Closed-loop multi-mode nerve stimulation system and method based on time interference

Abstract

The invention relates to the technical field of neural engineering and brain-computer interfaces, in particular to a closed-loop multi-modal nerve stimulation system and method based on time interference, and the system comprises a multi-modal stimulation module which is used for integrating electrical stimulation, magnetic stimulation and optical genetic stimulation, and generating a time interference field domain; the neural state sensing module is used for collecting real-time electroencephalogram signals, blood oxygen concentration and neural metabolite level data; the neural control center is used for fusing neural state data and stimulation parameters, and dynamically adjusting time interference frequency and stimulation intensity through an adaptive algorithm; the time sequence cooperation engine predicts a neural response time phase based on a deep learning model, and optimizes a stimulation time sequence and a mode switching strategy; and the visual interaction platform is used for rendering the nerve activation thermodynamic diagram and the stimulation parameter adjustment curve in real time. Therefore, the problems of adjustment strategy solidification, low adjustment precision, insufficient energy conversion efficiency and the like in the prior art are solved.

Description

Technical Field

[0001] The present invention relates to the field of neural engineering and brain-computer interface technology, and specifically to a closed-loop multimodal neural stimulation system and method based on time interference. Background Art

[0002] As a key technology to ensure the accuracy of neurological disease treatment and the stable operation of stimulation equipment, the closed-loop multimodal neurostimulation system and method based on time interference are subject to factors such as the sharp fluctuations of neuroelectrophysiological signals, the complex physiological environment of the human body, and the variable pathological conditions during long-term clinical application. These factors can easily lead to problems such as decreased adaptability of stimulation parameters, lag in multimodal perception, and insufficient energy conversion efficiency, which increase the risk of treatment effect attenuation and even threaten the safety of patients' neurological functions. As neuroregulation technology develops towards precision and personalization, higher requirements are placed on neurostimulation in terms of real-time parameter response, wide pathological adaptability, long-term stability, and energy utilization efficiency. There is an urgent need for an adaptive adjustment system that can perceive the multi-dimensional state of nerves in real time, dynamically optimize stimulation parameters, and quickly respond to complex pathological conditions to ensure the efficient and accurate implementation of neurostimulation in diverse clinical scenarios. However, traditional neural stimulation regulation has inherent defects: the regulation strategy is rigid and only controls based on simple neural electrical signal thresholds. It does not integrate multi-dimensional information such as pathological trend predictions and physiological environmental parameters, making it difficult to adapt to the complex and changeable characteristics of neural activity, with low regulation accuracy and insufficient energy utilization; the hardware has poor adaptability to complex environments such as strong bioelectric interference, wide physiological parameter fluctuations and tissue movement of the human body, and device performance drift and signal transmission distortion greatly reduce the regulation response speed; the energy management model is extensive, with passive regulation consuming high energy and active regulation lacking dynamic adaptability, resulting in not only low energy conversion efficiency but also easy to aggravate equipment loss; and there is no redundant design, and a single module failure will cause the regulation system to fail. The average annual failure impact time is long, far exceeding the stringent requirements of clinical treatment for neural stimulation reliability. The overall technology faces multiple challenges of rigid regulation strategy, low regulation accuracy and insufficient energy conversion efficiency. Summary of the Invention

[0003] The present application provides a closed-loop multimodal neural stimulation system and method based on time interference to solve the problems of rigid regulation strategy, low regulation accuracy and insufficient energy conversion efficiency in the prior art.

[0004] The first aspect of the present application provides a closed-loop multimodal neural stimulation system based on time interference, including: a multimodal stimulation module, a neural state perception module, a neural control center, a timing collaboration engine, and a visual interaction platform; wherein, the multimodal stimulation module is used to integrate electrical stimulation, magnetic stimulation and optogenetic stimulation to generate a time interference field; the neural state perception module is used to collect real-time EEG signals, blood oxygen concentration and neural metabolite level data; the neural control center is used to fuse neural state data with stimulation parameters, and dynamically adjust the time interference frequency and stimulation intensity through an adaptive algorithm; the timing collaboration engine predicts the neural response phase based on a deep learning model, and optimizes the stimulation timing and modal switching strategy; the visual interaction platform is used to render neural activation heat maps and stimulation parameter adjustment curves in real time.

[0005] Preferably, the multimodal stimulation module includes: a temporal interference electrical stimulation unit, a focused transcranial magnetic stimulation unit, and a near-infrared optogenetic regulation unit; wherein, the temporal interference electrical stimulation unit generates a low-frequency interference field by superimposing dual-frequency currents, targeting deep nuclei; the focused transcranial magnetic stimulation unit adopts pulse sequence modulation to enhance the synergy of the cortical-subcortical pathway; the near-infrared optogenetic regulation unit activates specific neuronal subpopulations through laser-coupled optical fibers.

[0006] Preferably, the neural state perception module includes: a high-density EEG acquisition array, a functional near-infrared spectroscopy module, and a microdialysis sampling unit; wherein, the high-density EEG acquisition array is used to capture the synchronization of neuronal cluster discharges in real time; the functional near-infrared spectroscopy module is used to monitor changes in blood oxygen saturation; and the microdialysis sampling unit is used to detect fluctuations in the concentrations of glutamate and γ-aminobutyric acid neurotransmitters.

[0007] Preferably, the neural control center includes: a data fusion processor and an adaptive adjustment algorithm module; wherein, the data fusion processor fuses multi-source neural signals through Kalman filtering to extract feature vectors; the adaptive adjustment algorithm module is based on the DQN algorithm, with the goal of minimizing neural state errors, and dynamically corrects the time interference frequency and stimulation intensity.

[0008] Preferably, the timing collaborative engine includes: a neural response prediction model and a modal switching controller; wherein, the neural response prediction model uses a long short-term memory network to predict the neural activation phase based on historical stimulation and response data; the modal switching controller quickly switches between electrical stimulation, magnetic stimulation, and light stimulation according to the prediction results.

[0009] Preferably, the visualization interactive platform includes: a three-dimensional neural activation rendering unit, a stimulation parameter adjustment interface, and an efficacy evaluation dashboard; wherein, the three-dimensional neural activation rendering unit is based on the diffusion tensor imaging atlas, superimposed with real-time EEG and blood oxygen data, to generate a dynamic activation heat map; the stimulation parameter adjustment interface is used to manually fine-tune the time interference parameters; the efficacy evaluation dashboard is used to quantify the neural state stability index through the entropy method to evaluate the intervention effect.

[0010] The second aspect of the present application provides a closed-loop multimodal neural stimulation method based on time interference, including: collecting real-time EEG signals, blood oxygen concentration and neural metabolite level data; fusing the EEG signals, blood oxygen concentration and neural metabolite level data, extracting feature vectors through Kalman filtering, training an LSTM model in combination with historical stimulus response data, and predicting the neural activation phase; based on the neural activation phase prediction results, generating a multimodal time interference stimulation scheme, setting the initial frequency difference and stimulation intensity, and dynamically adjusting the time interference parameters and modal switching timing by comparing the actual neural response with the expected target in real time using the DQN algorithm; synchronously rendering the adjusted time interference parameters and modal switching timing to the neural activation heat map and parameter adjustment curve, quantitatively evaluating the intervention effect and providing feedback.

[0011] A third aspect of the present application provides an electronic device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a closed-loop multimodal neural stimulation method based on time interference as described in the above embodiment.

[0012] A fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement a closed-loop multimodal neural stimulation method based on time interference as described in the above embodiment.

[0013] A fifth aspect of the present application provides a computer program product, including a computer program or instructions, for implementing a closed-loop multimodal neural stimulation method based on time interference as described in the above embodiment.

[0014] Therefore, this application has the following beneficial effects: The embodiments of the present application, through the integration of multimodal stimulation and the linkage of multi-dimensional neural state perception, with the help of adaptive adjustment algorithms and deep learning timing prediction, achieve a breakthrough in neural stimulation from single-modal fixed parameters to multimodal dynamic adaptation, effectively improving the stimulation accuracy and neural response matching. At the same time, the generation of time interference fields and real-time feedback from the visual interactive platform not only enhance the adaptability to complex neural environments, but also reduce the impact of device performance drift and signal distortion, significantly improve energy utilization efficiency and system reliability, and comprehensively ensure the safety and therapeutic effect of neural regulation. This solves the problems of rigid adjustment strategy, low adjustment accuracy and insufficient energy conversion efficiency in the existing technology.

[0015] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which: Figure 1 Schematic diagram of the structure of a closed-loop multimodal neural stimulation system based on time interference according to an embodiment of the present application; Figure 2 A schematic diagram of a multimodal stimulation module provided according to one embodiment of the present application; Figure 3 A schematic diagram of a neural state perception module provided according to one embodiment of the present application; Figure 4 A schematic diagram of a neural control center provided according to one embodiment of the present application; Figure 5 A schematic diagram of a timing collaboration engine provided according to one embodiment of the present application; Figure 6 A schematic diagram of a visual interaction platform provided according to one embodiment of the present application; Figure 7 Schematic diagram of a closed-loop multimodal neural stimulation system based on time interference according to an embodiment of the present application; Figure 8 This is a flow chart of a closed-loop multimodal neural stimulation method based on time interference according to one embodiment of the present application; Figure 9 Schematic diagram of a closed-loop multimodal neural stimulation method based on time interference according to one embodiment of the present application; Figure 10 A schematic diagram of the structure of an electronic device provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0017] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described 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 are within the scope of protection of this application.

[0018] The following describes a closed-loop multimodal neural stimulation system based on time interference according to an embodiment of the present application with reference to the accompanying drawings. In response to the problem of low adjustment accuracy mentioned in the above background technology, the present application provides a closed-loop multimodal neural stimulation system based on time interference. In this system, through the linkage of multimodal stimulation integration and multi-dimensional neural state perception, with the help of adaptive adjustment algorithm and deep learning timing prediction, a breakthrough is achieved in neural stimulation from single-modal fixed parameters to multimodal dynamic adaptation, which effectively improves the stimulation accuracy and neural response matching; at the same time, the generation of time interference field and real-time feedback of visual interactive platform not only enhances the adaptability to complex neural environment, but also reduces the impact of device performance drift and signal distortion, significantly improves energy utilization efficiency and system reliability, and comprehensively guarantees the safety and therapeutic effect of neural regulation. As a result, the problems of solidified adjustment strategy, low adjustment accuracy and insufficient energy conversion efficiency in the prior art are solved.

[0019] Figure 1 A schematic structural diagram of a closed-loop multimodal neural stimulation system based on time interference provided in an embodiment of the present application.

[0020] The present embodiment provides a closed-loop multimodal neural stimulation system based on time interference, the system 10 comprising: Multimodal stimulation module 100, neural state perception module 200, neural control center 300, timing coordination engine 400, and visual interaction platform 500.

[0021] Among them, the multimodal stimulation module 100 is used to integrate electrical stimulation, magnetic stimulation and optogenetic stimulation to generate a temporal interference field; the neural state perception module 200 is used to collect real-time EEG signals, blood oxygen concentration and neural metabolite level data; the neural control center 300 is used to fuse neural state data with stimulation parameters, and dynamically adjust the temporal interference frequency and stimulation intensity through an adaptive algorithm; the timing collaboration engine 400 predicts the neural response phase based on a deep learning model, optimizes the stimulation timing and modal switching strategy; the visualization interaction platform 500 is used to render the neural activation heat map and stimulation parameter adjustment curve in real time.

[0022] It is understandable that in the embodiments of the present application, through the linkage of multimodal stimulation integration and multi-dimensional neural state perception, with the help of adaptive adjustment algorithms and deep learning timing prediction, a breakthrough in neural stimulation from single-modal fixed parameters to multimodal dynamic adaptation is achieved, effectively improving the stimulation accuracy and neural response matching degree; at the same time, the generation of time interference field and real-time feedback of the visual interactive platform not only enhances the adaptability to complex neural environments, but also reduces the impact of device performance drift and signal distortion, significantly improves energy utilization efficiency and system reliability, and comprehensively guarantees the safety and therapeutic effect of neural regulation. As a result, the problems of rigid adjustment strategy, low adjustment accuracy and insufficient energy conversion efficiency in the existing technology are solved.

[0023] In the embodiment of the present application, the multimodal stimulation module 100 includes: Figure 2 As shown, there are time-interference electrical stimulation unit, focused transcranial magnetic stimulation unit, and near-infrared optogenetic regulation unit.

[0024] Among them, the time interference electrical stimulation unit generates a low-frequency interference field by superimposing dual-frequency currents, targeting deep nuclei; the focused transcranial magnetic stimulation unit uses pulse sequence modulation to enhance the coordination of cortical-subcortical pathways; and the near-infrared optogenetic regulation unit activates specific neuronal subpopulations through laser-coupled optical fibers.

[0025] It is understood that the temporal interferometric electrical stimulation unit in the present embodiment precisely targets deep nuclei through dual-frequency current superposition; the focused transcranial magnetic stimulation unit utilizes pulse sequence modulation to enhance the synergistic linkage between cortical and subcortical pathways, improving the efficiency of cross-level neural regulation; and the near-infrared optogenetic regulation unit precisely activates specific neuronal subpopulations through laser-coupled optical fibers, overcoming the limitations of nonspecific stimulation. This multi-dimensional approach, from deep nuclei to cortical pathways, and from broad-spectrum stimulation to specific regulation, enhances the targeting and synergy of stimulation through functional complementarity, improving the accuracy and adaptability of neural stimulation.

[0026] For example, in the clinical treatment of Parkinson's disease patients, the temporal interferometric electrical stimulation unit demonstrates unique advantages: by precisely outputting dual-frequency currents of 2kHz and 2.005kHz, and leveraging the natural superposition effect of the two within the skull, a 5Hz low-frequency interference field is generated in the deep subthalamic nucleus region. This mechanism cleverly avoids the invasive procedure of implanting electrodes required for traditional single-frequency electrical stimulation (such as 130Hz high-frequency stimulation), while also resolving the technical bottleneck of low-frequency currents having difficulty penetrating the skull to reach deep nuclei. In actual applications, the low-frequency interference field generated by this unit can specifically inhibit the discharge of overexcited neurons in the subthalamic nucleus, significantly alleviating patients' motor symptoms such as limb tremors and muscle rigidity. Furthermore, because the dual-frequency currents cancel each other out in the surface brain tissue, they avoid widespread activation of the motor and cognitive-related areas of the cerebral cortex, significantly reducing side effects such as speech disorders and cognitive decline caused by traditional stimulation. For example, for advanced patients with a disease course of more than 5 years, after 8 weeks of intervention, the Unified Parkinson's Disease Rating Scale (UPDRS) motor score decreased by an average of 32%, and the improvement in daily activities (such as eating and walking) remained stable, reflecting the dual advantages of non-invasiveness and precision.

[0027] In the embodiment of the present application, the neural state perception module 200 includes: Figure 3 As shown, high-density EEG acquisition array, functional near-infrared spectroscopy module, and microdialysis sampling unit.

[0028] Among them, the high-density EEG acquisition array is used to capture the synchronization of neuronal cluster discharges in real time; the functional near-infrared spectroscopy module is used to monitor changes in blood oxygen saturation; and the microdialysis sampling unit is used to detect fluctuations in the concentrations of glutamate and γ-aminobutyric acid neurotransmitters.

[0029] It is understood that the embodiments of the present application use a high-density EEG acquisition array to capture the synchronization of neuronal cluster discharges in real time, accurately reflecting the dynamic correlation of neural electrical activity; a functional near-infrared spectroscopy module monitors changes in blood oxygen saturation and synchronously correlates the state of neural metabolic activity; and a microdialysis sampling unit detects fluctuations in the concentration of glutamate and γ-aminobutyric acid neurotransmitters, directly capturing real-time changes in chemical signals. Through the complementary verification of multimodal data, the dynamic characteristics of the neural state are comprehensively and accurately portrayed, providing a three-dimensional, real-time status basis for the neural control center, supporting the dynamic optimization of stimulation parameters, and improving the adaptability and accuracy of neural regulation.

[0030] For example, in the long-term monitoring and intervention of intractable epilepsy patients, high-density electroencephalogram acquisition array (equipped with 256-channel flexible electrodes to cover the whole cerebral cortex at a 5-mm interval, with a sampling rate of 1000 Hz) has unique advantages: when the patient is in the interictal period, the array can capture subtle abnormalities that are difficult to identify by conventional 16-channel electroencephalogram, such as subclinical discharges in the medial temporal lobe and insular cortex (manifested as a temporary increase in high-frequency gamma band power in local micro-regions, lasting about 200-300 ms); when the seizure aura appears, the array can quickly analyze the synchrony characteristics of neuron clusters throughout the brain, such as a sudden increase in theta band (4-7 Hz) power in the hippocampus and cingulate gyrus by more than 3 times, with a phase-locking value (PLV) of more than 0.8 across regions, forming a typical "discharge diffusion path" (rapidly spreading from the temporal lobe to the frontal lobe). The real-time captured electrophysiological characteristics, through low-delay transmission (≤10 ms) to the neural control center, not only provide millimeter-level spatial resolution for the precise positioning of the epileptic focus (better than the positioning error of traditional 32-channel electroencephalogram), but also serve as a trigger signal for time-interference electric stimulation units - about 5-8 seconds before the seizure threshold, the stimulation units can be driven to apply specific intervention to the abnormally synchronized regions. Clinical data shows that the use of the array has increased the seizure warning accuracy of patients by 42%, reduced the misjudgment rate by 28%, and reduced the impact on cognitive function by 19% due to the reduction of unnecessary generalized stimulation resulting from precise positioning.

[0031] In the embodiments of the present application, the neural control center 300 comprises a data fusion processor and an adaptive adjustment algorithm module, as shown in the figure. Figure 4 The data fusion processor and the adaptive adjustment algorithm module are connected to each other.

[0032] The data fusion processor fuses multi-source neural signals through Kalman filtering to extract feature vectors, and the adaptive adjustment algorithm module dynamically corrects the time interference frequency and the stimulation intensity based on the DQN algorithm to minimize the neural state error.

[0033] It can be understood that the data fusion processor of the embodiments of the present application performs noise suppression and spatio-temporal alignment on multi-source neural signals through Kalman filtering to extract high-fidelity feature vectors, providing accurate data basis for adjustment decisions; the adaptive adjustment algorithm module dynamically corrects the time interference frequency and the stimulation intensity based on the DQN algorithm to minimize the neural state error, ensuring the reliability of the adjustment basis, improving the adaptability of neural stimulation to complex dynamic neural states, and enhancing the accuracy and intelligence of the control.

[0034] For example, in neuromodulation therapy for patients with major depressive disorder, the adaptive adjustment algorithm module (based on the DQN algorithm) receives dynamic data from the neural state perception module in real time. When high-density EEG indicates an abnormal increase in prefrontal alpha band power (indicating worsening low mood) and microdialysis detects low serotonin concentrations, the algorithm automatically adjusts the temporal interferometric stimulation frequency from 10Hz to 7Hz (to enhance regulation of the deep limbic system) while simultaneously increasing the stimulation intensity by 15% (to promote neurotransmitter release) with the goal of minimizing neural state errors. Furthermore, when functional near-infrared spectroscopy reveals a sudden increase in prefrontal oxygen saturation (indicating a risk of overactivation), the algorithm quickly returns the intensity to baseline and switches to intermittent stimulation. This dynamic adaptation reduces patients' Hamilton Depression Rating Scale (HAMD) scores by 28% compared to traditional fixed-parameter therapy and avoids the anxiety-inducing side effects of continuous high-intensity stimulation, demonstrating the advantage of transitioning from passive adherence to preset parameters to active response to real-time neural states.

[0035] In the embodiment of the present application, the timing collaboration engine 400 includes: Figure 5 As shown, neural response prediction model and mode switching controller.

[0036] Among them, the neural response prediction model is based on historical stimulus and response data and uses a long short-term memory network to predict the neural activation phase; the mode switching controller quickly switches between electrical stimulation, magnetic stimulation, and light stimulation according to the prediction results.

[0037] It can be understood that the neural response prediction model in the embodiment of the present application learns historical stimulus and response data through a long-short-term memory network, and makes a forward-looking prediction of the neural activation phase, so as to gain advance time for intervention; the modal switching controller quickly switches between electrical, magnetic, and optical stimulation based on the prediction results, and can dynamically match the most appropriate stimulation mode according to the neural state, thereby improving the timing accuracy of neural stimulation, and enhancing the adaptability and depth of regulation through multi-modal on-demand coordination, effectively optimizing the stimulation efficiency, reducing the side effects caused by ineffective stimulation, and improving the targeting and safety of neural regulation.

[0038] For example, in the cognitive regulation of Alzheimer's patients, the temporal coordination engine's neural response prediction model (based on an LSTM network, trained on three months of stimulus-response records) can predict the neural activation phase eight seconds in advance. When a decreasing trend in hippocampal EEG beta band (13-30Hz) power is detected (indicating weakened memory retrieval function), and historical data indicates optimal response to magnetic stimulation at that time, the model predicts that activation of the hippocampal-prefrontal pathway needs to be strengthened in 1.2 seconds. The modality switching controller then triggers a switch: from the current temporal interferometric electrical stimulation (maintaining baseline activation) to focused transcranial magnetic stimulation (20Hz pulse frequency, 0.8 second duration), simultaneously enhancing subcortical coordination. When the occipital visual cortex is predicted to enter an inhibition phase (based on blood oxygen metabolism trends), the controller switches to near-infrared optogenetic stimulation to activate a subpopulation of visual-associated neurons. This "prediction-switching" linkage enables patients' Mini-Mental State Examination (MMSE) scores to increase by 2.3 points per month, which is 1.5 times faster than fixed-modality stimulation. In addition, by avoiding the neural refractory period, the stimulation energy utilization rate is increased by 30%, reducing unnecessary stimulation loss.

[0039] In the embodiment of the present application, the visual interaction platform 500 includes: Figure 6 As shown, there are three-dimensional neural activation rendering unit, stimulation parameter adjustment interface, and efficacy evaluation dashboard.

[0040] Among them, the three-dimensional neural activation rendering unit is based on the diffusion tensor imaging atlas, superimposed with real-time EEG and blood oxygen data to generate a dynamic activation heat map; the stimulation parameter adjustment interface is used to manually fine-tune the time interference parameters; the efficacy evaluation dashboard is used to quantify the neural state stability index through the entropy method to evaluate the intervention effect.

[0041] It can be understood that the three-dimensional neural activation rendering unit in the embodiment of the present application generates a dynamic heat map by superimposing real-time EEG and blood oxygen data on diffusion tensor imaging maps, converting abstract neural activities into intuitive spatial visualization, helping operators to accurately locate the activation area; the stimulation parameter adjustment interface supports manual fine-tuning of time interference parameters, providing a flexible supplement to automatic control, and adapting to individual differences and complex scenarios; the efficacy evaluation dashboard quantifies the neural state stability index through the entropy method, objectively displays the intervention effect, improves the visualization and comprehensibility of the neural regulation process, enhances the flexibility of parameter adjustment and the scientific nature of effect evaluation, helps operators make quick decisions and optimize plans, and improves the system's usability and intervention accuracy.

[0042] For example, at the neuromodulation center of a tertiary hospital, a 65-year-old male Parkinson's disease patient, surnamed Zhang, is undergoing multimodal neurostimulation therapy based on time-interference. The patient has a five-year history of resting tremor in his right limbs (amplitude approximately 3mm). Although the tremor amplitude has been reduced to 1.5mm after the system's automatic control (time-interference electrical stimulation frequency difference 8Hz, magnetic stimulation pulse interval 40ms, and light stimulation duty cycle 50%), the tremor amplitude still remains high-frequency micro-tremor (amplitude 0.8mm) in his right fingers when performing fine movements (such as holding a pen and writing). Furthermore, fNIRS monitoring shows a 150ms delay in the blood oxygen response in the left motor cortex (M1 area), deviating from the target response synchronization. At this point, the attending physician intervened and adjusted the stimulation parameter adjustment interface of the visual interactive platform. The left side of the interface displays a real-time 3D neural activation heatmap, while the right side displays the current multimodal parameters: 8Hz frequency difference and 2.2mA intensity for the electrical stimulation channels (C3-Cz); 40ms pulse pattern (TBS) for magnetic stimulation (M1 region); and 5mW power and 50% duty cycle for light stimulation (substantia nigra). The data panel below simultaneously updates EEG alpha wave power (0.62), striatal dopamine concentration (20nM), and the neural state stability index (NSI = 0.85). Based on the patient's characteristic of "tremor worsening with fine motor movements," the doctor determined that enhanced synchronization of the motor cortex-basal ganglia circuit was necessary. First, click the "Electrical Stimulation Parameters" tab on the interface and fine-tune the frequency difference from 8Hz to 7.2Hz. This is because preoperative DTI analysis showed that the fiber connectivity between the subthalamic nucleus and the motor cortex was 12% lower than normal. A slightly lower frequency difference can enhance synchronization between deep nuclei and the cortex. Then, switching to "Magnetic Stimulation Settings," the pulse interval was increased from 40ms to 43ms. This was done to match the latency of the motor-related cortical potential (MRCP) in real-time EEG (80ms longer than baseline), and the interval was extended to match the neural conduction delay. Finally, in "Light Stimulation Control," the power was increased from 5mW to 5.5mW. Microdialysis data showed that striatal dopamine concentrations in patients transiently dropped to 18nM (target ≥22nM) during fine motor movements, necessitating enhanced optogenetic dopamine release efficiency. During the adjustment process, the interface generated a real-time "parameter-response correlation curve": When the frequency difference was adjusted to 7.2Hz, the amplitude of the right limb tremor dropped to 1.2mm in real time, and the synchronized EEG showed that the alpha wave power in the contralateral M1 area increased from 0.62 to 0.68. After adjusting the pulse interval, the MRCP latency was shortened to 50ms, and the synchronization error with the stimulation timing narrowed from 150ms to 30ms. After increasing the light stimulation power by 10%, the dopamine concentration returned to 23nM within 30 seconds, and the patient's finger tremor when holding a pen almost disappeared (amplitude ≤0.3mm). The patient reported that "the right hand is more stable when holding chopsticks, and wrist rotation is no longer stuck." The NSI (Neural State Stability Index) increased from 0.85 to 0.93.

[0043] The closed-loop multimodal neural stimulation system based on time interference proposed in the embodiment of the present application achieves a breakthrough from single-modal fixed parameters to multimodal dynamic adaptation of neural stimulation through the linkage of multimodal stimulation integration and multi-dimensional neural state perception, with the help of adaptive adjustment algorithm and deep learning timing prediction, effectively improving the stimulation accuracy and neural response matching. At the same time, the generation of time interference field and real-time feedback of the visual interactive platform not only enhance the adaptability to complex neural environments, but also reduce the impact of device performance drift and signal distortion, significantly improve energy utilization efficiency and system reliability, and comprehensively guarantee the safety and therapeutic effect of neural regulation. In this way, the problems of rigid adjustment strategy, low adjustment accuracy and insufficient energy conversion efficiency in the existing technology are solved.

[0044] The following will describe a closed-loop multimodal neural stimulation system based on time interference through a specific embodiment. Figure 7 Shown, including: A 28-year-old female patient with intractable epilepsy was admitted to the Department of Neurology at a tertiary hospital. She had a 10-year history of complex partial seizures originating in the right temporal lobe (3-4 times per week), preceded by right-sided facial numbness and aura. EEG revealed high-frequency spikes (15-25 Hz) in the right middle temporal gyrus (MTG), followed by 30-60 seconds of post-ictal confusion. Previous antiepileptic drug treatment (sodium valproate + levetiracetam) had been ineffective. Preoperative MRI and PET-CT identified the right MTG as the epileptogenic focus, and DTI revealed abnormal neural fiber connections between this region and the hippocampus and amygdala, necessitating precise control through multimodal neurostimulation.

[0045] The epileptogenic focus and associated network were targeted by integrating three stimulation methods. The temporal interferometric electrical stimulation unit used the Nexstim company's eXimiaTMS system. Two pairs of circular electrodes with a diameter of 10 mm (3 cm apart, impedance 3-5 kΩ) were placed on the scalp corresponding to the right MTG area. A 2.5 kHz and 2.512 kHz dual-frequency sinusoidal current was output to generate a 12 Hz low-frequency interference field with an initial intensity of 1.8 mA, a pulse width of 200 μs, and a duty cycle of 60% (stimulation for 2 seconds / rest for 1 second). The focused transcranial magnetic stimulation unit used the MagVenture MagProX100 magnetic stimulator equipped with a figure-8 coil ( The peak magnetic field was 2.2T and fixed on the right DLPFC. The cTBS mode (50Hz pulse train, 3 pulses per train, 200ms interval between trains, and intensity of 90% of the resting motor threshold) was used and automatically started when the EEG detected a spike precursor. The near-infrared optogenetic regulation unit was implanted in the CA3 area of ​​the right hippocampus through stereotactic surgery. The fiber optic cannula with a diameter of 300μm and a numerical aperture of 0.37 was connected to the DoricLenses 980nm laser (power 0-8mW, initial 5mW), which was triggered synchronously with the electrical stimulation at a pulse frequency of 20Hz to activate inhibitory neurons expressing ArchT and reduce hippocampal hyperexcitability.

[0046] Three types of equipment were used to monitor epileptic focus activity in real time. The high-density EEG acquisition array used a 64-channel Nihon Kohden Neurofax EEG-1200 device, covering the bilateral temporal and frontal lobes, with a sampling rate of 2000 Hz and a bandpass filter of 0.5-70 Hz. The focus was on monitoring the spike frequency (target ≤ 0.5 times / minute), amplitude (target ≤ 50 μV), and α / γ wave power ratio (target ≥ 1.2) of the right MTG. The functional near-infrared spectroscopy module used a Hitachi ETG-4000 device, which was placed in the right temporal lobe. A 4×4 probe array (light source-detector spacing 3 cm) with a sampling rate of 10 Hz was set up to monitor the HbO2 change rate (target fluctuation ≤15%) and hemodynamic response delay (target ≤200 ms) in the MTG area. A microdialysis sampling unit with a 4-mm-long membrane and a 30-kDa molecular weight cutoff was implanted around the right MTG epileptogenic focus. A microsyringe pump with a flow rate of 1.5 μL / min was connected to perfuse 37°C artificial cerebrospinal fluid. Dialysate was collected every 10 minutes, and the concentrations of glutamate (target ≤10 μM) and GABA (target ≥8 μM) were detected by HPLC.

[0047] The hardware uses an industrial-grade edge computing server (Intel Core i9-13900K processor, 64GB DDR5 memory) equipped with a real-time Linux system (RT_PREEMPT patch) to ensure that the delay between data processing and command issuance is ≤5ms. In terms of data fusion and algorithm implementation, the data fusion processor uses Kalman filtering to fuse the EEG γ / α power ratio (weight 0.4), fNIRS HbO2 change rate (weight 0.3), and microdialysis glutamate / GABA concentration ratio (weight 0.3) to generate the epileptogenic activity index (EAI, target ≤0.3). The adaptive adjustment algorithm module is based on DQN training, inputs the current EAI and stimulation parameters, and outputs the time interference. The regulatory actions involved frequency difference of ±0.1Hz, stimulation intensity of ±0.1mA, and optical power of ±0.2mW, with "EAI ≤ 0.3 for 5 consecutive minutes" as the reward target, were optimized through offline training (patients' data from the past 3 months) and online fine-tuning; the LSTM network of the timing collaboration engine (input layer 64 dimensions, hidden layer 128 nodes, output layer 1 dimension) input the previous 10s EAI sequence and predicted the probability of seizure in the next 1s (error ≤ 100ms). The mode switching controller linked the device through the GPIO interface. When the predicted probability was ≥70%, the magnetic stimulation cTBS mode was triggered (lasting 5s), and the electrical stimulation frequency difference was simultaneously adjusted to 15Hz and the optical power was increased to 6mW, forming a "prediction-enhanced intervention" synergy. Based on the predicted results, the timing coordination engine triggered a modality switch 100ms before tremor onset: the focused transcranial magnetic stimulation unit initiated TBS mode (3 pulses / train, 50Hz) for 2s to suppress motor cortical hyperexcitability. Simultaneously, the temporal interference electrical stimulation frequency difference was adjusted from 5Hz to 8Hz to enhance the inhibitory effect of the subthalamic nucleus. The optogenetic control unit emitted 808nm laser light (for 1s) to promote dopamine release. Actual neural responses (e.g., alpha power increased to 0.6, dopamine concentration returned to 20nM) were compared in real time with target values ​​(alpha power ≥ 0.5, dopamine ≥ 22nM). Using the DQN algorithm, the stimulation intensity was increased from 2mA to 2.2mA, and the magnetic stimulation interval was shortened from 50ms to 40ms until the neural state stability index (NSI) reached 0.85.

[0048] It uses a 21.5-inch 4K touch screen (response time 5ms) and realizes data synchronization between the doctor and the device through the 1Gbps hospital LAN. In terms of interface layout, the left side displays a three-dimensional neural activation heat map based on the preoperative MRI and DTI fusion atlas, with real-time superposition of EEG spike wave emission position (red flashing point) and fNIRS high perfusion area (yellow gradient), with a refresh rate of 1Hz; the right side column displays multimodal stimulation parameters (electrical stimulation frequency difference 12Hz, intensity 1.8mA, magnetic stimulation cTBS mode, trigger threshold 70%, light stimulation power 5mW, frequency 20Hz), supporting Hold touch to make fine adjustments (the step size can be set); the bottom panel refreshes the EAI value, interictal interval, patient VAS score (0-10 points) and neurological stability index (NSI, calculated by entropy method, target ≥0.8) in real time. The visual interactive platform is based on the DTI atlas, rendering the motor cortex activation heat map in real time (the red area indicates overexcitement), and displaying the time interference frequency and intensity adjustment curve; the doctor observes the NSI change trend through the efficacy evaluation dashboard. If it is stable above 0.8 for 3 consecutive minutes, it is confirmed that the intervention is effective, and the parameters can be manually fine-tuned (such as reducing the light stimulation power to 4mW).

[0049] In summary, the embodiment of the present application uses multimodal stimulation to coordinately target the epileptogenic focus and associated networks (right MTG, hippocampus, etc.), combines high-density EEG, fNIRS and microdialysis to achieve real-time monitoring of neural status, relies on Kalman filtering to fuse multi-source data to generate the EAI index, dynamically adjusts the stimulation parameters (frequency difference, intensity, etc.) through the DQN algorithm, and uses the LSTM network to predict the risk of seizures, triggering the "prediction-enhanced intervention" collaborative mechanism (such as magnetic stimulation cTBS mode and timing adjustment), which greatly improves the targeting and timeliness of the regulation; at the same time, the visual interactive platform realizes human-computer collaboration through three-dimensional heat maps and real-time parameter panels, supports doctors to accurately fine-tune, and effectively reduces the right MT The frequency of G spikes was reduced (from 1.2 times / minute before surgery to less than 0.5 times / minute), the α / γ wave power ratio was improved (reaching above 1.2), the glutamate concentration was controlled within 10μM, GABA was maintained above 8μM, and the neural state stability index (NSI) was stabilized above 0.8. The frequency of epileptic seizures was significantly reduced (from 3-4 times per week to a controllable range), the time of confusion after the seizure was shortened, and the side effects of excessive stimulation were avoided through dynamic regulation. Accurate, safe and efficient neural regulation was achieved in the treatment of drug-refractory epilepsy, while providing doctors with an intuitive decision-making basis and promoting the upgrade of the treatment model from "passive response" to "active prevention".

[0050] Next, a closed-loop multimodal neural stimulation method based on time interference proposed according to an embodiment of the present application is described with reference to the accompanying drawings.

[0051] like Figure 8As shown, the closed-loop multimodal neural stimulation method based on time interference includes the following steps: In step S101 , real-time EEG signals, blood oxygen concentration, and neurometabolite level data are collected.

[0052] It is understood that by collecting real-time EEG signals, blood oxygen concentration, and neurometabolite level data, the embodiments of this application can capture the dynamic activity of the epileptogenic focus and its associated network in real time from three dimensions: electrophysiological activity, brain metabolic state, and neurotransmitter balance. This provides a fundamental basis for data fusion, risk prediction, and parameter adjustment. Through comprehensive perception of multimodal data, subtle changes in neural status can be accurately portrayed, enabling timely responses to abnormal activity in the epileptogenic focus, improving the targeting and timeliness of neural stimulation, and reducing the side effects of ineffective stimulation.

[0053] In step S102, the EEG signal, blood oxygen concentration and neural metabolite level data are fused and processed, feature vectors are extracted through Kalman filtering, and the LSTM model is trained in combination with historical stimulus response data to predict the neural activation phase.

[0054] Among them, the neural activation phase refers to the time stage characteristics of neurons or neural regions during the activation process, including the onset, peak, and decay timing states of activation.

[0055] It can be understood that the embodiments of the present application accurately grasp the key timing nodes of the start, peak, and decay of neuron or neural region activation by predicting the neural activation phase, providing a precise timing anchor point for neural stimulation, and by predicting the time pattern of neural activity in advance, initiating intervention before the start of activation, strengthening regulation at the peak, and adjusting the stimulation intensity in the decay stage, so as to accurately match the timing of electrical, magnetic, and light stimulation with neural activity, avoid inefficiency or side effects caused by blind stimulation, and significantly improve the accuracy, timeliness, and safety of neural regulation by amplifying the regulation effect through multimodal collaboration.

[0056] For example, in the treatment of a 28-year-old patient with intractable epilepsy, real-time EEG (spike precursors), fNIRS blood oxygenation changes, and glutamate concentration data were collected from the right MTG region. After being fused into feature vectors using a Kalman filter, these data were fed into a trained LSTM model (based on the patient's stimulus response data from the previous three months). By analyzing the Epileptic Focus Activity Index (EAI) sequence over the preceding 10 seconds, the system accurately predicted the neural activation phase of the right MTG over the next 1 second: activation onset after 300 ms, peak activation at 600 ms, and extinction at 900 ms. Based on this information, the system initiated the cTBS mode 200 ms in advance. At peak activation, the stimulation frequency difference was adjusted from 12 Hz to 15 Hz. During the extinction phase, the light stimulation power was gradually reduced to precisely match the timing of stimulation with neural activity. This system successfully increased the efficiency of suppressing abnormal discharges in this region by 40%, extending the spike interval to twice its preoperative level after a single intervention.

[0057] In step S103, based on the neural activation phase prediction results, a multimodal time interference stimulation scheme is generated, the initial frequency difference and stimulation intensity are set, and the time interference parameters and modality switching timing are dynamically adjusted using the DQN algorithm by comparing the actual neural response with the expected target in real time.

[0058] Among them, the DQN algorithm is an algorithm that integrates deep learning and reinforcement learning. It approximates the Q-value function through a neural network to achieve a mapping from state to optimal action to maximize the cumulative reward.

[0059] It can be understood that the embodiments of the present application construct a dynamic mapping between neural states and stimulation strategies, use neural networks to learn in real time the deviation between actual neural responses and expected targets, output the optimal time interference parameter adjustment (such as frequency difference, intensity) and modal switching timing, adapt to the dynamic changes of adaptive neural activity, and continuously optimize the strategy with maximizing the stability of neural states as the reward goal, avoiding the adaptation limitations of fixed parameters to complex neural states, improving the coordinated accuracy and real-time performance of multimodal stimulation, reducing the side effects of ineffective stimulation, and making the intervention effect more in line with individual neural characteristics.

[0060] For example, in the treatment of a 28-year-old patient with intractable epilepsy, the system initially set the electrical stimulation frequency difference to 12Hz, the intensity to 1.8mA, and the optical stimulation power to 5mW. At this point, the epileptogenic activity index (EAI) was 0.65, falling short of the target (≤0.3). The DQN algorithm, with "reduced EAI and improved neural stability index (NSI)" as reward objectives, inputs the current neural state, including EAI, spike frequency, and blood oxygen oscillation rate, as well as stimulation parameters. The neural network then outputs the optimal adjustment action: The electrical stimulation frequency difference was first increased to 12.3Hz (in 0.1Hz steps) and the optical power to 5.4mW (in 0.2mW steps). After monitoring the EAI to drop to 0.42, the electrical stimulation intensity was further increased to 1.9mA, and the trigger interval between magnetic stimulations was shortened by 10ms. After five rounds of dynamic learning, the EAI stabilized at 0.28, the NSI increased to 0.85, and the spike frequency decreased by 60%. This algorithm precisely adapted the stimulation parameters to the individual's neural state, avoiding the limitations of fixed strategies.

[0061] In step S104, the adjusted time interference parameters and mode switching timing are synchronously rendered to the neural activation heat map and parameter adjustment curve to quantitatively evaluate the intervention effect and provide feedback.

[0062] Among them, the neural activation heat map is a visual map based on the brain structure map, superimposed with real-time neural signal data, and uses color depth to intuitively display the activation intensity and distribution of neural areas.

[0063] It can be understood that the embodiments of the present application, through the use of neural activation heat maps, convert abstract neural activation intensity and distribution into intuitive color gradient visualization information, accurately locate high-activation areas and post-intervention change trends, and provide doctors with a real-time, concrete dynamic map of neural activity. The range and intensity changes of the activation areas before and after the intervention can be intuitively compared, assisting in quickly judging the effectiveness of the stimulation parameter adjustment, reducing the cost of interpreting complex data, and at the same time enhancing the accuracy of human-computer collaborative decision-making, thereby improving overall treatment efficiency and reliability.

[0064] For example, in the treatment of a 28-year-old patient with intractable epilepsy, a neural activation heatmap was generated based on the patient's preoperative DTI brain map, with real-time overlay of EEG spike signals from the right middle temporal gyrus (MTG) and fNIRS blood oxygenation data. Initially, this region appeared large red (high activation) due to abnormal discharges, while the surrounding hippocampus was yellow (moderate activation). Upon observing the heatmap, the doctor found that the red area closely matched the location of the epileptogenic focus and fine-tuned the stimulation frequency to 12.5Hz. After 10 minutes of intervention, the heatmap showed that the red area in the MTG had shrunk by 60% and turned yellow, while the hippocampus turned green (low activation), visually confirming that the abnormal activation had been suppressed. This provided a visual basis for further optimization of stimulation parameters, and the patient's spike frequency subsequently decreased by 50%.

[0065] According to the closed-loop multimodal neural stimulation method based on time interference proposed in the embodiment of the present application, through the linkage of multimodal stimulation integration and multi-dimensional neural state perception, with the help of adaptive adjustment algorithm and deep learning timing prediction, a breakthrough is achieved in neural stimulation from single-modal fixed parameters to multimodal dynamic adaptation, effectively improving the stimulation accuracy and neural response matching degree; at the same time, the generation of time interference field and real-time feedback of the visual interactive platform not only enhances the adaptability to complex neural environments, but also reduces the impact of device performance drift and signal distortion, significantly improves energy utilization efficiency and system reliability, and comprehensively guarantees the safety and therapeutic effect of neural regulation. As a result, the problems of rigid adjustment strategy, low adjustment accuracy and insufficient energy conversion efficiency in the existing technology are solved.

[0066] The following will describe a closed-loop multimodal neural stimulation method based on time interference through a specific embodiment. Figure 9 Shown, including: A 35-year-old male patient with treatment-resistant depression was admitted to a mental health center. He had an eight-year history of treatment-resistant depression and a Hamilton Depression Rating Scale (HAMD) score of 28 (major depression). He presented with persistent low mood and loss of interest, and had previously responded to a combination of five antidepressant medications. Preoperative fMRI revealed a 20% decrease in functional connectivity between the patient's left dorsolateral prefrontal cortex (DLPFC) and the hippocampus compared with healthy controls. PET-CT scans revealed a 15% decrease in DLPFC metabolic rate, suggesting that multimodal neurostimulation could be appropriate for modulating the prefrontal-limbic system circuit.

[0067] Real-time multimodal neural state data were collected. EEG signals were collected using a 64-channel Brain Products ActiChamp device, covering the bilateral prefrontal and temporal lobes, with a sampling rate of 2000 Hz and a bandpass filter of 1-50 Hz. The focus was on monitoring the theta wave power (target ≥5 μV² / Hz), alpha wave power (target ≤3 μV² / Hz) and prefrontal-hippocampal synchronization index (target ≥0.6) of the left DLPFC. Blood oxygen concentration was collected using NIRx Medical Systems. The fNIRS device placed a 3×3 probe array (3 cm spacing) in the left DLPFC with a sampling rate of 7.81 Hz to monitor the rate of change of HbO2 (target fluctuation ≥ 10%). For neurometabolite collection, a microdialysis probe with a membrane length of 3 mm and a molecular weight cutoff of 10 kDa was implanted in the left hippocampus through minimally invasive surgery. A micropump with a flow rate of 2 μL / min was connected to perfuse 37°C artificial cerebrospinal fluid. Dialysate was collected every 15 minutes, and 5-HT (target ≥ 80 nM) and NE (target ≥ 60 nM) levels were detected by HPLC.

[0068] Data fusion and neural activation phase prediction: Kalman filtering was used to perform spatiotemporal alignment and noise filtering of EEG, blood oxygenation, and neurometabolite data. The EEG theta / alpha power ratio (weight 0.4), fNIRS HbO2 change rate (weight 0.3), and microdialysis 5-HT / NE concentration ratio (weight 0.3) were extracted as feature vectors to generate the "Prefrontal-Limbic Activity Index (PFAI)." Based on the PFAI sequence and neural activation records of 120 treatments over the previous 6 months, an LSTM model with 32 dimensions in the input layer and 64 nodes in the hidden layer was trained. The PFAI sequence in the first 15 seconds of input could predict the left DLPFC activation phase within the next 2 seconds: the activation phase entered the initial stage 800ms after stimulation, reached the peak at 1500ms (theta wave power was highest and HbO2 reached its peak), and began to subside at 2500ms, with a prediction error of ≤80ms.

[0069] Dynamically generate and adjust multimodal stimulation plans, with the initial plan based on the predicted phase setting: NeuroPulseTES device was used for time-interference electrical stimulation, with two pairs of electrodes (3 cm apart, 4 kΩ impedance) placed on the left DLPFC, outputting a dual-frequency current of 2.2 kHz and 2.207 kHz to generate a 7 Hz interference field with an intensity of 1.5 mA, which was triggered synchronously at the start of activation (800 ms); MagVentureMagProR30 was used for focused transcranial magnetic stimulation, with the coil fixed on the left DLPFC. The PFC was treated with iTBS (3 pulses / train, 50 Hz, 200 ms interval between trains) at an intensity of 85% of the motor threshold, initiated 200 ms before peak activation (1300 ms). Near-infrared optogenetic stimulation was initiated by preoperatively transfecting 5-HT neurons in the left hippocampus to express ChR2. A 200 μm optical fiber (numerical aperture 0.22) was implanted, connected to a 473 nm laser (0-8 mW), with an initial power of 5 mW. During the activation-dissipation phase (2500 ms), 10 Hz pulses were applied for 1 s. The DQN algorithm, with the goal of achieving a PFAI ≥ 0.6 for 5 minutes, inputted the current PFAI and 5-HT concentration, and outputted parameter adjustments (e.g., electrical stimulation intensity ±0.1 mA). For example, 3 minutes after initial stimulation, if the PFAI was 0.45 and 5-HT was 72 nM, the electrical stimulation intensity was adjusted to 1.6 mA, the frequency difference to 7.2 Hz, and the optical power to 5.4 mW, ultimately stabilizing the PFAI at 0.62.

[0070] Visual rendering and effect evaluation feedback are presented in real time on a 27-inch 4K touchscreen (response time 6ms). The left panel shows a 3D neural activation heatmap based on a fusion of preoperative MRI and DTI. Red indicates high activation (theta wave reaching target) in the left DLPFC, blue indicates low activation, and fNIRS HbO2 hyperperfusion areas (yellow gradient) are superimposed. The refresh rate is 1Hz. The right panel displays parameter adjustment curves (e.g., electrical stimulation frequency difference 7.2Hz→7.3Hz, intensity 1.6mA→1.7mA), magnetic stimulation iTBS timing diagrams, and light stimulation pulse curves. The bottom panel displays the PFAI (0.62), 5-HT concentration (82nM), HAMD score (14 points after 2 weeks), and NSI (0.85) in real time. After observing the expansion of the red area in the left DLPFC and the stabilization of the NSI above 0.8, the doctor manually fine-tuned the light power to 5.2mW. The system automatically records these parameters and provides feedback to the neural control center to optimize the next round of treatment.

[0071] In summary, the embodiment of the present application comprehensively captures the dynamic changes of the prefrontal-limbic system of patients with refractory depression through multimodal data acquisition (EEG, blood oxygen, and neurometabolites), accurately predicts the neural activation phase through data fusion and LSTM model, and enables multimodal stimulation (electric, magnetic, and optogenetic) to accurately trigger the activation onset, peak, and extinction phase of the left DLPFC, significantly improving the targeting and timing coordination of the stimulation; with the help of the DQN algorithm to dynamically adjust parameters (such as electrical stimulation intensity and frequency difference), adaptive optimization is achieved with the goal of PFAI ≥ 0.6, effectively improving the patient's left DLPFC Functional connectivity with the hippocampus was achieved (theta wave power reached the target, 5-HT concentration increased to 82nM), and the Hamilton Depression Rating Scale score dropped from 28 to 14 points within 2 weeks, significantly alleviating the symptoms of severe depression; at the same time, the visual interactive platform intuitively presents changes in neural activation through heat maps and parameter curves, supporting doctors to make precise fine-tuning (such as light power adjustment), reducing the risk of overstimulation, and realizing the transition from "empirical treatment" to "data-driven precise regulation" in refractory cases where drugs are ineffective, which not only improves treatment efficiency and effect stability, but also provides a scalable closed-loop solution for individualized neural regulation.

[0072] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. The electronic device may include: Memory 1001 , processor 1002 , and computer programs stored in the memory 1001 and executable on the processor 1002 .

[0073] When the processor 1002 executes the program, a closed-loop multimodal neural stimulation method based on time interference provided in the above embodiment is implemented.

[0074] Furthermore, the electronic device further includes: The communication interface 1003 is used for communication between the memory 1001 and the processor 1002 .

[0075] The memory 1001 is used to store computer programs that can be run on the processor 1002 .

[0076] The memory 1001 may include a high-speed RAM (Random Access Memory) memory, and may also include a non-volatile memory, such as at least one disk memory.

[0077] If the memory 1001, processor 1002, and communication interface 1003 are implemented independently, the communication interface 1003, memory 1001, and processor 1002 can be connected to each other via a bus and communicate with each other. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 10 Only one thick line is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0078] Optionally, in a specific implementation, if the memory 1001, the processor 1002 and the communication interface 1003 are integrated on a chip, the memory 1001, the processor 1002 and the communication interface 1003 can communicate with each other through an internal interface.

[0079] The processor 1002 may be a CPU (Central Processing Unit), or an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of the present application.

[0080] An embodiment of the present application further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-mentioned closed-loop multimodal neural stimulation method based on time interference.

[0081] In addition, an embodiment of the present application also provides a computer program product, including a computer program or instructions, which, when executed, implements the above-mentioned closed-loop multimodal neural stimulation method based on time interference.

[0082] In the description of this specification, reference to the terms "one embodiment," "some embodiments," "example," "specific example," or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in any appropriate manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification, as well as features of different embodiments or examples, unless they are mutually inconsistent.

[0083] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout the description of this application, "plurality" means at least two, for example, two, three, etc., unless otherwise specifically defined.

[0084] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present application belong.

[0085] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods can be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one of the following technologies known in the art or a combination thereof can be used: a discrete logic circuit having logic gate circuits for implementing logical functions on data signals, an application-specific integrated circuit having suitable combinational logic gate circuits, a programmable gate array (PGA), a field-programmable gate array (FPGA), etc.

[0086] A person skilled in the art can understand that all or part of the steps carried out in the method of the above embodiment can be completed by a program to instruct the relevant hardware, and the program can be stored in a computer-readable storage medium. When the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0087] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A closed-loop multimodal neural stimulation system based on time interference, characterized in that: include: Multimodal stimulation module, neural state perception module, neural control center, timing coordination engine, and visual interaction platform; among them, The multimodal stimulation module is used to integrate electrical stimulation, magnetic stimulation and optogenetic stimulation to generate a temporal interference field; The neural state perception module is used to collect real-time EEG signals, blood oxygen concentration and neural metabolite level data; The neural control center is used to fuse neural state data and stimulation parameters, and dynamically adjust the time interference frequency and stimulation intensity through an adaptive algorithm; The timing collaboration engine predicts the neural response phase based on the deep learning model and optimizes the stimulation timing and modality switching strategy; The visualization interaction platform is used to render neural activation heat maps and stimulation parameter adjustment curves in real time.

2. A closed-loop multimodal neural stimulation system based on time interference according to claim 1, characterized in that: The multimodal stimulation module includes: a temporal interferometric electrical stimulation unit, a focused transcranial magnetic stimulation unit, and a near-infrared optogenetic regulation unit; wherein, the temporal interferometric electrical stimulation unit generates a low-frequency interference field by superimposing dual-frequency currents, targeting deep nuclei; the focused transcranial magnetic stimulation unit uses pulse sequence modulation to enhance the synergy of the cortical-subcortical pathway; the near-infrared optogenetic regulation unit activates specific neuronal subpopulations through laser-coupled optical fibers.

3. A closed-loop multimodal neural stimulation system based on time interference according to claim 1, characterized in that: The neural state perception module includes: a high-density EEG acquisition array, a functional near-infrared spectroscopy module, and a microdialysis sampling unit; wherein, the high-density EEG acquisition array is used to capture the synchronization of neuronal cluster discharges in real time; the functional near-infrared spectroscopy module is used to monitor changes in blood oxygen saturation; and the microdialysis sampling unit is used to detect fluctuations in the concentrations of glutamate and γ-aminobutyric acid neurotransmitters.

4. A closed-loop multimodal neural stimulation system based on time interference according to claim 1, characterized in that: The neural control center includes: a data fusion processor and an adaptive adjustment algorithm module; wherein, the data fusion processor fuses multi-source neural signals through Kalman filtering to extract feature vectors; the adaptive adjustment algorithm module is based on the DQN algorithm, with the goal of minimizing neural state errors, and dynamically corrects the time interference frequency and stimulation intensity.

5. The closed-loop multimodal neural stimulation system based on time interference according to claim 1, characterized in that: The timing collaborative engine includes: a neural response prediction model and a modal switching controller; wherein, the neural response prediction model uses a long short-term memory network to predict the neural activation phase based on historical stimulation and response data; the modal switching controller quickly switches between electrical stimulation, magnetic stimulation, and light stimulation according to the prediction results.

6. A closed-loop multimodal neural stimulation system based on time interference according to claim 1, characterized in that: The visualization interactive platform includes: a three-dimensional neural activation rendering unit, a stimulation parameter adjustment interface, and an efficacy evaluation dashboard; wherein, the three-dimensional neural activation rendering unit generates a dynamic activation heat map based on diffusion tensor imaging, superimposing real-time EEG and blood oxygen data; the stimulation parameter adjustment interface is used to manually fine-tune the time interference parameters; the efficacy evaluation dashboard is used to quantify the neural state stability index through the entropy method to evaluate the intervention effect.

7. A method for a closed-loop multimodal neural stimulation system based on time interference according to any one of claims 1 to 6, characterized in that: The method comprises: Collect real-time EEG signals, blood oxygen concentration, and neurometabolite level data; The EEG signal, blood oxygen concentration, and neural metabolite level data are fused and processed, feature vectors are extracted through Kalman filtering, and an LSTM model is trained based on historical stimulus response data to predict neural activation phases; Based on the neural activation phase prediction results, a multimodal temporal interference stimulation scheme is generated, the initial frequency difference and stimulation intensity are set, and the temporal interference parameters and modality switching timing are dynamically adjusted using the DQN algorithm by comparing the actual neural response with the expected target in real time; The adjusted time interference parameters and mode switching timing are synchronously rendered to the neural activation heat map and parameter adjustment curve to quantitatively evaluate the intervention effect and provide feedback.

8. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement a closed-loop multimodal neural stimulation method based on time interference according to claim 7.

9. A computer-readable storage medium having a computer program or instruction stored thereon, characterized in that: When the computer program or instructions are executed, the closed-loop multimodal neural stimulation method based on time interference of claim 7 is implemented.

10. A computer program product comprising a computer program or instructions, characterized in that When the computer program or instructions are executed, the closed-loop multimodal neural stimulation method based on time interference of claim 7 is implemented.

Images