Photostimulation system

Abstract

The invention provides a light stimulation system. The light stimulation system comprises an acquisition module configured to acquire a first acquisition signal of a user; the detection module is configured to obtain a first feature by using an identification algorithm based on the first acquisition signal; determining a user state by using a state model based on the first feature; if the user state is an abnormal state or a pre-abnormal high-risk state, generating a closed-loop intervention control instruction and determining a target stimulation parameter according to the first feature; the light stimulation module is configured to receive a closed-loop intervention control instruction and drive the head-mounted visual light stimulation therapeutic apparatus to operate according to a target stimulation parameter; the problems that an existing photostimulation system can only control an LED array of a photostimulation therapeutic instrument to generate light beams with fixed wavelength, frequency and intensity, the fixed wavelength, frequency and intensity of the light beams generated by the LED array cannot be adjusted by receiving real-time state information of a user, and the application range of the photostimulation system is limited are solved.

Description

[0001] This application claims priority to a Chinese patent application filed with the Patent Office of China on May 23, 2025, application number 202510675505.6, with the invention name “A light stimulation system for treating epilepsy”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the technical field of light stimulation equipment, and in particular to a light stimulation system. Background Art

[0003] The light stimulation system is used to control the LED array of the light stimulation therapy device to generate a light beam of fixed wavelength, frequency, and intensity according to set instructions. The light beam is used to stimulate the user's visual cortex to regulate the excitability of the cerebral cortex, thereby reducing the frequency and intensity of epileptic seizures.

[0004] Currently, the photostimulation system can only control the LED array of the photostimulation therapy device to produce a light beam of fixed wavelength, frequency, and intensity by inputting setting instructions, but is unable to receive real-time status information from the user and adjust the fixed wavelength, frequency, and intensity of the light beam generated by the LED array according to the real-time status information, resulting in limited application scope of the photostimulation system. Summary of the Invention

[0005] The present application provides a light stimulation system to solve the technical problem that the current light stimulation system can only control the LED array of the light stimulation therapy device to generate a light beam with a fixed wavelength, frequency and intensity, but cannot adjust the fixed wavelength, frequency and intensity of the light beam generated by the LED array by receiving real-time status information from the user, resulting in a limited scope of application of the light stimulation system.

[0006] The present application provides a light stimulation system, comprising:

[0007] Acquisition module, detection module, light stimulation module;

[0008] The acquisition module is configured to:

[0009] Acquire a first acquisition signal of the user; the first acquisition signal is acquired by electrodes disposed on the user's head;

[0010] The detection module is configured to:

[0011] Based on the first collected signal, a first feature is obtained using a recognition algorithm; the recognition algorithm includes: a fast Fourier transform algorithm, a wavelet transform algorithm, a time-frequency joint threshold discrimination mechanism, a frequency band energy mutation detection method, a time domain feature extraction method, a synchronization analysis method, a dynamic threshold adjustment mechanism, and cloud-based model feedback adjustment;

[0012] Based on the first feature, a state model is used to determine the user state; the state model is generated by training historical signals and corresponding historical features; the user state includes: normal state, abnormal state, and pre-abnormal high-risk state;

[0013] If the user state is the abnormal state or the high-risk state before the abnormality, generating a closed-loop intervention control instruction and determining target stimulation parameters according to the first feature; the target stimulation parameters include: stimulation frequency, wavelength, and intensity;

[0014] The light stimulation module is configured as follows:

[0015] Receive the closed-loop intervention control instruction and drive the head-mounted visual light stimulation therapeutic device to operate with the target stimulation parameters; the head-mounted visual light stimulation therapeutic device is set on the user's head; the head-mounted visual light stimulation therapeutic device includes: a light-emitting diode array and a pulse width modulation drive circuit; the light-emitting diode array is set in the area in front of the user's eyes.

[0016] In some embodiments, the acquisition module includes:

[0017] An acquisition unit, wherein the acquisition unit is configured to:

[0018] Acquiring a first acquisition signal from the user using electrodes disposed on the user's head; the electrodes are disposed on the forehead, central, parietal, occipital, and bilateral earlobe regions of the user's head;

[0019] A signal processing unit, wherein the signal processing unit is configured to:

[0020] A low-noise preamplifier and a band-pass filter circuit are used to generate input noise and filtered signals respectively;

[0021] The input noise and the filtered signal are used to filter out power frequency interference and high-frequency myoelectric artifacts in the first acquisition signal.

[0022] In some embodiments, the detection module is further configured to:

[0023] Based on the first feature, using a state model, determining whether there are abnormal features; the abnormal features include: spikes and slow waves, high-amplitude synchronous discharges, rhythmic burst slow waves, and abnormal spectral power features;

[0024] If not, determining that the user status is normal;

[0025] If so, determining whether the number of features of the abnormal feature is greater than a preset value sum, or whether a feature in the abnormal feature is greater than a feature threshold;

[0026] If not, the user status is determined to be a pre-abnormal high-risk state;

[0027] If so, it is determined that the user status is abnormal.

[0028] In some embodiments, the detection module is further configured to:

[0029] Obtaining a second collected signal of the user after the head-mounted light stimulation therapeutic device has finished running; the second collected signal is obtained by electrodes arranged on the user's head;

[0030] Based on the second collected signal, using a recognition algorithm, obtaining a second feature;

[0031] Based on the second feature, using a state model, determining whether there is an abnormal feature;

[0032] If not, it is determined that the user's intervention result is effective;

[0033] If so, determining whether the number of features of the abnormal feature is greater than a preset value sum, or whether a feature in the abnormal feature is greater than a feature threshold;

[0034] If not, it is determined that the user's intervention result is partially effective;

[0035] If so, it is determined that the user's intervention result is invalid.

[0036] In some embodiments, the detection module is configured with a stimulation parameter model, and the stimulation parameter model is configured to generate closed-loop intervention control instructions and determine target stimulation parameters based on the abnormal characteristics;

[0037] The system further comprises:

[0038] A parameter optimization module, wherein the parameter optimization module is configured to:

[0039] Storing light stimulation data; the light stimulation data includes: the first acquisition signal, the second acquisition signal, stimulation parameters, and the user's intervention results after the head-mounted light stimulation therapeutic device is completed; the intervention results include: effective, partially effective, and ineffective;

[0040] The stimulation parameter model is trained using the light stimulation data, and the weights of the stimulation parameter model are updated.

[0041] In some embodiments, the system further comprises:

[0042] A data encryption module, wherein the data encryption module is configured to:

[0043] The light stimulation data is encrypted.

[0044] In some embodiments, the stimulation parameter model is configured with a database, wherein the database is configured to store the light stimulation data; the stimulation parameter model is further configured to:

[0045] Based on the abnormal characteristics, a closed-loop intervention control instruction is generated and sent to the light stimulation module, and the target stimulation parameters are determined according to the set rules; the set rules are to select each parameter in the stimulation parameters in the database step by step according to the set levels to determine the target stimulation parameters; the levels include: first level, second level, and third level; the first level is the abnormal characteristics, the second level is the intervention result, and the third level is the numerical value of the stimulation parameter.

[0046] In some embodiments, the stimulation parameter model is further configured to:

[0047] determining a first stimulation parameter corresponding to the abnormal feature in the database;

[0048] Obtaining the corresponding intervention result when the head-mounted light stimulation therapeutic device operates with the first stimulation parameter;

[0049] determining a second stimulation parameter that is effective corresponding to the intervention result;

[0050] sorting the parameters of the second stimulation parameters from large to small according to their values, and determining the parameter corresponding to the minimum value;

[0051] The parameters corresponding to the minimum values ​​are combined to determine target stimulation parameters.

[0052] In some embodiments, the light stimulation module is further configured to:

[0053] receiving the closed-loop intervention control instruction and driving the pulse width modulation drive circuit;

[0054] The pulse width modulation driving circuit controls the light emitting diode array to operate with the target stimulation parameters.

[0055] The present application provides a light stimulation system, which includes: an acquisition module, a detection module, and a light stimulation module; the acquisition module is configured to: acquire a first acquisition signal of the user; the first acquisition signal is acquired by an electrode arranged on the user's head; the detection module is configured to: obtain a first feature based on the first acquisition signal using a recognition algorithm; the recognition algorithm includes: a fast Fourier transform algorithm, a wavelet transform algorithm, a time-frequency joint threshold discrimination mechanism, a frequency band energy mutation detection method, a time domain feature extraction method, a synchronization analysis method, a dynamic threshold adjustment mechanism, and a cloud-based model feedback adjustment; based on the first feature, a state model is used to determine the user state; the state model is generated by historical signals and corresponding historical feature training; the user state includes: normal state, abnormal state, and state model. state, high-risk state before abnormality; if the user state is the abnormal state or the high-risk state before abnormality, then according to the first feature, a closed-loop intervention control instruction is generated and the target stimulation parameters are determined; the target stimulation parameters include: stimulation frequency, wavelength, and intensity; the light stimulation module is configured to: receive the closed-loop intervention control instruction, and drive the head-mounted light stimulation therapy device to operate with the target stimulation parameters; the head-mounted light stimulation therapy device is arranged on the user's head; the head-mounted light stimulation therapy device includes: a light-emitting diode array and a pulse width modulation drive circuit; the light-emitting diode array is arranged in the area in front of the user's eyes, so as to realize real-time monitoring and response to the user state through the light stimulation system, and adjust the operating parameters of the light stimulation therapy device according to the user state, thereby improving the applicability of the light stimulation system. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] In order to more clearly illustrate the technical solution of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0057] Figure 1 Schematic diagram of the structure of the light stimulation system used in this application;

[0058] Figure 2 A schematic diagram of the structure of the acquisition module in this application;

[0059] Figure 3 Flowchart for confirming user status for the state model in this application;

[0060] Figure 4 Flowchart to identify intervention outcomes for the state model in this application;

[0061] Figure 5 This is the first characteristic diagram of epileptic seizures in the different power light stimulation groups in this application;

[0062] Figure 6This is the second characteristic diagram of epileptic seizures in the different power light stimulation groups in this application;

[0063] Figure 7 This is the first characteristic diagram of epileptic seizures in different frequency light stimulation groups in this application;

[0064] Figure 8 This is the second characteristic diagram of epileptic seizures in different frequency light stimulation groups in this application;

[0065] Figure 9 This is the first characteristic graph of epileptic seizures in the different wavelength light stimulation groups in this application;

[0066] Figure 10 This is the second characteristic diagram of epileptic seizures in the different wavelength light stimulation groups in this application.

[0067] Description of reference numerals:

[0068] 1-acquisition module; 11-acquisition unit; 12-signal processing unit; 2-detection module; 3-light stimulation module; 4-parameter optimization module; 5-data encryption module. DETAILED DESCRIPTION

[0069] In order to enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts should fall within the scope of protection of this application.

[0070] For example, epilepsy (EP) is a common neurological disease in children and adults, affecting more than 65 million people and is a major global health problem. Currently, there are nearly 10 million epilepsy patients, and the number is increasing year by year. The average life expectancy of patients with epilepsy is reduced by 2 to 10 years, and the overall mortality rate is 3.0 to 7.9 per 100,000. The early mortality rate is 2 to 3 times that of the general population, and the mortality rate of childhood epilepsy is 5.3 to 8.8 times higher than that of adults. Currently, the number of sudden unexpected deaths in epilepsy alone may be as high as 20,000 per year. Current clinical anti-epileptic seizure drugs (ASMs) are mainly used to control the symptoms of epileptic seizures and lack a causal treatment effect. 30% of patients still suffer from drug-refractory epilepsy. This application proposes a new non-drug anti-epileptic strategy, which envisions interfering with the epilepsy circuit through visual flash stimulation to achieve the purpose of controlling epilepsy.

[0071] Visual light stimulation interferes with epileptic circuits: On the one hand, the visual pathway can induce epileptic seizures. The use of visual light stimulation to induce epileptic seizures, namely the flash stimulation induction test (IPS), has been used in the clinical diagnosis of epilepsy for decades. There are more than 10 types of epilepsy clinically that can induce epileptic seizures by IPS, suggesting that the visual pathway is closely related to epileptic seizures. On the other hand, through reverse thinking, it is speculated that the application of a light stimulation method opposite to the IPS parameters is likely to control epilepsy through the visual pathway. Based on the above concept, the present application proposes a light stimulation system for controlling epilepsy.

[0072] Among them, the light stimulation system uses visual light stimulation with different parameters (including wavelength, frequency, and intensity) from IPS to interfere with the epileptic brain network and related oscillations, so as to inhibit epileptic seizures and epileptic-like discharges and achieve the purpose of controlling epilepsy.

[0073] Closed-loop light stimulation monitors the electroencephalogram (EEG) in real time and, based on the EEG's warning signals before an epileptic seizure, drives the visual light stimulator to deliver appropriate light stimulation, thereby achieving optimally timed light stimulation interference. Artificial intelligence is also introduced to continuously optimize the timing of individual light stimulation, achieving the best effect of light stimulation in controlling epilepsy.

[0074] Epilepsy patients often undergo a latent period after their first seizure, gradually forming a brain network of epileptic circuits, ultimately developing into recurrent epilepsy (the chronic phase). Currently, clinically diagnosed epilepsy patients are already in the chronic phase, with epileptic circuits already established. Initiating treatment during this phase clearly puts them in a delayed and passive position. However, appropriate intervention methods are currently lacking during the latent phase of epilepsy. Therefore, introducing noninvasive, non-pharmacological visual light stimulation therapy during this period is feasible.

[0075] For example, in recent years, several groundbreaking studies have revealed the potential of light stimulation in treating other neurological diseases. For example, a groundbreaking study published in Nature in 2016 demonstrated that 40Hz gamma-band light stimulation can induce gamma oscillations (30-80Hz) in the visual cortex, enhancing the ability of microglia to clear beta-amyloid, thereby improving cognitive function in Alzheimer's disease mice by nearly 50%. A subsequent study published in Cell in 2019 further revealed that multimodal gamma stimulation (light stimulation + sound stimulation) can synergistically regulate neuroimmune networks. This discovery has inspired interdisciplinary research in the field of epilepsy treatment, by screening light stimulation parameters with anti-epileptic effects (such as specific gamma frequencies) to inhibit or intervene in the synchronization process of epilepsy networks. Further research has demonstrated that the cortical-thalamic-temporal lobe circuits of the visual system have close neural functional connections with epilepsy-susceptible brain regions (such as the hippocampus and amygdala), providing a neuroanatomical basis for targeted light stimulation to modulate epilepsy networks. It is hypothesized that light stimulation with specific parameters may inhibit abnormal epileptic discharges, providing a theoretical basis for non-invasive, non-drug neuromodulatory treatments for diseases such as epilepsy.

[0076] At the level of technological innovation, the development of closed-loop neuromodulation systems has opened up new paths for precise intervention in epilepsy. Invasive neuromodulation technologies, represented by responsive neurostimulation (RNS) and deep brain stimulation (DBS), can effectively suppress focal epileptic seizures by real-time monitoring of electroencephalography (EEG) and triggering electrical stimulation. However, its invasive nature leads to high surgical risks, high costs for implanted devices, and a limited number of beneficiaries. In contrast, the development of a non-invasive visual light stimulation system has the advantages of high safety, low cost, and broad applicability. If it can be combined with closed-loop control and artificial intelligence optimization technology, it is expected to break through the limitations of existing anti-epileptic therapies.

[0077] For example, the solutions currently closest to the present application mainly include three invasive anti-epileptic neuromodulation technologies: Vagus Nerve Stimulation (VNS), Deep Brain Stimulation (DBS) and Responsive Neurostimulation (RNS).

[0078] Vagus nerve stimulation (VNS) is applied through an open-loop neuromodulatory device. The system mainly consists of an electric pulse generator implanted subcutaneously in the chest, connecting wires, and stimulating electrodes wrapped around the vagus nerve on the left side of the patient's neck. The device stimulates the vagus nerve through preset periodic electric pulses to regulate the excitability of the cerebral cortex, thereby reducing the frequency and intensity of epileptic seizures. However, VNS is an open-loop stimulation system that cannot monitor and respond to the patient's EEG state in real time, making it difficult to achieve individualized and precise control. In addition, the VNS implantation surgery has certain risks, limited battery life, high long-term maintenance costs, and limited patient applicability. It is currently mainly used for the control of clinical drug-refractory epilepsy, and about half of the patients can reduce epileptic seizures.

[0079] Deep brain stimulation (DBS) is commonly used to treat drug-resistant epilepsy or other neurological diseases. The DBS system consists of stimulating electrodes implanted in specific brain nuclei (such as the anterior thalamic nucleus or hippocampus), an electrical stimulation generator implanted subcutaneously in the chest, and wires connecting the two. The system can continuously monitor the EEG activity of specific brain areas in a closed-loop or semi-closed-loop manner, and automatically or semi-automatically trigger electrical stimulation when abnormal discharges are detected to inhibit abnormal neural network activity. However, the DBS system implantation surgery is complicated, and patients face the risk of intracranial implantation and may experience postoperative complications (such as infection, bleeding, etc.). In addition, the equipment is expensive, and long-term management and maintenance after implantation are difficult. It is mostly used in patients with severe drug-resistant epilepsy, and the benefits to children are limited.

[0080] The responsive neurostimulation (RNS) system is a closed-loop neuromodulation therapy device, which mainly consists of an implantable electrode array, an intracranial signal processing unit, an electrical stimulation generator, and an external control program. First, electrodes are implanted in the patient's intracranial lesion area to monitor the EEG activity in the focal area in real time; second, the EEG signal is transmitted to the processing unit in real time, and abnormal discharges are identified through a preset epileptic seizure detection algorithm; when the abnormal discharge exceeds the predetermined threshold, the stimulation device is automatically triggered to send electric pulses of specific intensity and frequency to the abnormal area to inhibit the progression of epileptic seizures. RNS devices are often equipped with an external program control module for remote adjustment and optimization of treatment parameters to achieve individualized control effects. Although the RNS system achieves closed-loop treatment, it is an invasive solution that requires surgical implantation of intracranial electrodes. Patients face surgical risks and long-term equipment maintenance issues after implantation. At the same time, the cost is high and the beneficiary population is limited.

[0081] In addition, there are currently some non-invasive neuromodulation technologies in clinical practice, such as transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS), which use magnetic fields and direct current, respectively, to achieve non-invasive stimulation and regulation of the cerebral cortex. These technologies do not require surgical implantation of equipment, are relatively safe, and are easier to promote clinically. However, these non-invasive technologies usually lack a real-time closed-loop feedback mechanism, making it difficult to accurately locate and set appropriate stimulation parameters. The interference effect on patients is limited and unstable, making it difficult to implement home treatment and limiting its scope of application.

[0082] In summary, although VNS, DBS, and RNS in the existing technologies have certain anti-interference effects, they are all invasive solutions with disadvantages such as surgical risks, high treatment costs, and limited scope of application. Existing non-invasive stimulation technologies have problems such as insufficient precision, low degree of individualization, and insufficient efficacy. To overcome the above shortcomings, this application proposes a non-invasive, non-drug closed-loop light stimulation anti-epileptic system based on the visual pathway, which combines the advantages of high safety of visual stimulation, flexible parameter adjustment, real-time EEG monitoring feedback, and individualized regulation through artificial intelligence learning.

[0083] For example, although existing non-invasive regulation technologies have been applied in clinical practice, such as transcranial magnetic stimulation (TMS) and transcranial direct current stimulation (tDCS), the stimulation sites obtained by the systems used by the above technologies lack accuracy, making it difficult to achieve precise intervention on the epileptic network and unable to implement individualized intervention models. The interference effect is limited and uncertain, and the equipment is relatively expensive.

[0084] In summary, the obvious defects of the current epilepsy interference technologies include: drug treatment plans are limited to controlling epileptic seizure symptoms and fail to achieve fundamental intervention in the abnormal brain network of epilepsy; invasive VNS, DBS and RNS technologies have surgical trauma and long-term equipment maintenance problems, and the beneficiary population is limited; non-invasive stimulation technologies (such as TMS, tDCS) lack precise positioning and parameter optimization, the interference effect is unstable, and it is expensive.

[0085] Therefore, the current clinical and research fields of epilepsy urgently need a non-invasive, highly safe, precise, reliable and parameter-adjustable personalized portable non-drug anti-epileptic interference technology solution. This solution must have the following characteristics: the ability to intervene in the epileptic network through the visual pathway; the ability to flexibly adjust parameters (timing, duration, frequency, wavelength, intensity) to avoid the risk of inducing epileptic seizures; real-time EEG monitoring and closed-loop feedback control mechanism; personalized and precise regulation based on AI technology (deep learning and online learning); and the ability to move the epilepsy treatment window forward to the epilepsy latency stage, achieving earlier and more effective anti-epileptic intervention.

[0086] In some technologies, the light stimulation system can only control the LED array of the light stimulation therapy device to generate a light beam with a fixed wavelength, frequency, and intensity, but cannot adjust the fixed wavelength, frequency, and intensity of the light beam generated by the LED array by receiving real-time status information from the user. As a result, the scope of application of the light stimulation system is limited. To solve this technical problem, the present application provides a light stimulation system, which is described below:

[0087] like Figure 1 The figure shows a schematic diagram of the structure of the light stimulation system in this application.

[0088] The present application provides a light stimulation system, including: an acquisition module 1, a detection module 2, and a light stimulation module 3.

[0089] The acquisition module 1 is configured to acquire a first acquisition signal of the user. The first acquisition signal is an EEG signal of the patient acquired when the electrodes are just activated.

[0090] like Figure 2 As shown, it is a structural diagram of acquisition module 1 in this application.

[0091] The acquisition module 1 includes:

[0092] The acquisition unit 11 is configured to:

[0093] The first acquisition signal of the user is obtained by using electrodes arranged on the user's head; the electrodes are arranged in the forehead area, central area, parietal lobe area, occipital lobe area and bilateral earlobe areas of the user's head.

[0094] Specifically, taking 10 electrodes as an example, the EEG acquisition module 1 is based on the international standard 10-20 system layout. By configuring 10 highly sensitive Ag / AgCl electrodes, it is used for non-invasive EEG signal acquisition. The specific distribution of electrodes includes: frontal area (Fp1, Fp2), central area (C3, C4), parietal area (P3, P4), occipital area (O1, O2) and bilateral earlobe reference electrodes (A1, A2), which can effectively cover epilepsy-related cortical areas and meet the needs of synchronous monitoring of electrical activities in different brain areas. In order to improve wearing comfort and daily usability, the electrodes adopt a dry or semi-dry structure, combined with flexible conductive materials and an adjustable head-mounted system to ensure good skin contact impedance (≤10kΩ) and electrical signal stability.

[0095] The signal processing unit 12 is configured to:

[0096] A low-noise preamplifier and a bandpass filter circuit are used to generate input noise and a filter signal respectively; and the input noise and the filter signal are used to filter out power frequency interference and high-frequency myoelectric artifacts in the first acquisition signal.

[0097] Specifically, the acquisition module 1 integrates a low-noise preamplifier (input noise density <1μV / Hz) and a bandpass filter circuit (cut-off frequency 1~45Hz) to suppress power frequency interference (50 / 60Hz) and high-frequency electromyographic artifacts, and retain the main frequency band information of the EEG. The digital sampling part adopts a high-performance microcontroller unit (MCU), such as the STM32F4 series, which supports a sampling rate of ≥256Hz and a 24-bit resolution analog-to-digital converter (ADC), and has low-power operation and real-time data transmission capabilities. The processed first EEG signal can be stably transmitted to the detection module 2 via a Bluetooth module or a wireless LAN communication technology module, providing a high-quality data basis for subsequent epileptic wave identification and closed-loop intervention decisions. The acquisition module 1 has the characteristics of high precision, multi-channel, low latency, and strong portability, and is suitable for continuous EEG monitoring needs in clinical and home scenarios.

[0098] The detection module 2 is configured to obtain a first feature based on the first acquired signal using a recognition algorithm; the recognition algorithm includes a fast Fourier transform algorithm, a wavelet transform algorithm, a time-frequency joint threshold discrimination mechanism, a frequency band energy mutation detection method, a time-domain feature extraction method, a synchronization analysis method, a dynamic threshold adjustment mechanism, and cloud-based model feedback adjustment; the first EEG feature includes but is not limited to EEG signal strength and frequency. The first feature is a representative EEG signal within the EEG signal that can detect the user's state and is used to determine the patient's epileptic state, paving the way for subsequently determining the operating parameters of the head-mounted light stimulation therapy device.

[0099] Based on the first feature, a state model is used to determine the user state; the state model is generated by training historical signals and corresponding historical features; the user states include normal state, seizure state, and pre-seizure high-risk state. The seizure state is characterized as an epileptic seizure state; the pre-seizure high-risk state is characterized as a pre-seizure high-risk state. The historical signals and historical features both correspond to historical information of the first acquired signal and the first feature.

[0100] like Figure 3 Shown is a flow chart of the epileptic state model in this application for confirming the epileptic state.

[0101] Specifically, the detection module 2 is further configured to:

[0102] Based on the first feature, the state model is used to determine whether there are abnormal features; the abnormal features include: spike slow waves, high-amplitude synchronous discharges, rhythmic burst slow waves and spectral power abnormal features; if not, the user state is determined to be normal; if so, it is determined whether the number of features of the abnormal features is greater than the preset value, or whether the features in the abnormal features are greater than the feature threshold; if not, the user state is determined to be a high-risk state before an epileptic seizure; if so, the user's epileptic state is determined to be an epileptic seizure state. Among them, the abnormal features are characterized by epileptic EEG features. The patient's epileptic state is determined by the number of abnormal features and whether the abnormal features are greater than the feature threshold (i.e., the danger value). It can be understood that when the patient is in different states (high-risk state before an epileptic seizure, epileptic seizure state), the EEG signal (epileptic EEG features) will also change. This application determines the patient's epileptic state through the above features.

[0103] If the abnormal state is the epileptic seizure state or the high-risk state before the epileptic seizure, a closed-loop intervention control instruction is generated and target stimulation parameters are determined based on the first feature; the target stimulation parameters include: stimulation frequency, wavelength, intensity, string length, duration, and circadian rhythm stimulation.

[0104] Specifically, the detection module 2 is used to perform high-precision real-time analysis on the collected EEG signal, i.e., the first collected signal, and has the dual functions of prediction and identification of epileptic seizures. It is the core control unit for realizing closed-loop control. This module is deployed in an embedded digital signal processor (DSP) or a microcontroller based on the ARM Cortex-M architecture, and has low-latency and high-efficiency data processing capabilities. The algorithm core includes Fast Fourier Transform (FFT), Wavelet Transform (Wavelet Transform) and a time-frequency joint threshold discrimination mechanism, which is used to identify typical EEG features related to epilepsy, such as spike-and-wave discharges (SWD), high-amplitude synchronous discharges, rhythmic burst slow waves and spectral power abnormalities. The system can not only identify clear epileptic seizures, but also predict high-risk EEG changes before the attack by constructing an individualized pre-epileptic state model, thereby achieving prospective intervention. When the real-time analysis algorithm determines that the current EEG is in a pre-epileptic state or a high-probability pre-epileptic state, the system automatically generates a closed-loop intervention instruction, triggering detection module 2 to generate target stimulation parameters and calling light stimulation module 3 to quickly implement the personalized stimulation intervention strategy. The entire process response time is controlled within 200 milliseconds, ensuring timely and accurate intervention. This module also supports collaboration with cloud-based deep learning (online learning) models, dynamically updating recognition thresholds and pattern recognition strategies, and continuously optimizing epileptic seizure identification and prediction capabilities.

[0105] like Figure 4 Shown is a flow chart for confirming intervention results of the epileptic status model in this application.

[0106] In this embodiment, the detection module 2 is further configured to:

[0107] A second signal of the user is obtained after the head-mounted light stimulation therapeutic device is completed; the second signal corresponds to the EEG signal of the patient after the treatment with the head-mounted light stimulation therapeutic device is completed.

[0108] Based on the second signal, a second feature is obtained using a recognition algorithm; based on the second feature, a state model is used to determine whether an abnormal feature exists; if not, the user's intervention result is determined to be valid; if so, it is determined whether the number of features of the abnormal feature is greater than a preset value, or whether the feature in the abnormal feature is greater than a feature threshold; if not, the user's intervention result is determined to be partially valid; if so, the user's intervention result is determined to be invalid.

[0109] Exemplarily, after the patient completes treatment with the head-mounted light stimulation therapy device, the patient's intervention result is determined by the number of abnormal features and whether the abnormal features are greater than the feature threshold (i.e., the danger value). It can be understood that after the patient completes treatment with the head-mounted light stimulation therapy device, the EEG signal will also change (i.e., return to a normal EEG signal). This application determines the patient's intervention result after the head-mounted light stimulation therapy device intervention through the above-mentioned features.

[0110] The light stimulation module 3 is configured as follows:

[0111] Receive the closed-loop intervention control instruction and drive the head-mounted light stimulation therapy device to operate with the target stimulation parameters; the head-mounted light stimulation therapy device is set on the user's head; the head-mounted light stimulation therapy device includes: a light-emitting diode array and a pulse width modulation drive circuit; the light-emitting diode array is set in the area in front of the user's eyes.

[0112] The light stimulation module 3 is further configured as follows:

[0113] The closed-loop intervention control instruction is received to drive the pulse width modulation drive circuit; the pulse width modulation drive circuit controls the light emitting diode array to operate with the target stimulation parameters.

[0114] Specifically, the light stimulation module 3 is primarily responsible for performing light stimulation intervention tasks and consists of a high-brightness adjustable light emitting diode array (LED) and a pulse width modulation (PWM) drive circuit. The LED array is integrated into the forehead area of ​​the device and in front of the visual field, using 660–665nm red light or white light as the primary wavelength. The stimulation frequency is set to 40Hz by default, and the power range is adjustable to 20–80W to adapt to the photosensitivity response characteristics and treatment needs of different individuals. When the main control unit receives the intervention instruction from the closed-loop control module, the microcontroller unit (MCU) immediately drives the PWM modulation circuit based on real-time optimized parameters, precisely controlling the light frequency, wavelength, and intensity of the LED light source to achieve a rapid response in milliseconds. Light stimulation is transmitted to the central nervous system through the visual pathway, regulating abnormal synchronized brain activity, thereby effectively suppressing epileptic discharges within the critical time window and intervening in the initiation and propagation of epileptic seizures. This module has the characteristics of adjustable parameters, rapid response and high control precision. It can work together with the EEG recognition module and AI optimization system to realize individualized and precise visual light stimulation treatment plans.

[0115] In this embodiment, the detection module 2 is configured with a stimulation parameter model, and the stimulation parameter model is configured to generate closed-loop intervention control instructions and determine target stimulation parameters according to the abnormal characteristics;

[0116] The system further comprises:

[0117] The parameter optimization module 4 is configured to:

[0118] Storing light stimulation data; the light stimulation data includes: the first acquisition signal, the second acquisition signal, stimulation parameters, and the user's intervention results after the head-mounted light stimulation therapy device has completed operation; the intervention results include: effective, partially effective, and ineffective; using the light stimulation data to train the stimulation parameter model and update the weights of the stimulation parameter model. The first acquisition signal and the second acquisition signal correspond to the user's EEG signal within a preset time before the head-mounted light stimulation therapy device is operated, and the user's EEG signal within a preset time after the head-mounted light stimulation therapy device is operated, respectively.

[0119] The stimulation parameter model is configured with a database, and the database is configured to store the light stimulation data; the stimulation parameter model is further configured to:

[0120] Based on the abnormal characteristics, a closed-loop intervention control instruction is generated and sent to the light stimulation module 3. The target stimulation parameters are determined according to a set rule. The set rule is to gradually select each stimulation parameter in the database according to a set level to determine the target stimulation parameters. The levels include: first level, second level, and third level. The first level represents the abnormal characteristics, the second level represents the intervention result, and the third level represents the value of the stimulation parameter. By setting a grading mechanism, the most suitable stimulation parameters for the patient can be determined to achieve the best treatment effect.

[0121] Specifically, the stimulation parameter model is further configured as follows:

[0122] Determine a first stimulation parameter in the database corresponding to the abnormal feature; obtain the corresponding intervention result when the head-mounted light stimulation therapy device is operated with the first stimulation parameter; determine a second stimulation parameter that is valid for the intervention result; sort each parameter in the second stimulation parameter from large to small according to the value of the parameter, and determine the parameter corresponding to the minimum value; combine the parameters corresponding to the minimum value to determine the target stimulation parameter.

[0123] It is worth noting that after obtaining the patient's EEG signal, the stimulation frequency, wavelength, and intensity will be determined first based on the abnormal characteristics in the EEG signal; then, based on the patient's EEG signal, it will be determined whether the patient is in an abnormal state or a high-risk state before abnormality. The stimulation parameter model will select the optimal stimulation string length, duration, and circadian rhythm stimulation parameters according to the above rules, which is equivalent to the operating mode of the light stimulation therapy device. Finally, the light stimulation therapy device operates in the optimal operating mode according to the determined optimal light stimulation parameters (stimulation frequency, wavelength, intensity).

[0124] Specifically, when the stimulation parameter model receives the abnormal feature, first, by determining the first stimulation parameter corresponding to the abnormal feature from the database, such as selecting the first stimulation parameter corresponding to the epileptic EEG feature including spike-slow waves and high-amplitude synchronous discharges, the stimulation parameter data in the database is preliminarily screened, and the abnormal feature of the patient is matched with the abnormal feature corresponding to the stimulation parameter data in the database; secondly, the second stimulation parameter with an effective intervention result is screened out from the first stimulation parameter through the intervention result, that is, the stimulation parameter that is effective for the patient's treatment effect is selected, and the first stimulation parameter is screened for the second time; finally, the various parameters in the second stimulation parameter are sorted from large to small according to the numerical value of the parameter, such as sorting the various stimulation parameters corresponding to the stimulation frequency, wavelength, and intensity, and selecting the parameters corresponding to the minimum stimulation frequency, wavelength, and intensity, so as to combine them to obtain the target stimulation parameter. It can be understood that the smaller the intensity of the light stimulation, the higher the patient's acceptance, so selecting a smaller light stimulation intensity is more beneficial to the patient's treatment.

[0125] It is worth noting that the judgment conditions of the above-mentioned grading mechanism are not limited. For example, the patient's age and physical parameters can be added as judgment conditions.

[0126] Specifically, the parameter optimization module 4 is the core support system for realizing individualized and dynamic optimization of treatment strategies. It relies on cloud computing resources to continuously train deep learning models and adaptively optimize reinforcement learning strategies. It covers four sub-processes: data cache upload, model incremental learning, weight update push, and stimulation parameter regulation:

[0127] 1. Real-time data caching and uploading

[0128] During each closed-loop intervention, the device collects 10-second EEG signal segments before and after the intervention in real time, simultaneously recording the stimulation parameters used (including frequency, wavelength, intensity, etc.) and the intervention effect label (such as "effective," "ineffective," or "partially effective"). The data is first cached locally on the device for a short period of time and then regularly transmitted to the cloud training server via Wi-Fi or 4G / 5G wireless networks. The Advanced Encryption Standard (AES) is used for encryption during transmission to ensure data security and privacy compliance.

[0129] 2. Cloud-based online learning and model optimization

[0130] The cloud server deploys a computing platform equipped with a high-performance Graphics Processing Unit (GPU) and runs a deep learning model, namely the stimulation parameter model, which is a hybrid structure of a Convolutional Neural Network–Long Short-Term Memory (CNN-LSTM). The model uses an incremental learning strategy, updating the weight structure in real time as new data is continuously received. Combined with Experience Replay and Elastic Weight Consolidation (EWC) technology, it effectively prevents the forgetting of historical learning content and ensures the model's stability and generalization capabilities during long-term training. The cloud model will be regularly iterated and updated to continuously improve the accuracy of epileptic seizure prediction and the personalized matching of stimulation parameter recommendations.

[0131] 3. Model Weight Differential Update and Terminal Hot Update

[0132] After each round of incremental model training, the cloud generates a weight difference package (ΔW) and pushes the updated content to the device through the Federated Learning mechanism. After receiving and verifying the difference package, the terminal device automatically performs a hot update of the local model without manual intervention, ensuring that online learning results are quickly fed back into the closed-loop control system, achieving continuous optimization and long-term adaptation of the model.

[0133] 4. Online reinforcement learning control strategy

[0134] At the same time, reinforcement learning agent models, such as Deep Q-Network (DQN) or Proximal Policy Optimization (PPO), are integrated into the cloud system. This module analyzes historical stimulation records and treatment feedback in real time to build a reward and punishment mechanism (such as a grading mechanism), using "effective seizure suppression" as the reward signal and "light stimulation energy consumption" as the penalty term. It automatically optimizes the individual stimulation parameter combination (including stimulation timing, duration, frequency, wavelength, and intensity) to achieve the optimal intervention strategy for different patient conditions.

[0135] Through the collaborative work of the above four sub-modules, this system builds an integrated closed-loop learning system of "perception-analysis-feedback-optimization", realizing the dynamic adjustment of stimulation parameters and the continuous evolution of individualized treatment strategies, effectively improving the accuracy, efficiency and long-term control capabilities of epilepsy treatment.

[0136] In this embodiment, the system further comprises:

[0137] The data encryption module 5 is configured to encrypt the light stimulation data.

[0138] Specifically, the data encryption module 5 is used to ensure the security of data communication between the device and the cloud, as well as the long-term stability of the system. The system uses the Advanced Encryption Standard (AES) 256-bit encryption algorithm, combined with the Transport Layer Security (TLS) protocol, to achieve full-link data encryption transmission and authentication, effectively preventing the interception, tampering, or leakage of light stimulation data during transmission, and meeting the high standards of data privacy and compliance required by medical information systems.

[0139] The device supports over-the-air (OTA) updates, enabling remote, automated updates and version management of software, firmware, and deep learning models. The system includes functional modules such as version verification, differential updates, and failure rollback to ensure the stability and security of each upgrade process, preventing system interruptions or operational errors caused by update anomalies. Through regular security patch deployment and model hot-update mechanisms, the data encryption module 5 ensures the platform's long-term security, reliability, and self-evolution capabilities, providing critical support for the intelligent and clinically applicable closed-loop epilepsy intervention system.

[0140] The light stimulation system provided in this application is based on the treatment logic of "real-time monitoring - intelligent identification - closed-loop intervention - adaptive optimization", and has built a closed-loop treatment process that integrates EEG acquisition, light stimulation execution, and cloud learning. The specific treatment steps are as follows:

[0141] Step 1: EEG acquisition and real-time monitoring:

[0142] The child wears a head-mounted EEG acquisition device (acquisition module 1) equipped with multiple (e.g., 10) lead electrodes. Once activated, EEG acquisition module 1 collects multi-channel EEG signals in real time, covering the forehead, central lobe, parietal lobe, occipital lobe, and binaural reference areas. The signals are filtered, amplified, and digitized before being transmitted to detection module 2.

[0143] Step 2: EEG analysis and seizure prediction and identification:

[0144] Seizure Detection Module 2 uses algorithms such as Fast Fourier Transform and Wavelet Transform to extract features and dynamically analyze EEG signals, identifying epileptic EEG characteristics, including spikes and slow waves and high-amplitude synchronous discharges. Simultaneously, Seizure Detection Module 2 runs an epileptic state model to proactively predict and identify seizures in real time, enabling precise timing assessment.

[0145] Step 3: Trigger closed-loop intervention instructions:

[0146] When the detection module 2 determines that the child's EEG enters a high-risk state of epileptic seizure or pre-seizure, it automatically triggers the closed-loop intervention control instruction and calls the light stimulation module 3 to generate the best stimulation plan based on the patient's EEG characteristics.

[0147] Step 4: Implement individualized visual light stimulation:

[0148] The light stimulation module 3 immediately drives the LED array according to the optimized target stimulation parameters (including stimulation frequency, wavelength, and intensity), providing the patient with visual light stimulation of a specific rhythm, regulating abnormal synchronized activities in the brain through the visual cortex-thalamus loop, and quickly interrupting or alleviating epileptic seizures.

[0149] Step 5: Collecting intervention feedback and marking results:

[0150] Detection module 2 synchronously collects changes in EEG signals before and after light stimulation, and evaluates the intervention effect in combination with clinical observations (such as behavioral interruption during epileptic seizures), forming intervention record entries marked as "effective", "ineffective" or "partially effective".

[0151] Step 6: Data upload and cloud training:

[0152] The device locally caches the EEG signals, stimulation parameters, and intervention results before and after the intervention, and regularly uploads them to the detection module 2 via Wi-Fi or 4G / 5G network encryption, providing a data basis for the subsequent training of the stimulation parameter model.

[0153] Step 7: Stimulus parameter model update and push optimization strategy:

[0154] The cloud runs deep learning and reinforcement stimulation parameter models, continuously training and optimizing strategies based on historical data. After training is complete, the system generates weight difference packets and updated parameters, which are pushed to the terminal device through the federated learning mechanism, enabling automatic hot updates of the local model.

[0155] Step 8: Continuous closed-loop adaptive optimization:

[0156] The light stimulation system continuously improves the accuracy of epileptic seizure identification and the efficiency of stimulation intervention through a continuously iterative "identification-intervention-feedback-optimization" process, ultimately forming an adaptive, closed-loop control treatment strategy that meets individual characteristics, achieving the goal of accurate, efficient, and long-term stable non-drug intervention treatment for epilepsy.

[0157] In summary, the above process can be automatically executed and run continuously, supporting long-term daily wear and use. It is particularly suitable for daily monitoring and intervention treatment of children with epilepsy, and has good clinical feasibility and intelligent application prospects.

[0158] This application provides a non-invasive, non-drug, closed-loop photostimulation anti-epileptic treatment system and method based on the visual pathway. By integrating real-time EEG monitoring, closed-loop visual photostimulation intervention, and artificial intelligence parameter optimization, it proposes an innovative, safe, and personalized treatment solution. The key innovations of the technical solution are clearly summarized as follows:

[0159] (1) An anti-epileptic method using a combination of non-inducing visual light stimulation parameters.

[0160] This application proposes and verifies for the first time a combination of visual light stimulation parameters with anti-epileptic effects that is different from induced light stimulation (IPS), including specific frequency (such as 20-60Hz), wavelength (660-665nm red light or white light) and intensity (20-80W), which is proven to be able to effectively inhibit epileptic seizures and abnormal EEG discharges, innovatively expanding the application of visual light stimulation in the treatment of epilepsy.

[0161] (2) A method for real-time EEG feature analysis and closed-loop triggered visual light stimulation.

[0162] This application constructs a detection module 2, which uses feature extraction methods such as fast Fourier transform and wavelet transform to identify epileptic EEG features related to epileptic seizures (such as spikes and slow waves, etc.) in real time, and automatically triggers visual light stimulation immediately after the seizure is predicted or identified, realizing true real-time closed-loop intervention and significantly improving the timeliness and accuracy of intervention.

[0163] (3) A system for individualized online optimization of visual light stimulation parameters based on cloud-based deep learning (online learning).

[0164] This application designs a cloud-based incremental online deep learning model (stimulation parameter model), which uses the convolutional neural network-long short-term memory network (CNN-LSTM) structure to continuously train and optimize the model online through the EEG and light stimulation response data collected on the device side, automatically generate and push the optimal parameter combination of individualized light stimulation, and form a personalized and adaptive treatment model.

[0165] (4) A method for automatic hot updating of device-side models using a federated learning differential update mechanism.

[0166] This application proposes a novel federated learning differential update technology, which is used to achieve the secure push and rapid verification of the differential weight (ΔW) after cloud model training to the device side. The terminal device automatically performs model hot update (HotUpdate), continuously optimizes the treatment effect and model accuracy, and ensures the intelligence and stability of the long-term operation of the device.

[0167] (5) A method for automatic optimization of visual light stimulation parameters with online reinforcement learning capabilities.

[0168] This application introduces a reinforcement learning agent (such as deep Q network DQN or proximal policy optimization PPO) in the cloud system to analyze intervention data in real time and establish an automatic optimization mechanism with treatment effect as positive reward and stimulation energy consumption as negative penalty, dynamically adjusting the timing, duration, frequency, wavelength and intensity of light stimulation to form the optimal visual light stimulation strategy for each patient.

[0169] (6) A therapeutic strategy for early visual light stimulation intervention during the latent period of epilepsy.

[0170] This application proposes and verifies an early visual light stimulation treatment plan, which significantly delays the formation process of epileptic network by intervening in the epileptic latency period, clearly proposes a new treatment concept and implementation path for advancing the epilepsy intervention window, effectively improves the initiative and overall efficacy of treatment, and fills the gap in current clinical early intervention methods for epilepsy.

[0171] (7) A complete visual light stimulation therapy system that integrates real-time monitoring, data analysis, closed-loop control, and AI optimization.

[0172] This application provides a head-mounted integrated system that synergistically integrates the EEG acquisition module, data processing module, closed-loop control module, visual light stimulation implementation module and cloud-based AI optimization module to form a complete "perception-analysis-feedback-optimization" closed-loop system, which can adaptively adjust treatment strategies in real time and has the comprehensive technical advantages of portability, intelligence, safety and efficiency.

[0173] The light stimulation system provided by this application has the following beneficial effects:

[0174] Advantage 1: Non-invasive visual light stimulation, high safety, and good patient compliance

[0175] Current anti-epileptic drugs (ASMs) have significant side effects with long-term use, and some patients develop drug resistance. Invasive therapies such as vagus nerve stimulation (VNS), deep brain stimulation (DBS), and responsive neurostimulation (RNS) require surgical implantation, which carries significant risks and is costly. This application utilizes a completely non-invasive visual light stimulation method, implemented through a head-mounted wearable device, avoiding both the side effects of medications and the risks of invasive surgery, significantly improving patient safety and treatment compliance. It is particularly suitable for long-term, daily use in children and patients who are not suitable for surgery.

[0176] Advantage 2: Flexible and accurate visual light stimulation parameters, avoiding induction risks and providing therapeutic stimulation mode

[0177] Intermittent photostimulation (IPS) is used to induce seizures and diagnose epilepsy. Through systematic animal experiments and clinical studies, this study has identified a specific light stimulation parameter combination of 40Hz frequency, 660-665nm wavelength (red or white light), and 40W intensity, demonstrating its significant anti-epileptic effect. This breakthrough successfully transitions visual light stimulation from diagnostic to therapeutic use, avoiding the risk of epilepsy induction associated with traditional IPS.

[0178] Advantage 3: Real-time EEG closed-loop monitoring and feedback mechanism to improve the timeliness and accuracy of intervention

[0179] Current non-invasive stimulation methods (such as TMS and tDCS) generally lack real-time EEG monitoring and cannot effectively capture the optimal intervention time for epileptic seizures, resulting in insufficient treatment precision, insufficient efficacy, and instability. This application integrates a real-time EEG feature recognition module that can monitor and predict epileptic seizures in real time. Through a closed-loop control algorithm, it automatically triggers visual light stimulation, allowing precise intervention before or in the early stages of an epileptic seizure, significantly improving the timeliness, precision, and effectiveness of treatment interventions.

[0180] Advantage 4: Introducing cloud AI technology to achieve individualized treatment and adaptively optimize efficacy

[0181] Current treatments are all fixed, programmed plans that fail to account for individual patient differences and lack self-learning and optimization mechanisms, making personalized treatment plans impossible. This application innovatively introduces cloud-based centralized training of deep learning technology (CNN-LSTM) and online reinforcement learning methods. By leveraging EEG and treatment feedback data continuously uploaded from the device, it dynamically optimizes treatment strategies in real time and automatically achieves adaptive adjustment of individualized visual light stimulation parameters, significantly enhancing the individualized adaptability and long-term stability of treatment.

[0182] Advantage 5: Achieve early intervention in the latent period of epilepsy and improve overall patient prognosis

[0183] Current clinical treatments are mostly limited to the chronic phase of epilepsy, with delayed intervention timing, failing to effectively prevent the formation of epileptic networks, and resulting in limited overall efficacy. This application proposes and validates a novel treatment strategy, namely, visual light stimulation intervention during the latent phase of epilepsy. By preemptively preventing the formation of epileptic networks or delaying their development, this strategy significantly expands the time window for epilepsy treatment and has the potential to improve patients' long-term prognosis.

[0184] Advantage 6: Integrate a complete closed-loop technology system to build a new intelligent and portable epilepsy treatment platform

[0185] Existing technologies typically employ single modules or open-loop designs, failing to form an effective "perception-analysis-feedback-optimization" closed-loop system and resulting in limited intelligence and automation. This invention successfully constructs a complete closed-loop technology system integrating EEG acquisition, real-time analysis, closed-loop visual stimulation, cloud-based AI learning (online learning), and adaptive optimization. The device boasts comprehensive portability, intelligence, and automation, and is expected to be widely used in both home and clinical settings, significantly enhancing the clinical value of epilepsy treatment and the convenience of daily use for patients.

[0186] For example, although there are some potential alternatives in the field of epilepsy treatment that may overlap with the present application in some functions or goals, these existing technical solutions still have significant gaps with the present invention in terms of overall structure, core mechanism, safety, intelligence, and actual treatment effect, making it difficult to substantially replace the closed-loop visual light stimulation anti-epileptic technology proposed in this application. The specific analysis is as follows:

[0187] (1) Transcranial Magnetic Stimulation (TMS) and Transcranial Direct Current Stimulation (tDCS)

[0188] As noninvasive neuromodulation technologies, TMS and tDCS theoretically possess some potential for treating epilepsy. However, in practice, these technologies lack a clear, real-time EEG closed-loop feedback mechanism, limited stimulation site precision, and insufficient parameter flexibility. This makes them incapable of achieving precise, real-time closed-loop intervention and personalized, optimized treatment of epileptic networks. Consequently, their therapeutic effects are unstable, making them a poor substitute for the visual light stimulation closed-loop control system proposed in this paper.

[0189] (2) Auditory Rhythm Stimulation or Olfactory Rhythm Stimulation

[0190] Some studies have proposed using auditory or olfactory pathways to regulate brain rhythms for epilepsy intervention. However, the mechanisms linking the stimulation pathways and epilepsy networks in these approaches remain unclear, they lack precise EEG closed-loop feedback, their parameter systems have not been systematically established, and the scientific evidence base is weak. Currently, there is no clear and reliable clinical or animal research evidence supporting their efficacy. Therefore, they are unlikely to replace the precise closed-loop control scheme based on the visual pathway of the present invention.

[0191] (3) Constant parameter wearable light stimulation device

[0192] Some existing wearable devices attempt to improve sleep or cognitive function through visual illumination with fixed parameters (such as constant frequency, intensity, and wavelength), but have not yet been used for anti-epileptic treatment. These devices lack real-time EEG monitoring and closed-loop control mechanisms, cannot dynamically adjust stimulation parameters based on the individual EEG state of epileptic patients, and lack intelligent, adaptive treatment capabilities. Therefore, they cannot achieve the precise intervention and long-term efficacy optimization proposed by this invention, and are unlikely to completely replace the technical solution of this application.

[0193] (4) Combined drug therapy and neuromodulation therapy

[0194] Clinically, treatment with anti-epileptic drugs combined with neuromodulatory devices (such as DBS and VNS) can theoretically enhance efficacy, but this still poses challenges such as drug resistance and the accumulation of side effects, preventing long-term drug dependence. Furthermore, invasive neuromodulatory devices carry high surgical risks and maintenance costs, and intervention is limited to the chronic phase of epilepsy. Therefore, this combined approach also fails to meet the goals of the present invention: non-invasive, highly safe, early intervention, and long-term, personalized, optimized treatment.

[0195] In summary, all currently available alternative technical solutions have significant limitations, primarily manifested in a lack of precise closed-loop control capabilities, insufficient adaptability of individualized parameters, delayed intervention timing, and poor safety and convenience. None of these solutions can substantially replace the innovative closed-loop visual light stimulation anti-epileptic treatment technology proposed in this application. Therefore, this application is clearly irreplaceable under current technological conditions, fully demonstrating its technological advancement and unique anti-epileptic treatment strategy.

[0196] For example, Figures 5 to 10 In the figure, respectively, are the first characteristic graph and the second characteristic graph of epileptic seizures in the light stimulation groups with different powers; the first characteristic graph and the second characteristic graph of epileptic seizures in the light stimulation groups with different frequencies; and the first characteristic graph and the second characteristic graph of epileptic seizures in the light stimulation groups with different wavelengths. Figure 5 Middle (a) to (d) represent the number of seizures, average duration of each seizure, total seizure time, and average power of the different power light stimulation groups, respectively; Figure 6 (a) to (d) represent the average waveform length / s, average waveform length / time, average amplitude, and latency of the different power light stimulation groups, respectively; Figure 7 Middle (a) to (d) represent the number of seizures, average duration of each seizure, total seizure time, and average power of the different frequency light stimulation groups, respectively; Figure 8 (a) to (d) represent the average waveform length / s, average waveform length / time, average amplitude, and latency of the different frequency light stimulation groups, respectively; Figure 9 Middle (a) to (d) represent the number of seizures, average duration of each seizure, total seizure time, and average power of the different wavelength light stimulation groups, respectively; Figure 10 (a) to (d) represent the average waveform length / s, average waveform length / time, average amplitude, and latency for different wavelength light stimulation groups, respectively. The characteristic graphs above indicate that the visual light stimulation parameters with significant anti-epileptic effects are: a frequency of 40 Hz, a power of 40 W (equivalent to an ambient illumination of 3100 to 8250 lux), and a wavelength of white light or red light of 660 to 665 nm, with red light being more effective.

[0197] The above specific implementation methods further explain in detail the purpose, technical solutions and beneficial effects of the embodiments of the present application. It should be understood that the above are only specific implementation methods of the embodiments of the present application and are not intended to limit the scope of protection of the embodiments of the present application. Any modifications, equivalent replacements, improvements, etc. made on the basis of the technical solutions of the embodiments of the present application should be included in the scope of protection of the embodiments of the present application.

Claims

1. A light stimulation system, characterized in that: include: Acquisition module (1), detection module (2), light stimulation module (3); The acquisition module (1) is configured to: Acquire a first acquisition signal of the user; the first acquisition signal is acquired by electrodes disposed on the user's head; The detection module (2) is configured to: Based on the first collected signal, using a recognition algorithm to obtain a first feature; The recognition algorithm includes: fast Fourier transform algorithm, wavelet transform algorithm, time-frequency joint threshold discrimination mechanism, frequency band energy mutation detection method, time domain feature extraction method, synchronization analysis method, dynamic threshold adjustment mechanism and cloud model feedback adjustment; Based on the first feature, a state model is used to determine the user state; the state model is generated by training historical signals and corresponding historical features; the user state includes: normal state, abnormal state, and pre-abnormal high-risk state; If the user state is the abnormal state or the high-risk state before the abnormality, generating a closed-loop intervention control instruction and determining target stimulation parameters according to the first feature; the target stimulation parameters include: stimulation frequency, wavelength, intensity, string length, duration, and circadian rhythm stimulation; The light stimulation module (3) is configured as follows: Receive the closed-loop intervention control instruction and drive the head-mounted visual light stimulation therapeutic device to operate with the target stimulation parameters; the head-mounted visual light stimulation therapeutic device is set on the user's head; the head-mounted visual light stimulation therapeutic device includes: a light-emitting diode array and a pulse width modulation drive circuit; the light-emitting diode array is set in the area in front of the user's eyes.

2. A light stimulation system according to claim 1, characterized in that: The acquisition module (1) comprises: An acquisition unit (11), wherein the acquisition unit (11) is configured to: Acquiring a first acquisition signal from the user using electrodes disposed on the user's head; the electrodes are disposed on the forehead, central, parietal, occipital, and bilateral earlobe regions of the user's head; A signal processing unit (12), wherein the signal processing unit (12) is configured to: A low-noise preamplifier and a band-pass filter circuit are used to generate input noise and filtered signals respectively; The input noise and the filtered signal are used to filter out power frequency interference and high-frequency myoelectric artifacts in the first acquisition signal.

3. A light stimulation system according to claim 1, characterized in that: The detection module (2) is further configured to: Based on the first feature, using a state model, determining whether there is an abnormal feature; The abnormal characteristics include: spikes and slow waves, high-amplitude synchronous discharges, rhythmic burst slow waves and abnormal spectral power characteristics; If not, determining that the user status is normal; If so, determining whether the number of features of the abnormal feature is greater than a preset value sum, or whether a feature in the abnormal feature is greater than a feature threshold; If not, the user status is determined to be a pre-abnormal high-risk state; If so, it is determined that the user status is abnormal.

4. A light stimulation system according to claim 3, characterized in that: The detection module (2) is further configured to: Obtaining a second acquisition signal of the user after the head-mounted visual light stimulation therapeutic device has finished running; the second acquisition signal is obtained by electrodes arranged on the user's head; Based on the second collected signal, using a recognition algorithm, obtaining a second feature; Based on the second feature, using a state model, determining whether there is an abnormal feature; If not, it is determined that the user's intervention result is effective; If so, determining whether the number of features of the abnormal feature is greater than a preset value sum, or whether a feature in the abnormal feature is greater than a feature threshold; If not, it is determined that the user's intervention result is partially effective; If so, it is determined that the user's intervention result is invalid.

5. A light stimulation system according to claim 4, characterized in that: The detection module (2) is configured with a stimulation parameter model, and the stimulation parameter model is configured to generate a closed-loop intervention control instruction and determine a target stimulation parameter according to the abnormal characteristics; The system further comprises: A parameter optimization module (4), wherein the parameter optimization module (4) is configured to: Storing light stimulation data; the light stimulation data includes: the first acquisition signal, the second acquisition signal, stimulation parameters, and the user's intervention results after the head-mounted light stimulation therapeutic device is completed; the intervention results include: effective, partially effective, and ineffective; The stimulation parameter model is trained using the light stimulation data, and the weights of the stimulation parameter model are updated.

6. A light stimulation system according to claim 5, characterized in that: The system further comprises: A data encryption module (5), wherein the data encryption module (5) is configured to: The light stimulation data is encrypted.

7. A light stimulation system according to claim 5, characterized in that: The stimulation parameter model is configured with a database, and the database is configured to store the light stimulation data; the stimulation parameter model is further configured to: According to the abnormal characteristics, a closed-loop intervention control instruction is generated and sent to the light stimulation module (3), and the target stimulation parameters are determined according to the set rules; the set rules are to select each parameter in the stimulation parameters in the database step by step according to the set levels to determine the target stimulation parameters; the levels include: first level, second level, and third level; the first level is the abnormal characteristics, the second level is the intervention result, and the third level is the value of the stimulation parameter.

8. A light stimulation system according to claim 7, characterized in that: The stimulation parameter model is further configured as follows: determining a first stimulation parameter corresponding to the abnormal feature in the database; Obtaining the corresponding intervention result when the head-mounted visual light stimulation therapeutic device operates with the first stimulation parameter; determining a second stimulation parameter that is effective corresponding to the intervention result; sorting the parameters of the second stimulation parameters from large to small according to their values, and determining the parameter corresponding to the minimum value; The parameters corresponding to the minimum values ​​are combined to determine target stimulation parameters.

9. The light stimulation system according to claim 1, characterized in that: The light stimulation module (3) is further configured as follows: receiving the closed-loop intervention control instruction and driving the pulse width modulation drive circuit; The pulse width modulation driving circuit controls the light emitting diode array to operate with the target stimulation parameters.

Images