. Therefore, there is a need for a structurally integrated device that enables continuous EEG monitoring, real-time pattern recognition, and adaptive neurostimulation control with high precision and reliability.

[0005] Neurostimulation has advanced significantly in recent decades due to the increasing prevalence of neurological disorders such as epilepsy, Parkinson's disease, essential tremor, depression, and chronic pain syndromes. Conventional neurostimulation systems, including deep brain stimulation (DBS), vagus nerve stimulation (VNS), and transcranial electrical stimulation (tES), have demonstrated clinical efficacy in modulating neural circuits to alleviate symptoms. However, most of these systems operate in an open-loop system, where stimulation parameters such as amplitude, frequency, and pulse width are pre-programmed and delivered continuously or intermittently without considering the patient's physiological state in real time.This lack of adaptability limits the therapeutic precision of such systems, since neuronal activity is inherently dynamic and changes over time, with behavioral state and disease progression.

[0006] Open neurostimulation systems are typically configured empirically during clinical examination. Clinicians adjust the parameters by trial and error to achieve symptom relief. Once set, these parameters remain unchanged until the next adjustment, which may occur weeks or months later. This approach does not account for transient pathological events such as the onset of an epileptic seizure or fluctuating motor symptoms in Parkinson's disease. Therefore, stimulation may be unnecessarily administered, resulting in wasted energy and potential side effects, or it may be omitted at critical moments when intervention is needed. Furthermore, continuous stimulation can lead to neuronal adaptation or habituation, reducing long-term effectiveness and necessitating further parameter adjustments.

[0007] Furthermore, closed neurostimulation systems have been developed that use physiological signals as feedback to dynamically adjust the stimulation dose. Early approaches of this type were based on relatively simple biomarkers such as local field potentials (LFPs) or threshold-based detection of signal amplitude changes. Although these systems represent an improvement over open systems, they are often limited by the specificity and reliability of the chosen biomarkers. For example, threshold-based detection methods are highly sensitive to noise and can produce false positives or false negatives, thus compromising the accuracy of stimulation control. Moreover, these systems often lack the necessary computing power to interpret complex neuronal dynamics, leading to suboptimal results under heterogeneous clinical conditions.

[0008] Electroencephalography (EEG) has established itself as a promising method for real-time monitoring of brain activity due to its high temporal resolution, non-invasive nature, and comparatively low cost. EEG signals capture a broad spectrum of neuronal oscillations and transient events that can serve as indicators of pathological or functional brain states. However, integrating EEG into closed neurostimulation systems presents several technical challenges. EEG signals are inherently noisy and prone to artifacts caused by muscle activity, eye movements, environmental influences, and electrode impedance fluctuations. Conventional signal processing techniques such as bandpass filtering and simple spectral analysis are often insufficient to reliably isolate relevant neuronal patterns from this noise.

[0009] Existing EEG-based neurostimulation systems often use offline analysis or near real-time processing, resulting in significant latency in data acquisition and analysis. This delay compromises the fundamental goal of control, which requires rapid detection of and response to neuronal events. Furthermore, many systems employ manually created features and rule-based procedures that are not universally applicable across different patients and clinical conditions. These approaches often fail to capture the nonlinear and non-stationary characteristics of EEG signals, leading to lower detection accuracy and unreliable control decisions.

[0010] Recent advances in machine learning and artificial intelligence have enabled more complex approaches to EEG pattern recognition. Techniques such as support vector machines, artificial neural networks, and deep learning architectures are employed to classify neural states and detect pathological events. While these methods offer greater accuracy, their implementation in real-time neurostimulation systems is limited by computational complexity, power consumption, and latency requirements. Many machine learning models require extensive training datasets and computational resources that are not readily available in embedded or implantable devices. Furthermore, the lack of interpretability of complex models poses a challenge for clinical validation and regulatory approval.

[0011] Another limitation of existing systems lies in their lack of personalization and adaptability. Neural signatures associated with specific diseases can vary considerably between individuals and change over time due to disease progression, medication effects, or neuronal plasticity. Most current systems lack mechanisms for continuous learning or adaptation, resulting in static models whose effectiveness may decline over time. Furthermore, the absence of robust feedback mechanisms to assess the effect of delivered stimulation limits these systems' ability to optimize their control strategies.

[0012] Hardware limitations play a crucial role in the performance of current neurostimulation systems. High-resolution EEG data acquisition requires multiple channels, high sampling rates, and low-noise amplification, all of which lead to increased power consumption and more complex devices. Implantable systems, in particular, must adhere to strict power and size constraints, which complicates the integration of advanced signal processing and machine learning capabilities. Furthermore, ensuring reliable and stable electrode function over extended periods remains a significant challenge, as electrode wear and tissue changes can degrade signal quality.

[0013] Latency is another crucial factor influencing the effectiveness of closed-loop neurostimulation. The time delay between signal acquisition, processing, decision-making, and stimulation delivery must be minimized to ensure timely intervention. Many existing systems suffer from processing delays due to inefficient techniques, data bottlenecks, or hardware limitations. Such delays can render the system ineffective in rapidly changing conditions such as epileptic seizures, where milliseconds can be critical.

[0014] Safety and robustness are also of paramount importance in the development of closed neurostimulation systems. Inaccurate detection or inappropriate stimulation can lead to adverse effects, including tissue damage, exacerbation of seizures, or unintended modulation of neural circuits. Existing systems often lack comprehensive safety mechanisms for validating control decisions prior to stimulation. Furthermore, the variability and unpredictability of EEG signals complicate the establishment of reliable thresholds and control strategies that ensure consistent performance under varying operating conditions.

[0015] Another drawback of current solutions is the insufficient integration of sensors, data processing, and stimulation. Many systems consist of individual modules that communicate via external interfaces. This leads to increased latency, higher energy consumption, and reduced reliability. The lack of tightly coupled architectures prevents smooth data flow and real-time responsiveness for an effective control loop. Furthermore, such modular designs complicate the miniaturization of devices and limit their use in wearable or implantable formats.

[0016] Furthermore, data management and communication present significant challenges. Continuous EEG monitoring generates large volumes of data that must be processed, stored, or transmitted. Existing systems often rely on external computing resources or cloud-based processing, leading to additional latency and raising concerns about data privacy and security. Wireless communication modules, on the other hand, can be resource-intensive and prone to interference, further impacting system performance.

[0017] The clinical application of advanced neurostimulation systems is hampered by regulatory and application-related challenges. Complex systems with numerous configurable parameters require extensive validation and training for medical professionals. The lack of standardized protocols for EEG-based control further complicates clinical implementation. Patient comfort and compliance also play a crucial role, particularly with wearable systems where electrode placement and device ergonomics are critical.

[0018] Given the aforementioned limitations, there remains an urgent need for an integrated system that combines high-resolution EEG data acquisition, robust real-time signal processing, advanced pattern recognition, and adaptive stimulation control in a unified device architecture. Such a system should operate with low latency, high accuracy, and energy efficiency, while also providing mechanisms for personalization, continuous adaptation, and safety assurance. Developing such a system would represent a significant advancement over existing solutions and would enable more effective and responsive neurostimulation therapies tailored to individual patient needs. Summary of the invention

[0019] The present disclosure describes a system in the form of a medical device, consisting of a housing that encloses an EEG acquisition unit, a signal conditioning circuit, a real-time processing unit, a pattern recognition processor, a decision control unit, and a neurostimulation delivery unit. The system is configured to continuously acquire multi-channel EEG signals, process them to extract relevant features, identify predefined or adaptive neural patterns, and generate stimulation control signals in a closed-loop system to modulate neural activity.

[0020] The device features a compact hardware architecture whose EEG acquisition unit comprises multiple electrodes spatially corresponding to the target brain regions. The signal processing circuitry includes amplification stages, analog filter components, and analog-to-digital converters for generating digitized EEG data streams. The real-time processing unit performs signal preprocessing operations such as artifact removal, baseline correction, and normalization.

[0021] The pattern recognition processor is configured to extract features using time, frequency, and time-frequency analysis, and then classify them using trained models such as support vector machines, convolutional neural networks, or adaptive thresholding mechanisms. The decision control unit generates stimulation commands based on the detected patterns and integrates feedback loops that adjust stimulation parameters such as amplitude, frequency, pulse width, and timing.

[0022] The neurostimulation unit consists of implantable or surface electrodes coupled to a stimulation signal generator, enabling the delivery of controlled electrical impulses to the target tissue. The system operates in a closed-loop control system, where continuous feedback from EEG signals dynamically influences the stimulation, thereby increasing therapeutic precision and reducing unnecessary stimulation.

[0023] The main objective of the present invention is a system for controlling neurostimulation by means of real-time EEG pattern recognition. A structurally integrated device continuously monitors neuronal activity and dynamically modulates the stimulation parameters depending on the recognized electrophysiological patterns, thereby improving therapeutic precision and efficacy. A further objective of the invention is a device for acquiring high-resolution multi-channel EEG signals using multiple electrodes and associated signal processing circuits. This ensures the precise acquisition of neuronal dynamics while simultaneously minimizing noise, artifacts, and interference that could impair signal quality.

[0024] A further objective of the invention is to provide a real-time processing architecture integrated into the device, in which digitized EEG signals undergo preprocessing steps such as filtering, normalization, and artifact suppression. This enables reliable subsequent analysis without significant latency. Another objective of the invention is the integration of an advanced pattern recognition processor configured to extract meaningful features from EEG signals across time, frequency, and time-frequency domains and to classify neuronal states using adaptive or trained models. This enables the precise, real-time detection of pathological or targeted neuronal patterns.

[0025] A further aim of the invention is to provide a decision control unit that generates stimulation control signals based on identified EEG patterns. The control unit adaptively adjusts stimulation parameters such as amplitude, frequency, pulse width, and timing to the current neuronal state and previous response patterns. Furthermore, the invention aims to ensure a true closed-loop control system through continuous feedback of EEG signals after stimulation. This serves to evaluate the effectiveness of the delivered stimulation and to optimize subsequent control decisions, thereby enabling continuous improvement of therapy outcomes.

[0026] A further aim of the invention is to provide a neurostimulation unit integrated into the device structure, consisting of a programmable pulse generator and stimulation electrodes for delivering controlled electrical impulses to targeted neuronal regions with high temporal precision and safety. A further aim is to minimize the latency between signal acquisition, processing, and stimulation delivery through a tightly coupled hardware and processing architecture, thereby enabling timely intervention in rapidly developing neurological events such as seizures or abnormal oscillatory activity.

[0027] A further objective of the invention is to provide a customizable system whose pattern recognition and control techniques can be adapted to patient-specific neural profiles and can evolve over time to account for changes in physiological state, disease progression, or therapeutic response. Furthermore, it is an objective of the invention to integrate safety mechanisms into the control unit to prevent excessive or inappropriate stimulation by applying predefined thresholds and validation criteria before stimulation delivery.

[0028] A further objective of the invention is to provide an energy-efficient device for long-term operation, including implantable or wearable versions. This is achieved through the use of energy management circuits and optimized processing techniques to reduce energy consumption without compromising performance. Another objective is the seamless integration of sensor, processing, and stimulation components into a compact housing, thereby improving reliability, reducing system complexity, and facilitating practical use in clinical and home environments.

[0029] A further objective of the invention is to provide communication and data processing functions within the device to enable external monitoring, parameter configuration, and data logging while maintaining data integrity and security. Finally, an overarching objective of the invention is to overcome the limitations of conventional open and existing closed neurostimulation systems by providing a technologically advanced, adaptive, and responsive device that utilizes real-time EEG pattern recognition to enable targeted and efficient neuromodulation therapy. BRIEF DESCRIPTION OF THE IMAGE

[0030] These and other features, aspects and advantages of the present invention will be better understood if the following detailed description is read with reference to the accompanying drawing, in which the same symbols represent the same parts: Fig. Figure 1 shows a block diagram of a system for controlling neurostimulation using real-time electroencephalography.

[0031] Furthermore, those skilled in the art will recognize that the elements in the drawing are simplified and not necessarily drawn to scale. For example, the flowcharts illustrate the process by highlighting the main steps to facilitate understanding of the present disclosure. With regard to the construction of the device, one or more components may be represented in the drawing by conventional symbols. The drawing may show only the specific details relevant to understanding the embodiments of the present disclosure, so as not to clutter the drawing with details that are already apparent to those skilled in the art from the description contained herein. Detailed description of the invention

[0032] To facilitate understanding of the principles of the invention, reference is made below to the embodiment shown in the drawing, which is described using specific terms. It is understood, however, that this does not limit the scope of protection of the invention. Rather, modifications and further developments of the depicted system, as well as further applications of the inventive principles shown therein, are conceivable, insofar as they would normally occur to a person skilled in the art in the field of the invention.

[0033] It will be clear to those skilled in the art that the foregoing general description and the following detailed description are exemplary and explanatory of the invention and are not to be understood as a limitation of it.

[0034] References to “an aspect”, “another aspect”, or similar phrases in this description mean that a particular feature, structure, or property described in connection with the embodiment is included in at least one embodiment of the present disclosure. Therefore, phrases such as “in one embodiment”, “in another embodiment”, and similar expressions in this description may, but do not necessarily, all refer to the same embodiment.

[0035] The terms "includes," "comprehensive," or similar expressions denote non-exclusive inclusion. Thus, a procedure or method containing a list of steps does not only include those steps but may also include further steps not explicitly listed or inherent in the procedure or method. Likewise, the statement "includes..." for one or more devices, subsystems, elements, structures, or components, without further limitations, does not preclude the existence of other devices, subsystems, elements, structures, or components.

[0036] Unless otherwise defined, all technical and scientific terms used herein have the same meanings generally known to those skilled in the art in the field to which this invention belongs. The systems, methods, and examples described herein serve only for illustration and are not to be understood as limiting.

[0037] Embodiments of the present disclosure are described in detail below with reference to the attached drawing.

[0038] Fig.Figure 1 shows the block diagram of a system for controlling neurostimulation using real-time EEG pattern recognition. The system 100 comprises: one or more EEG signal acquisition units (102) connected to a scalp or intracranial electrode array for the continuous acquisition of bioelectrical brain signals from a subject; an analog input unit (104) electrically connected to the EEG signal acquisition units, comprising amplification, filtering, and analog-to-digital converter circuits for converting the acquired EEG signals into digitized signal streams;a signal conditioning processor (106) connected to the analog input unit, which performs real-time noise reduction, base correction, motion artifact removal, and band-specific filtering of the digitized signal streams; a feature extraction processor (108) configured to receive processed electroencephalographic signals and compute temporal, spectral, and spatial features, including waveform morphology descriptors, frequency-band power distributions, phase relationships, and interchannel coherence metrics; a pattern recognition processor (110) configured to classify the extracted features into predefined or adaptive neurophysiological states using trained computational models and to generate state identification outputs corresponding to the recognized neuronal states;a control processor (112) that is operationally coupled to the pattern recognition processor and configured to determine neurostimulation control parameters based on the identified neuronal states, wherein the control processor dynamically adjusts stimulation amplitude, pulse width, frequency, and waveform according to a control policy; a neurostimulation delivery unit (114) that is electrically coupled to one or more stimulation electrodes and configured to deliver controlled electrical stimulation signals to the target tissue in response to the determined neurostimulation control parameters; and a feedback synchronization unit (116) that is configured to temporally synchronize the acquired electroencephalographic signals with the delivered stimulation signals and update the control processor based on real-time feedback for a continuous closed-loop control system.

[0039] In one embodiment, the signal conditioning processor (106) further comprises an adaptive filter circuit configured to suppress mains interference, electromyographic interference and electrode impedance fluctuations by means of continuously updated filter coefficients derived from real-time signal statistics.

[0040] In one embodiment, the feature extraction processor (108) is configured to compute multi-resolution time-frequency representations using window transformations and to generate feature vectors corresponding to transient neuronal events such as epileptiform discharges, oscillatory bursts, and phase-locked responses.

[0041] In one embodiment, the pattern recognition processor (110) comprises a hierarchical classification arrangement with a first processor for anomaly detection and a second processor for multi-class classification of neuronal states, thereby enabling stepwise decision-making for triggering neurostimulation.

[0042] In one embodiment, the pattern recognition processor (110) is further configured to update classification parameters by incremental learning based on newly acquired electroencephalographic data and corresponding stimulation results stored in a memory unit.

[0043] In one embodiment, the control processor (112) is configured to implement a predictive control strategy that estimates future neuronal state trajectories based on historical electroencephalographic patterns and adjusts the stimulation parameters before the occurrence of undesired neuronal events.

[0044] In one embodiment, the neurostimulation unit (114) comprises a programmable pulse generation circuit configured to generate biphasic or monophasic stimulation pulses with charge balancing to minimize tissue damage.

[0045] In one embodiment, the feedback synchronization unit (116) is configured to compensate for the latency of signal acquisition and the delay of stimulation delivery by applying time correction using time-stamped data buffers.

[0046] In an embodiment further comprising a safety monitoring processor configured to continuously evaluate stimulation parameters and physiological responses to ensure compliance with predefined safety thresholds and to interrupt stimulation upon detection of abnormal or unsafe conditions.

[0047] In one embodiment, the electroencephalographic signal acquisition units (102) consist of high-density electrode arrays arranged to capture spatially distributed neuronal activity across multiple cortical regions, thereby enabling spatial pattern recognition.

[0048] The described system is implemented using concrete, physical hardware elements that operate with measurable electrical signals rather than abstract constructs. The electroencephalographic (EEG) signal acquisition units consist of physical electrodes and conductive interfaces that directly capture the bioelectrical activity of nerve tissue. The analog input stage is implemented using discrete and integrated electronic circuits, including amplifiers, analog filters, and analog-to-digital converters, which convert low-amplitude biological signals into usable electrical data streams. Subsequent processing stages are performed by dedicated hardware processors, such as digital signal processing circuits and embedded processing units, configured for deterministic operations like noise reduction, filtering, and feature computation using fixed electronic architectures.Pattern recognition is performed by physically instantiated computing circuits that apply preconfigured mathematical transformations to incoming signals, while the control processor is implemented in a hardware-based controller that generates electrical control signals to regulate the stimulation parameters. The neurostimulation unit includes pulse generation circuits, current drivers, and electrode interfaces that physically deliver controlled electrical impulses to the nerve tissue. Additionally, synchronization and safety monitoring functions are implemented through timing circuits, buffers, and monitoring hardware that ensure correct timing alignment and enforce electrical and physiological limits.

[0049] The system for controlling neurostimulation using real-time EEG pattern recognition operates with a tightly integrated sequence of signal acquisition, preprocessing, feature extraction, classification, decision-making, and stimulation delivery. Each stage is implemented in a uniform device architecture to ensure minimal latency and high reliability. The EEG signal acquisition unit continuously records multichannel neural signals from a subject using multiple electrodes arranged in a predefined spatial topology corresponding to the target regions of the cortex or subcortical bone. The acquired signals, typically in the microvolt range, are first amplified using low-noise amplifier circuits to preserve signal quality while suppressing common-mode interference.The amplified signals then pass through the signal processing unit, which applies analog filtering to remove unwanted frequency components, including baseline deviation and high-frequency interference. The signals are then converted into digitized signal streams using high-resolution analog-to-digital converters operating at a sampling rate sufficient for capturing fast neural oscillations.

[0050] The digitized signal streams are fed to the real-time processing unit, where a series of preprocessing techniques are performed to improve signal quality and prepare the data for pattern recognition. Preprocessing begins with adaptive artifact suppression. This dynamically identifies and attenuates artifacts caused by eye movements, muscle activity, and external electrical interference. This is achieved by combining reference-based subtraction methods with adaptive filtering techniques that continuously update the filter coefficients based on signal statistics. After artifact removal, baseline correction is performed to eliminate slowly changing deviations and ensure the signal is centered around a stable reference level.Subsequently, normalization is applied to standardize the signal amplitude across all channels, thus enabling consistent feature extraction regardless of impedance variations of the electrodes or differences between the channels.

[0051] The preprocessed signals are segmented into time windows using a buffering procedure. These windows may overlap to ensure temporal continuity and capture transient neural events. Within each time segment, the pattern recognition processor performs multi-stage feature extraction to capture the complex and non-stationary properties of electroencephalographic signals. Time-domain features are calculated by analyzing signal amplitude statistics, including mean, variance, and higher-order moments, which provide information about the signal distribution and transient fluctuations. Frequency-domain features are derived through spectral analysis, decomposing the signal into its frequency components and quantifying the power distribution across different frequency bands.Time-frequency features are obtained using decomposition methods that simultaneously capture temporal and spectral variations, thus enabling the identification of transient oscillation patterns that may indicate pathological or functional neuronal states.

[0052] In addition to these primary functions, the system computes advanced descriptors such as entropy-based measures to quantify signal complexity, synchronization indices to assess phase relationships between channels, and coherence metrics to evaluate the functional connectivity of different brain regions. These features are aggregated into a multidimensional feature vector that represents the current neural state within each time window. The pattern recognition processor then applies a classification procedure to the extracted feature vectors. This procedure is configured using a trained computational model that can incorporate subject-specific calibration data.The classification process compares the feature vector with learned decision boundaries or probabilistic representations to determine whether the observed neural activity corresponds to a predefined pattern of interest, such as a pathological event or a specific cognitive or physiological condition.

[0053] The classification results are transmitted to the decision control unit, which interprets the identified neuronal pattern and determines the appropriate stimulation response. The decision control unit uses an adaptive control procedure that maps detected neuronal states to stimulation parameters such as amplitude, pulse width, frequency, and timing. This mapping is not static but is dynamically adjusted based on historical response data and continuous feedback from the electroencephalography signals. The control procedure includes a feedback mechanism that evaluates the effects of previously delivered stimulations by analyzing subsequent neuronal activity, thus enabling iterative optimization of the stimulation parameters to achieve optimal therapeutic outcomes.

[0054] Before issuing stimulation commands, the control unit performs a safety validation to ensure that the proposed stimulation parameters are within predefined safety limits. This includes checking whether the stimulation amplitude and frequency exceed permissible limits and whether the timing of the stimulation conflicts with existing neuronal activity and could therefore cause undesirable effects. After successful validation, the control signals are transmitted to the neurostimulation unit. This unit consists of a programmable pulse generation circuit that produces electrical waveforms with precise temporal and amplitude characteristics. The generated pulses are delivered to the target tissue via stimulation electrodes positioned according to the therapeutic goal.

[0055] Following stimulation, the EEG signal acquisition unit continues to monitor neuronal activity, recording both the immediate and delayed effects of the stimulation. This data is processed as previously described, and the resulting information is fed back to the decision control unit. The system's closed-loop control ensures that each cycle of signal acquisition, processing, classification, and stimulation contributes to a continuously evolving control strategy that adapts to the subject's neuronal dynamics.

[0056] The system is implemented to prioritize computational efficiency and low latency, thus enabling real-time operation within the limitations of embedded or implantable devices. Data buffering, parallel processing, and optimized signal processing routines minimize delays between signal acquisition and stimulation delivery. Integrated power management strategies ensure continuous operation, particularly in implantable configurations with limited power resources. The overall architecture enables seamless and continuous interaction between sensor and stimulation components, providing a robust and adaptive platform for closed-loop neurostimulation based on real-time EEG pattern recognition.

[0057] The system is implemented as an integrated device, consisting of a rigid or semi-flexible housing that contains electronic circuits, power supplies, and communication interfaces. The EEG acquisition unit comprises several biocompatible electrodes that are positioned on the scalp, subcutaneously, or intracranially, depending on the application. Each electrode is electrically coupled to a low-noise preamplifier that captures EEG signals in the microvolt range while minimizing interference.

[0058] The captured signals are routed to the signal processing circuit, which includes cascaded amplifier stages and bandpass filters for isolating relevant frequency bands such as delta, theta, alpha, beta, and gamma bands. The processed analog signals are then digitized using high-resolution analog-to-digital converters operating at sampling rates that ensure temporal accuracy.

[0059] The digitized EEG data are processed by the real-time processing unit. This unit consists of a microprocessor or digital signal processor configured to perform preprocessing techniques. These techniques include adaptive filters for removing motion artifacts, suppressing mains hum, and normalization procedures for stabilizing signal amplitude fluctuations. The processed signals are segmented into time windows for feature extraction.

[0060] The pattern recognition processor calculates features such as spectral power density, wavelet coefficients, entropy measures, phase synchronization indices, and coherence metrics. These features are fed into classification models trained to recognize specific neural patterns such as epileptic seizures, abnormal oscillatory activity, or transitions between cognitive states. The processor can utilize pre-trained machine learning models that are subsequently adapted using patient-specific calibration data.

[0061] Upon detection of a target pattern of neuronal activity, the decision control unit generates a control signal that determines the stimulation parameters. The control logic incorporates adaptive mechanisms that adjust stimulation intensity and timing based on previous responses and the current neuronal state. The decision control unit may also include safety constraints to prevent overstimulation and ensure adherence to predefined therapeutic limits.

[0062] The neurostimulation unit includes a pulse generator capable of producing electrical waveforms with programmable properties. The stimulation electrodes are configured to deliver these waveforms to specific brain regions. The system ensures synchronization between recorded EEG patterns and stimulation to achieve optimal therapeutic results.

[0063] The device also includes a feedback monitoring pathway in which EEG signals are analyzed after stimulation to evaluate the effectiveness of the delivered stimulation. This feedback serves to optimize subsequent control decisions, thus forming a continuous adaptive control loop. The system can also include wireless communication circuits for external monitoring, parameter adjustment, and data logging.

[0064] The system's structural integration ensures minimal latency between signal acquisition, processing, and stimulation delivery. Designed for low-energy operation, the device features energy-efficient components and a power management circuit that enables long-term operation, particularly in implantable configurations.

[0065] The invention thus provides a technically advanced, device-oriented closed-loop neurostimulation system capable of real-time EEG pattern recognition and adaptive therapeutic intervention, overcoming the limitations of conventional open-loop and semi-closed-loop systems.

[0066] The present invention relates to the field of biomedical engineering and neurotechnology, in particular a device-oriented system for controlling neurostimulation by means of real-time EEG pattern recognition. The invention integrates the acquisition of EEG signals, real-time signal processing, adaptive pattern recognition, and controlled neurostimulation in a single device that dynamically modulates neuronal activity in response to continuously monitored electrophysiological signals.

[0067] The drawing and the preceding description illustrate embodiments. Those skilled in the art will recognize that one or more of the described elements can be combined to form a single functional element. Alternatively, certain elements can be divided into several functional elements. Elements of one embodiment can be added to another. For example, the process flows described here can be modified and are not limited to the manner described herein. Furthermore, the actions of a flowchart need not be performed in the sequence shown; nor do all actions necessarily need to be carried out. Actions that do not depend on other actions can be performed in parallel with the other actions. The scope of protection of the embodiments is in no way limited by these specific examples. Numerous variations, whether explicitly stated in the description or not, such as...Differences in structure, dimensions, and materials are possible. The scope of protection of the embodiments is at least as comprehensive as described by the following claims.

[0068] The advantages, other benefits, and problem solutions have been described above with reference to specific embodiments. However, the advantages, benefits, problem solutions, and any components that can effect or enhance an advantage, benefit, or solution are not to be construed as critical, necessary, or essential features or components of the claims. REFERENCES 100 A system for controlling neurostimulation in a closed control loop based on real-time electroencephalographic pattern recognition. 102 electroencephalographic signal acquisition units 104 Analog Front-End Unit 106 Signal conditioning processor 108 Feature Extraction Processor 110 Pattern Recognition Processor 112 Control processor 114 Neurostimulation unit 116 Feedback synchronization unit

Claims

A closed-loop neurostimulation control system based on real-time electroencephalographic pattern recognition, comprising: one or more electroencephalographic signal acquisition units configured to be operatively coupled to a scalp or intracranial electrode array to continuously acquire bioelectrical brain signals from a subject; an analog input unit electrically coupled to the electroencephalographic signal acquisition units, the analog input unit comprising an amplification circuit, a filter circuit, and an analog-to-digital conversion circuit configured to convert acquired electroencephalographic signals into digitized signal streams;a signal conditioning processor that is operationally coupled to the analog input stage and configured to perform real-time noise reduction, base correction, motion artifact removal, and band-specific filtering on the digitized signal streams; a feature extraction processor configured to receive processed electroencephalographic signals and compute temporal, spectral, and spatial features, including waveform morphology descriptors, frequency-band power distributions, phase relationships, and interchannel coherence metrics; a pattern recognition processor configured to classify the extracted features into predefined or adaptive neurophysiological states using trained computational models and to generate state identification outputs corresponding to the recognized neural states;a control processor operationally coupled to the pattern recognition processor, configured to determine the neurostimulation control parameters based on the identified neuronal states, the control processor dynamically adjusting the stimulation amplitude, pulse width, frequency, and waveform according to a control policy; a neurostimulation unit electrically connected to one or more stimulation electrodes and configured to deliver controlled electrical stimulation signals to the target tissue depending on the specified neurostimulation control parameters; and a feedback synchronization unit configured to temporally synchronize the acquired electroencephalographic signals with the delivered stimulation signals and update the control processor based on real-time feedback for a continuous closed-loop control system. System according to claim 1, wherein the signal conditioning processor further comprises an adaptive filter circuit configured to suppress mains interference, electromyographic interference and electrode impedance fluctuations using continuously updated filter coefficients derived from real-time signal statistics. System according to claim 1, wherein the feature extraction processor is configured to compute multiresolution time-frequency representations by means of window transformations and generate feature vectors corresponding to transient neuronal events such as epileptiform discharges, oscillatory bursts and phase-locked responses. System according to claim 1, wherein the pattern recognition processor comprises a hierarchical classification arrangement comprising a first processor for anomaly detection and a second processor for multi-class classification of neuronal states, thereby enabling stepwise decision-making for triggering neurostimulation. System according to claim 1, wherein the pattern recognition processor is further configured to update classification parameters by incremental learning based on newly acquired electroencephalographic data and corresponding stimulation results stored in a memory unit. System according to claim 1, wherein the control processor is configured to implement a predictive control strategy that estimates future neuronal state trajectories based on historical electroencephalographic patterns and adjusts the stimulation parameters before the occurrence of undesired neuronal events. System according to claim 1, wherein the neurostimulation unit comprises a programmable pulse generation circuit configured to generate biphasic or monophasic stimulation pulses with charge balancing to minimize tissue damage. System according to claim 1, wherein the feedback synchronization unit is configured to compensate for the latency of signal acquisition and the delay of stimulation delivery by applying a time correction using time-stamped data buffers. System according to claim 1, further comprising a safety monitoring processor configured to continuously evaluate stimulation parameters and physiological responses to ensure compliance with predefined safety thresholds and to interrupt stimulation upon detection of abnormal or unsafe conditions. System according to claim 1, wherein the electroencephalographic signal acquisition units comprise high-density electrode arrays arranged to capture spatially distributed neuronal activity across multiple cortical regions, thereby enabling spatial pattern recognition.

Images