Multi-mode closed-loop feedback thumbtack needle electrical stimulation anti-depression system

Abstract

The invention discloses a multi-mode closed-loop feedback thumbtack needle electrical stimulation anti-depression system, belongs to the technical field of drug delivery systems, and particularly relates to collaborative treatment of acupoint stimulation and drug sustained release. The problems that in the prior art, cooperation of acupoint stimulation and medicine slow release cannot be achieved, and further improvement of the treatment effect is limited are solved. The system comprises an AI regulation and control module and a collaborative treatment module. The AI regulation and control module predicts treatment parameters; and the collaborative treatment module is used for carrying out collaborative treatment of acupoint electrical stimulation and skin administration on the patient based on the predicted treatment parameters. The multi-mode closed-loop feedback thumbtack needle electrical stimulation anti-depression system is suitable for collaborative treatment of acupoint stimulation and drug sustained release.

Description

Technical Field

[0001] This invention relates to the field of drug delivery system technology, and more particularly to the synergistic treatment of acupoint stimulation and sustained drug release. Background Technology

[0002] Depression is a mental illness with a high relapse rate and a high disability rate. Current treatments for depression mainly include medication and physical therapy.

[0003] in: Drug therapy primarily utilizes drug-loaded microspheres. However, existing methods for preparing drug-loaded microspheres suffer from problems such as uneven particle size, high drug burst release rates, and low encapsulation efficiency, leading to significant side effects and easy drug resistance in current drug therapies.

[0004] Physical therapy mainly includes traditional acupuncture (using traditional press needles). Traditional acupuncture has a high effectiveness rate in treating depression (53.3%~85.1%), with a low recurrence rate and few adverse reactions. However, it has the problem of high operator dependence: the efficacy is limited by the operator's experience and individual differences. Professionals are needed to accurately locate acupoints, making it unsuitable for ordinary patients to perform on their own, and thus difficult to popularize.

[0005] In addition, there is a lack of synergy between existing drug therapy and physical therapy techniques. Traditional press needles only contain a single metal needle and can only physically stimulate acupoints, while traditional drug-loaded microspheres can only provide sustained drug release. In short, existing technologies cannot achieve synergy between acupoint stimulation and sustained drug release, which limits further improvement in treatment efficacy. Summary of the Invention

[0006] This invention proposes a multimodal closed-loop feedback press needle electrostimulation antidepressant system, which solves the problem that existing technologies cannot achieve synergy between acupoint stimulation and drug sustained release, thus limiting further improvement in treatment efficacy.

[0007] The multimodal closed-loop feedback press needle electrical stimulation antidepressant system of the present invention includes the following modules: Patient profiling module: Collects patient personal information and builds patient profiles; Multi-source physiological data acquisition module: Collects multi-source physiological data from patients; Sleep data acquisition module: Employs multimodal sensor fusion technology to collect patient sleep data through a non-invasive wearable device; Sleep quality analysis module: Obtains sleep quality data based on patient sleep data analysis; Effect feedback information collection module: Collects patient effect feedback information for each acupoint electrical stimulation and skin drug administration; AI regulation module: Generates a synergistic antidepressant plan of acupoint electrical stimulation and skin drug delivery based on patient profile, multi-source physiological data, sleep quality data and treatment effect feedback information, and predicts treatment parameters; the treatment parameters include acupuncture points and electrical stimulation parameters for acupoint electrical stimulation and drug release rate for skin drug delivery; Synergistic Therapy Module: Based on predicted treatment parameters, patients receive synergistic therapy involving acupoint electrical stimulation and transdermal drug delivery.

[0008] Furthermore, a preferred embodiment is provided in which the multi-source physiological data includes the patient's electroencephalogram (EEG), electrocardiogram (ECG), skin conductance (GSR), and heart rate data.

[0009] Furthermore, a preferred embodiment is provided, wherein the patient sleep data includes the patient's body movement frequency during sleep, the patient's blood oxygen saturation and pulse rate variability, environmental temperature and humidity data, and environmental noise data.

[0010] Furthermore, a preferred embodiment is provided, wherein the sleep quality data includes: sleep stage data, sleep apnea event data, and sleep statistics; the sleep statistics include sleep efficiency, sleep latency, REM sleep percentage, and number of awakenings.

[0011] Furthermore, a preferred embodiment is provided, wherein the sleep quality analysis module includes: a sleep staging model, a sleep apnea identification model, and a sleep data statistics unit; The sleep staging model is a random forest classifier: it distinguishes sleep stages based on body movement frequency analysis to obtain sleep stage data; The sleep apnea recognition model is an LSTM network: based on the synchronous monitoring results of the patient's blood oxygen saturation and pulse rate variability during sleep, sleep apnea events are identified and sleep apnea event data is obtained; The sleep data statistics unit calculates sleep efficiency, sleep latency, REM sleep percentage, and number of awakenings based on the patient's sleep data.

[0012] Furthermore, a preferred embodiment is provided, wherein the AI ​​control module includes a control software device, an edge computing device, a cloud platform, and a user interaction terminal; The control software device includes an AI model, a storage unit, a training unit, and an interactive analysis unit; The AI ​​model is integrated into an edge computing device and is used to predict treatment parameters; The storage unit, training unit, and interactive analysis unit are integrated in the cloud platform; the storage unit is used to store data; the training unit is used to train AI models; and the interactive analysis unit is used to interact with users through user interaction terminals.

[0013] Furthermore, a preferred embodiment is provided in which the AI ​​model uses reinforcement learning and online learning mechanisms to predict treatment parameters; The reinforcement learning involves learning the optimal treatment parameters by continuously adjusting treatment parameters based on patient profiles, multi-source physiological data, sleep quality data, and treatment effect feedback. The online learning involves continuously updating and learning the AI ​​model based on multi-source physiological data, sleep quality data, and treatment effect feedback.

[0014] Furthermore, a preferred embodiment is provided in which the training unit uses few-shot learning, transfer learning, federated learning, and encrypted learning mechanisms to train the AI ​​model.

[0015] Furthermore, in a preferred embodiment, the interactive analysis unit includes an AI interpretability component; the user interaction terminal includes a doctor's mobile terminal. The AI ​​explainability component integrates SHAP and LIME technologies to provide explanations for the AI ​​model's decisions and pushes the explanation results to doctors via their mobile devices.

[0016] Furthermore, a preferred embodiment is provided, wherein the synergistic treatment module includes an electrical stimulation device, a microneedle triggering device, and a drug-loaded microneedle patch; The drug-loaded microneedle patch includes: a group of medical stainless steel main needles, a group of hydrogel drug-loaded auxiliary needles, and adhesive tape. The adhesive tape is thicker in the middle and thinner around the edges; the medical stainless steel main needle group and the hydrogel drug-loaded auxiliary needle group are fixed in the middle area of ​​the adhesive tape. The medical stainless steel main needle group includes 4 medical stainless steel main needles; the 4 medical stainless steel main needles are symmetrically distributed around the center of the adhesive tape; The hydrogel drug-loaded needle array is made of hydrogel material; the hydrogel drug-loaded needle array includes multiple drug-loaded needles, which are drug-loaded microneedles; the tips of the drug-loaded microneedles are uniformly distributed with drug-loaded microspheres; the drug-loaded microspheres are pH-responsive microspheres, and their drug release rate is regulated by changes in the pH value of the skin surface. The hydrogel drug-loaded auxiliary needle group is divided into 4 regions, each region is arranged in an octagonal array; the 4 regions are distributed around the periphery of the medical stainless steel main needle group, forming the shape of 4 petals of a lilac flower; the needle height of the medical stainless steel main needle group is greater than the needle height of the hydrogel drug-loaded auxiliary needle group. The medical stainless steel main needle group is electrically connected to the electrical stimulation device and receives electrical stimulation pulse current; The hydrogel drug-loaded accessory needle group is equipped with electrodes for electrical connection with the microneedle triggering device; The electrical stimulation device is signal-connected to the AI ​​control module and is used to transmit electrical stimulation pulse current to the medical stainless steel main needle group according to the electrical stimulation parameters of the acupoints predicted by the AI ​​control module. The medical stainless steel main needle group is used to insert into the predicted acupoints on the patient's body for acupoint electrical stimulation, thereby performing acupoint electrical stimulation on the patient. The microneedle triggering device is signal-connected to the AI ​​control module and is used to conduct a drug release triggering current to the hydrogel microneedles according to the drug release rate predicted by the AI ​​control module for skin administration. The drug release triggering current is used to regulate the pH value of the skin surface through electrolysis, thereby regulating the drug release rate of the drug-loaded microspheres.

[0017] The present invention has the following beneficial effects: The multimodal closed-loop feedback press needle electrostimulation antidepressant system described in this invention predicts treatment parameters through an AI control module and uses drug-loaded microneedle patches as a synergistic treatment module, achieving synergistic treatment of acupoint stimulation and drug sustained release.

[0018] The multimodal closed-loop feedback press needle electrostimulation antidepressant system described in this invention is suitable for synergistic treatment of acupoint stimulation and drug sustained release. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of the structure of a multimodal closed-loop feedback press needle electrical stimulation antidepressant system in one embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a microfluidic chip in one embodiment of the present invention; Figure 3 This is a schematic diagram of the distribution structure of the stainless steel main needle group and the hydrogel drug-loaded auxiliary needle group in a drug-loaded microneedle patch according to one embodiment of the present invention. Figure 4 This is a side view of the structure of a drug-loaded microneedle patch in one embodiment of the present invention; Figure 5 This is a schematic diagram of a drug-loaded microneedle patch according to one embodiment of the present invention; Figure 6 This is a schematic diagram of a conventional snap needle (excluding adhesive tape) in one embodiment of the present invention. Figure 7 This is a schematic diagram of a conventional snap needle (including adhesive tape) in one embodiment of the present invention.

[0021] Figures 8 to 10 In one embodiment of the present invention, a high-performance liquid chromatogram is provided; wherein, Figure 8 This is a high-performance liquid chromatogram of a pharmaceutical standard. Figure 9 The image shows the high-performance liquid chromatogram of the blank microspheres. Figure 10 This is a high-performance liquid chromatogram of drug-loaded microspheres; Figure 11 This is a schematic diagram of rat acupoint location in one embodiment of the present invention; Figure reference numerals: a, main channel; b, first channel; c, second channel; d, drug-loaded microspheres; 1, medical stainless steel main needle group; 2, hydrogel drug-loaded auxiliary needle group; 3, limiting and fixing layer. Detailed Implementation

[0022] To make the technical solutions and advantages of the present invention clearer, the specific embodiments of the present invention will be described in further detail and completely below with reference to the accompanying drawings. The various embodiments described below are only some preferred embodiments of the present invention, and not all of them; the various embodiments described below are intended to explain the present invention and should not be construed as limiting the present invention; reasonable combinations of the technical features defined in the various embodiments of the present invention, as well as all other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort, are all within the scope of protection of the present invention.

[0023] Implementation Method 1: A multimodal closed-loop feedback press needle electrical stimulation antidepressant system, the system comprising the following modules: Patient profiling module: Collects patient personal information and builds patient profiles; Multi-source physiological data acquisition module: Collects multi-source physiological data from patients; Sleep data acquisition module: Employs multimodal sensor fusion technology to collect patient sleep data through a non-invasive wearable device; Sleep quality analysis module: Obtains sleep quality data based on patient sleep data analysis; Effect feedback information collection module: Collects patient effect feedback information for each acupoint electrical stimulation and skin drug administration; AI regulation module: Generates a synergistic antidepressant plan of acupoint electrical stimulation and skin drug delivery based on patient profile, multi-source physiological data, sleep quality data and treatment effect feedback information, and predicts treatment parameters; the treatment parameters include acupuncture points and electrical stimulation parameters for acupoint electrical stimulation and drug release rate for skin drug delivery; Synergistic Therapy Module: Based on predicted treatment parameters, patients receive synergistic therapy involving acupoint electrical stimulation and transdermal drug delivery.

[0024] In this embodiment, the acupuncture point for the acupoint electrical stimulation may be a single acupoint (such as Neiguan acupoint) or a combination of multiple acupoints (such as the synergistic effect of Baihui acupoint + Taichong acupoint).

[0025] In this embodiment, based on the patient profile, the AI ​​regulation module can mine potential features in the data (such as the patient's lifestyle and psychological state) to improve the model's predictive ability.

[0026] In this embodiment, based on patient profiles and collected multi-source physiological data, the AI ​​regulation module can generate personalized treatment recommendations based on each patient's health status, medical history, reactions, and physiological data.

[0027] In this embodiment, the AI ​​control module implements a closed-loop integrated dynamic treatment strategy: If REM sleep percentage is detected to be <15% for 3 consecutive days, the drug release rate of the skin delivery device (such as drug-loaded microspheres) will be increased by 20%. If sleep efficiency is less than 65%, increase the intensity of acupoint electrical stimulation (using low-frequency pulsed current) to enhance the intensity of electrical stimulation (in 0.5mA increments, with an upper limit of 5mA). If the Apnea Index (AHI) is greater than 15 times per hour, an alert can be sent to the doctor via mobile devices (doctor's terminal) and suggestions for combined CPAP treatment can be provided.

[0028] In this embodiment, the multimodal closed-loop feedback press-needle electrical stimulation antidepressant system aims to optimize the treatment effect of depression through intelligent electrical stimulation technology. This system combines modern neuromodulation and real-time physiological monitoring technologies, achieving personalized and precise electrical stimulation therapy through a closed feedback loop.

[0029] In another embodiment, the patient's personal information includes medical history, genomic information (such as 5-HTTLPR genotype), and lifestyle information.

[0030] In this embodiment, the lifestyle information includes patient activity trajectory information (collected via mobile phone / GPS) and voice emotion analysis information (NLP).

[0031] In this embodiment, by constructing a patient profile, the AI ​​control module can customize a personalized treatment plan for each patient. By uncovering the potential characteristics within these data (such as patients' lifestyles and psychological states), we can improve the predictive capabilities of AI models.

[0032] In another embodiment, the multi-source physiological data includes the patient's electroencephalogram (EEG), electrocardiogram (ECG), skin conductance (GSR), and heart rate data.

[0033] In this embodiment, the multi-source physiological data is obtained by standardizing and augmenting the raw data collected by the sensors: Data standardization: Ensure that data from different sources are in the same format to facilitate subsequent analysis and processing; Data augmentation: If the actual amount of data collected is insufficient, more samples can be generated through synthetic data techniques (such as GAN) to enhance the robustness of the AI ​​control module (AI model).

[0034] In this embodiment, the multi-source physiological data acquisition module includes a heart rate detection device; The heart rate detection device: calculates and collects heart rate data, including real-time heart rate and heart rate variability, by monitoring changes in blood flow to the wrist.

[0035] In this embodiment, the heart rate detection device monitors changes in wrist blood flow for 10 minutes in order to calculate real-time heart rate (HR) and heart rate variability (HRV).

[0036] In this embodiment, the heart rate detection device uses a photoplethysmography (PPG) device, equipped with a green LED light source (wavelength 520 nm) and a silicon photodiode (response wavelength 400-1100 nm).

[0037] In another embodiment, the patient sleep data includes the patient's body movement frequency during sleep, the patient's blood oxygen saturation, the patient's pulse rate variability, environmental temperature and humidity data, and environmental noise data.

[0038] In this embodiment, the sleep data acquisition module includes a triaxial accelerometer, an improved PPG module, a temperature and humidity sensor, and a noise sensor; The triaxial accelerometer integrates a MEMS accelerometer for collecting body motion frequencies; The improved PPG module is used to simultaneously monitor the patient's blood oxygen saturation and pulse rate variability during sleep. The temperature and humidity sensor is used to collect ambient temperature and humidity data during sleep. The noise sensor is used to collect ambient noise data during sleep.

[0039] In this embodiment, the sampling frequency of the MEMS accelerometer is 50 Hz.

[0040] In this embodiment, the triaxial accelerometer collects the body motion frequency, which is obtained by preprocessing the raw signal collected by the triaxial accelerometer; the preprocessing includes: The acquired raw signal was filtered by Butterworth low-pass filter (cutoff frequency 10 Hz), and the PPG signal was denoised using wavelet transform.

[0041] In this embodiment, the improved PPG module is an upgraded photoplethysmography configuration, adding dual light sources of infrared LED (wavelength 850 nm) and red LED (wavelength 660 nm) to simultaneously monitor the patient's blood oxygen saturation (SpO2) and pulse rate variability (PRV) during sleep.

[0042] In this embodiment, the temperature and humidity sensor has an accuracy of ±0.5℃.

[0043] In this embodiment, the noise detection range of the noise sensor is 30-100 dB.

[0044] In another embodiment, the sleep quality data includes: sleep stage data, sleep apnea event data, and sleep statistics; the sleep statistics include sleep efficiency, sleep latency, REM sleep percentage, and number of awakenings.

[0045] In this embodiment, the sleep quality analysis module includes: a sleep staging model, a sleep apnea identification model, and a sleep data statistics unit; The sleep staging model is a random forest classifier: it distinguishes sleep stages based on body movement frequency analysis to obtain sleep stage data; The sleep apnea recognition model is an LSTM network: based on the synchronous monitoring results of the patient's blood oxygen saturation and pulse rate variability during sleep, sleep apnea events are identified and sleep apnea event data is obtained; The sleep data statistics unit calculates sleep efficiency, sleep latency, REM sleep percentage, and number of awakenings based on the patient's sleep data.

[0046] In this embodiment, the sleep efficiency is: total sleep time / time spent in bed × 100%.

[0047] In this embodiment, the sleep latency period is the time required from lying down to falling asleep, also known as the sleep latency period. In this embodiment, the REM sleep percentage is determined by combining PPG pulse morphology with acceleration frequency domain analysis.

[0048] In this embodiment, the number of awakenings is defined as follows: a body movement amplitude > 0.5g and a duration > 5 seconds is counted as one awakening.

[0049] In this embodiment, the patient's sleep data, after data preprocessing, is input into the sleep quality analysis module: Let the original sleep monitoring dataset (patient sleep data be D={x}) be... i ,y i}i=1; Data preprocessing (data cleaning process): (1) Handling missing values:

[0050] Delete records containing missing values ​​to ensure data integrity.

[0051] (2) Outlier correction:

[0052] Where μ j ,σ j Let be the mean and standard deviation of feature j, respectively.

[0053] In this embodiment, the sleep data statistics unit extracts features from the preprocessed patient sleep data and obtains sleep statistics data based on the extracted feature vectors. Extracted feature vector: x i =[t lat ,T total , n awake ,r REM ]; Among them, t lat Sleep latency (min); T total Total sleep time (h); nawake: number of awakenings; r REM REM ratio (%); y i ∈{0,1}: Improvement marker label.

[0054] In this embodiment, the sleep stages include wakefulness, light sleep, deep sleep, and REM sleep.

[0055] In this embodiment, body motion frequency analysis (ACT) is performed as follows: Body movement frequency refers to the number of times a person turns over and the slight movements of their limbs at night. Sleep stages are distinguished based on body movement frequency.

[0056] In this embodiment, the accuracy of the sleep staging model (random forest classifier) ​​is >92%.

[0057] In this embodiment, the sleep staging model is deployed using TensorFlow Lite to achieve lightweight edge computing with a single inference power consumption of <3mAh, meeting the requirements for 72 hours of continuous monitoring.

[0058] In this embodiment, the input features of the random forest classifier include body energy entropy, PRV low-frequency / high-frequency power ratio, and SpO2 descent slope.

[0059] In this embodiment, the random forest classifier: Model definition: Let the decision tree set T = {Tb} b=1 Each tree T b The prediction function is: h b (x): R4→{0,1}; Integrated prediction: y =mode({h b (x)} b=1 ).

[0060] Training optimization goal: (1) Node splitting criterion (minimization of Gini impurity):

[0061] Where, p c =∣S c | / |S| represents the proportion of category c in node S.

[0062] (2) Category balance constraint:

[0063] Where Nc is the number of samples in category c; Complexity control: Maximum tree depth D max =5, number of trees B=150.

[0064] In this embodiment, the sleep quality analysis module undergoes model validation and performance evaluation: Stratified sampling strategy: Let the original dataset class distribution be p(y=1)=α p ( y =1)= α Then the validation set is constructed as follows:

[0065] Where I is a random index set, |I| = 0.2N.

[0066] Accuracy calculation:

[0067] Feature importance quantification:

[0068] N b S: The set of all nodes of tree b; t : The sample set of node t; S tL ,S tR The left and right subsets after splitting.

[0069] In this embodiment, the apnea recognition model (LSTM network) has a sensitivity of ≥88% in recognizing apnea events (AHI index).

[0070] In this embodiment, the sleep apnea recognition model (LSTM network) also recognizes periodic limb movements (PLMS).

[0071] In this embodiment, sleep quality data is encrypted with AES-256 and then stored in segments.

[0072] In another embodiment, the sleep quality analysis module further includes a sleep quality trend visualization model.

[0073] The sleep quality trend visualization model: Input: Time series dataset {(d k ,T k ,r k )} k=1 (dk: number of monitoring days) Function: Using a dual-axis trend chart generation algorithm, generate the following curves and indicators: (1) Sleep duration trend curve: f T (d)=LOESS({(d k T k )},span=0.3) Locally weighted regression smoothing (LOESS) was used.

[0074] (2) REM ratio trend curve: f T (d) = Moving Average({(d) k r k )}, window=5).

[0075] (3) Visual indicators

[0076] Among them, w T =0.6, w r =0.4 is the weighting coefficient.

[0077] In this embodiment, through multi-sensor collaboration, body movement, blood oxygen, and environmental parameters are integrated for the first time for closed-loop management of sleep disorders in depression. In this embodiment, the system breaks through the one-way stimulation mode of traditional devices and realizes a three-way feedback regulation of "sleep quality-electric stimulation-drug release".

[0078] In this embodiment, the sleep data acquisition module and the sleep quality analysis module improve the physiological monitoring dimensions of the system, provide key biomarker data for the AI ​​model (AI regulation module), and further enhance the personalization and dynamic optimization capabilities of the treatment plan.

[0079] In another embodiment, the effect feedback information includes patient self-reports, medical records, and physiological change data.

[0080] In this embodiment, the patient self-report is obtained through a patient self-assessment scale.

[0081] In this implementation, feedback information is obtained from data such as patient self-reports, medical records, and physiological changes, and the AI ​​control module (AI model) can learn the effects of different treatment methods.

[0082] In another embodiment, the AI ​​control module includes a control software device, an edge computing device, a cloud platform, and a user interaction terminal. The control software device includes an AI model, a storage unit, a training unit, and an interactive analysis unit; The AI ​​model is integrated into an edge computing device and is used to predict treatment parameters; The storage unit, training unit, and interactive analysis unit are integrated in the cloud platform; the storage unit is used to store data; the training unit is used to train AI models; and the interactive analysis unit is used to interact with users through user interaction terminals.

[0083] In this embodiment, the AI ​​model is integrated into the edge computing device to improve computing efficiency and ensure that the AI ​​model can process data in real time.

[0084] In this embodiment, the edge computing device can be replaced with an embedded AI chip, and the AI ​​model can be integrated into the embedded AI chip. This can also process data in real time and provide instant feedback, ensuring system response speed.

[0085] In this embodiment, the AI ​​model is adapted to the needs of embedded devices, and has high computing performance and low latency.

[0086] In another embodiment, the AI ​​model uses reinforcement learning and online learning mechanisms to predict treatment parameters; The reinforcement learning involves learning the optimal treatment parameters by continuously adjusting treatment parameters based on patient profiles, multi-source physiological data, sleep quality data, and treatment effect feedback. The online learning involves continuously updating and learning the AI ​​model based on multi-source physiological data, sleep quality data, and treatment effect feedback.

[0087] In this embodiment, the treatment effect feedback information constitutes a real-time feedback mechanism (loop), including the patient's real-time response (such as symptom improvement), and based on this, combined with reinforcement learning and online learning mechanisms, the treatment plan is dynamically adjusted.

[0088] In this embodiment, the real-time adjustment of the treatment process is achieved through reinforcement learning (RL) algorithms, which optimize the treatment strategy based on real-time feedback. Reinforcement learning can gradually learn the optimal treatment plan (the best combination of treatment parameters) by continuously adjusting treatment parameters (acupuncture points, electrical stimulation parameters, and drug release rate) (i.e., a process of continuous trial and error and optimization).

[0089] In this implementation, reinforcement learning constructs a decision loop using a four-tuple of State, Action, Reward, and Next State, with the core objective of learning a long-term optimal policy. Key data used is as follows: Patient profile data: Purpose: To define the initial state, such as static characteristics like age, gender, history of chronic diseases, and history of drug allergies; Example: If a patient with depression also has sleep apnea (AHI>15), the state space needs to be marked "requires combined CPAP treatment" to affect action selection.

[0090] Multi-source physiological data (such as heart rate, blood pressure, blood oxygen): Purpose: To dynamically update the state, reflecting real-time changes in physiological state; Example: The RL model triggers an action to "increase electrical stimulation intensity" based on a sudden drop in blood oxygen (SpO2 < 90%) to avoid hypoxic events.

[0091] Sleep quality data (sleep stages, number of awakenings, AHI index): use: Status characteristics: Quantify sleep depth (e.g., REM sleep percentage <15% is abnormal); Reward function: Improved sleep efficiency → positive reward; Increased AHI index → ​​negative penalty.

[0092] Feedback information on effectiveness (pain score and symptom relief after stimulation / administration): use: Reward signal: Immediately quantifies the effect of therapeutic actions (e.g., pain score decrease = +R); Strategy optimization: If electrical stimulation of a certain acupoint causes skin allergy (feedback = -R), RL will reduce the probability of selecting that action.

[0093] Treatment parameter action recording (electrical stimulation parameters, drug dosage): use: Define the action space, such as "0.5mA low-frequency stimulation" or "microsphere release of drug X 10mg".

[0094] In this embodiment, based on an online learning mechanism, the AI ​​model can update and learn itself according to the patient's real-time response, ensuring continuous optimization of the treatment plan.

[0095] In this implementation, the model is updated regularly based on an online learning mechanism and continuously optimized with new data to ensure that the model is suitable for different types of depression and different populations.

[0096] In this embodiment, the combination of reinforcement learning and online learning mechanisms ensures that the treatment is continuously optimized in application, while the feedback mechanism provides continuous learning opportunities.

[0097] In another embodiment, the training unit uses few-shot learning, transfer learning, federated learning, and encrypted learning mechanisms to train the AI ​​model.

[0098] In this embodiment, the few-shot learning refers to training a model with strong generalization ability using a very small amount of labeled data (e.g., ≤5 samples per category), with meta-learning as its core. Classifying depression subtypes requires a large amount of labeled data, but initial clinical data is scarce; few-shot learning is used to address this data scarcity problem.

[0099] The role of the training phase: Support Set: A small number of samples guide the model to learn a "similarity measure".

[0100] Query Set: Used to evaluate the model's generalization ability.

[0101] Relationship with transfer learning: Pre-trained models that rely on transfer learning provide basic feature representations.

[0102] In this embodiment, the transfer learning improves the learning efficiency and performance of the target domain task by reusing the knowledge (such as feature representation) learned by the pre-trained model in the source domain.

[0103] The role of the training phase: Pre-training: The base model is trained using large-scale datasets (such as ImageNet or Med-PaLM medical large model). The data sources used for pre-training are implemented using data generation and augmentation techniques: multimodal physiological signal data is synthesized using generative adversarial networks (GANs) (which must meet the FDA's synthetic data validation standards).

[0104] Fine-tuning: Adjusting model parameters using small sample data from the target domain (such as physiological signals of depression) to adapt to new tasks, i.e., feature transfer.

[0105] In this embodiment, federated learning is a distributed collaborative training framework in which multiple clients (such as hospital / patient devices) train the model locally and only upload model parameter updates (not the original data), while the server aggregates the global model.

[0106] The role of the training phase: Local training: Each client updates its local model using private data; Parameter aggregation: The server integrates and updates to generate a global model, which is then iterated over in a loop.

[0107] Privacy protection mechanism: For sensitive medical data, the data must not leave the local area, in compliance with regulations such as GDPR / HIPAA.

[0108] Combine encryption technology (such as isomorphic encryption) to prevent parameter leakage.

[0109] In summary, the federated learning described above: Sharing learning results among different patients improves the model's generalization ability; At the same time, ensure that the data does not leave the patient's end, and update the model through training to ensure privacy and protect patient privacy.

[0110] AI models can be trained across institutions without leaking the original data within a federated learning framework.

[0111] In this embodiment, the encrypted learning: a set of technical tools (non-independent learning paradigm) is embedded in the training process (such as parameter encryption in federated learning) to ensure data security during the training process.

[0112] Commonly used techniques include homomorphic encryption (HE), differential privacy (DP), and secure multi-party computation (MPC).

[0113] Homomorphic encryption (HE): Gradients are calculated in ciphertext, which is suitable for preventing leakage of parameter exchange in federated learning during transmission.

[0114] Differential privacy (DP): Adding noise makes individual data points unidentifiable, suitable for model publishing or data sharing.

[0115] Secure Multi-Party Computation (MPC): Multi-party joint computation of encrypted data, suitable for cross-institutional joint modeling where only results are output.

[0116] In this embodiment, cryptographic learning (or cryptographic technology) is a type of privacy-preserving algorithm that ensures data security by using cryptographic technology during data transmission and storage.

[0117] In this implementation, few-shot learning is used to address the problem of scarce labeled samples; transfer learning is used to address the problem of insufficient data / domain differences; federated learning is used to address the problems of data silos and privacy compliance; and cryptographic learning is used to address the problems of transmission and storage security.

[0118] In this embodiment, an example of technical collaboration is as follows: AI models are trained using federated learning: federated learning is used to collaboratively train AI models across multiple centers. Each local center trains its own model and uploads the parameters in encrypted form. The server aggregates the data to generate a global model, ensuring that the original data never leaves the local machine.

[0119] Combining transfer learning and few-shot learning: Transfer learning is used to initialize the global model (such as loading Med-PaLM feature layers) and solve the cold start problem.

[0120] Few-shot learning is used to adapt patient-specific data (such as rare subtypes) locally on the client side, and then fine-tunes it rapidly through meta-learning.

[0121] Privacy-preserving algorithms (cryptographic learning) are embedded throughout the process: Homomorphic encryption protects parameter transmission in federated learning, while differential privacy adds noise during model deployment to meet GDPR / HIPAA requirements.

[0122] In another embodiment, the interaction analysis unit includes an AI interpretability component; the user interaction terminal includes a doctor's mobile terminal. The AI ​​explainability component integrates SHAP and LIME technologies to provide explanations for the AI ​​model's decisions and pushes the explanation results to doctors via their mobile devices.

[0123] In this embodiment, the AI ​​interpretability component is: Explanation toolsets: technical methods for revealing the decision-making logic of models (such as SHAP, LIME).

[0124] Core objective: To increase model transparency, enabling doctors or users to understand "why the AI ​​made this decision," thereby increasing trust.

[0125] Regulatory basis: Meeting the requirements of regulatory agencies FDA / NMPA for the traceability (or explainability) of medical AI decisions.

[0126] In this embodiment, the AI ​​interpretability component is: SHAP (Shapley Additive Explanations) value analysis: quantifies the contribution of each feature to the decision and shows key decision factors (such as "choosing 10Hz electrical stimulation due to reduced heart rate variability"). LIME (Local Proxy Model): Constructs simplified, interpretable models that approximate the local behavior of complex models; Attention mechanism: Visualizing the data region that the model focuses on (such as abnormal waveforms in an electroencephalogram); Output formats: Feature importance heatmap, Natural Language Decision Report.

[0127] In another embodiment, the interactive analysis unit includes an AI risk management component; The AI ​​risk management component ensures that the model's behavior complies with ethical, legal (ISO / IEC 23894 standard) and business requirements, and prevents and controls risks.

[0128] In this embodiment, the AI ​​risk management component: Governance Framework: A systematic risk management system covering the entire lifecycle of the model.

[0129] Core objective: To ensure that the model's behavior complies with ethical, legal, and business requirements, and to prevent and control risks.

[0130] Regulatory basis: Complies with the systematic requirements for AI risk management in standards such as ISO / IEC 23894.

[0131] In this embodiment, the AI ​​risk management component: Core components: Real-time deviation monitoring: Detects data distribution drift (such as sensor data offset) or model performance degradation (such as a decrease in prediction accuracy). Risk Cards: Structured records of risk types (such as privacy breaches, amplification of bias), impact levels, and corresponding strategies; Dynamic compliance engine: Embedded with the ISO / IEC 23894 standard, automatically verifying whether the model output complies with regulatory requirements (such as HIPAA privacy terms).

[0132] Technical support: Drift detection algorithms (such as KS test, PSI index); Automated audit logs (record all model operations and decision paths).

[0133] In another embodiment, the interactive analysis unit further includes an alarm component and a correlation analysis component; Alarm component: Analyzes abnormalities in sleep quality data, generates alarms and provides recommendations for combined CPAP treatment, and pushes them to doctors via their mobile devices (e.g., alerting to suicide risk when an abnormal decrease in HRV is detected). Correlation analysis component; generates simplified polysomnography (PSG) reports based on sleep quality data, and allows interaction with doctors via a doctor's terminal, supporting correlation analysis with PHQ-9 depression scores.

[0134] In another embodiment, the interactive analysis unit further includes a heatmap component; the user interaction terminal further includes a patient mobile terminal. Heatmap component: Generates a sleep structure heatmap based on sleep quality data and pushes it to the patient via their mobile device.

[0135] In another embodiment, the interactive analysis unit further includes a treatment progress component and an emotion fluctuation component; The treatment progress component: provides patients with a visual representation of their treatment progress via their mobile devices; The mood fluctuation component provides patients with mood fluctuation trend information via their mobile devices.

[0136] In this embodiment, the doctor's mobile terminal is used by the doctor to remotely monitor and adjust the treatment; the patient's mobile terminal is used by the patient to view the treatment progress.

[0137] In this embodiment, both the doctor's mobile terminal and the patient's mobile terminal have user-friendly mobile application interfaces, which facilitates interaction and data monitoring between medical staff and patients to obtain real-time and visualized treatment feedback.

[0138] The application interface is simple and clear, easy to use, and convenient for medical staff and patients to operate and obtain timely feedback.

[0139] In this embodiment, the multimodal closed-loop feedback press needle electrical stimulation antidepressant system enables users to interact with the big data model through a patient client (APP). Users can view treatment progress, heart rate data, electrical stimulation parameters, and sleep quality parameters in real time through the APP, and adjust some parameters as needed.

[0140] In this embodiment, the multimodal closed-loop feedback press needle electrical stimulation antidepressant system enables the medical team to remotely monitor patient data and adjust treatment plans through a doctor's client, ensuring personalized and efficient treatment.

[0141] In another embodiment, the AI ​​model includes: an acupoint location sub-model, a physical stimulation optimization sub-model, a drug therapy regulation sub-model, a dynamic balance integration sub-model, and a personalized treatment sub-model. Acupoint location sub-model: Based on multi-source physiological data, sleep quality data, and treatment effect feedback information, it dynamically generates meridian activation state vectors; Physical stimulation optimization sub-model: Dynamically generate electrical stimulation parameter vectors based on multi-source physiological data, sleep quality data, and treatment effect feedback information; Drug therapy regulation sub-model: Dynamically generate drug release vectors based on multi-source physiological data, sleep quality data, and treatment effect feedback information; Dynamic equilibrium integration sub-model: Based on the acupoint selection vector, electrical stimulation parameter vector and drug release vector, the synergistic effect is quantified to generate an integrated treatment plan; Personalized treatment sub-model: Based on patient profiles, predict the differences in patients' responses to integrated treatment plans, optimize integrated treatment plans through counterfactual reasoning, generate synergistic antidepressant plans of acupoint electrical stimulation and skin drug delivery, and obtain predicted treatment parameters.

[0142] In another embodiment, the acupoint location sub-model is implemented using a combination of 3D CNN and graph neural network (GNN). In this embodiment, the acupoint location sub-model dynamically recommends the optimal combination of acupuncture points (such as the synergistic effect of Baihui and Taichong points) by analyzing a large amount of historical efficacy data.

[0143] In this embodiment, the acupoint location sub-model includes a meridian feature processor mathematical model: Input: Patient's real-time biosignal matrix B∈R T×C (T is the time step, and C is the number of channels); Output: Meridian activation state vector fm∈R K (K meridian number); Calculation process: (1) Acupoint topological mapping:

[0144] Where N is the number of acupoints and Wp is the pre-trained weight matrix, which maps physiological signals to the acupoint space.

[0145] (2) Graph convolution fusion:

[0146] Where A∈{0,1} N×N : Acupoint adjacency matrix (based on preloaded topology); D=diag(∑ j A ij ) : Degree matrix; Θ (l) : Trainable parameters; σ is the activation function; (3) Meridian feature extraction: f m =MAXPOOL(H (L) The final layer output is max-pooled to obtain the meridian activation state vector. Based on the meridian activation state vector, the optimal acupuncture points can be obtained.

[0147] In another embodiment, the physical stimulus optimization sub-model includes a physical stimulus optimizer mathematical model: Input: Skin impedance z Pain threshold s Meridian activation state vector f m ; Output: Electrical stimulation parameter vector p = (waveform, frequency, intensity); Calculation process: (1) Neural activation model

[0148] Where I: current intensity; f: frequency; f0: target meridian resonance frequency (derived from f m Mapping).

[0149] (2) Discomfort constraint:

[0150] in, α , β , γ As a calibration parameter, γs defines the individualized tolerance threshold.

[0151] (3) Multi-objective optimization maxA(p)stD(p)≤ γs ,p∈P safe Among them, P safe Preset safety parameter space (loaded from SAFETY_PROFILE).

[0152] It should be noted that electrical stimulation modulation requires a millisecond-level response, but complex model calculations are time-consuming. Therefore, this implementation performs real-time edge computing optimization on the physical stimulation optimization sub-model: Lightweight Model: Knowledge Distillation of DRL Policy Networks (Teacher Model: Large-scale network in the cloud; Student Model: Micro-network on the device). Hardware co-design: Low-power real-time inference of spiking neural networks (SNN) is achieved using a neuromorphic chip (Loihi 2).

[0153] In another embodiment, the drug therapy regulation sub-model is implemented using deep reinforcement learning (DRL) combined with digital twins.

[0154] It should be noted that the physical control of drug release is as follows: A hydrogel microneedle is inserted into the skin; the tip of the hydrogel microneedle is equipped with a drug-loaded microsphere; the drug-loaded microsphere is a pH-responsive microsphere; an electrode is provided on the hydrogel microneedle; by adjusting the current output of the hydrogel microelectrode (0.1-5mA accuracy), the pH-responsive microsphere is triggered to release the drug (error <5%).

[0155] In this embodiment, the drug therapy regulation sub-model is based on the patient's real-time physiological feedback (such as changes in EEG γ wave power) and adjusts the electrical stimulation parameters (frequency and intensity) of the microneedle electrodes through Q-learning; a virtual digital twin of the patient is constructed to simulate the effect of different drug release curves on the therapeutic effect.

[0156] In this embodiment, the drug therapy regulation sub-model includes a pharmacodynamic module mathematical model: Input: Meridian activation state vector f m Circadian rhythm signal c∈R D ; Output: Drug release vector d=[d1,d2,…,d M ]; Calculation process: (1) Pharmacokinetic modeling:

[0157] in, Ct Blood drug concentration; ka , ke Absorption / elimination rate (loaded from PHARMA_DB) (2) Personalized parameter prediction

[0158] The MLP network outputs patient-specific pharmacokinetic parameters.

[0159] (3) Optimize the release plan

[0160] Wherein, constraint: d t ∈[0,d max ], where λ is the coefficient of the sparse regularization term.

[0161] In another embodiment, the dynamic equilibrium integration sub-model: Input: Meridian activation state vector f m Drug release vector d, electrical stimulation parameter vector p; Output: Integrated treatment plan T; (1) Quantification of synergistic effects:

[0162] in, μ i These are weighting coefficients, calibrated using historical data.

[0163] (2) Security verification:

[0164] Among them, g i Preset safety constraint functions (such as maximum current, drug accumulation, etc.).

[0165] In another embodiment, the personalized treatment sub-model is implemented through multi-task learning (MTL) combined with causal reasoning.

[0166] Based on patient profiles, the study predicts differences in patient responses to integrated treatment plans, optimizes integrated treatment plans through counterfactual reasoning, generates synergistic antidepressant plans combining acupoint electrical stimulation and dermal drug delivery, and obtains predicted treatment parameters.

[0167] For example, integrating genomic data (such as the 5-HTTLPR genotype) with clinical scales (PHQ-9) can predict patient response differences to active ingredients of Bupleurum and Paeonia lactiflora versus electrical stimulation; and can optimize treatment plans through counterfactual reasoning (such as "if the electrical stimulation frequency is increased from 10Hz to 20Hz, the depression score is expected to decrease by 15%").

[0168] Implementation Method 2: The acupoint stimulation module includes an electrical stimulation device and a metal main needle; The electrical stimulation device is connected to the AI ​​control module and is used to transmit electrical stimulation pulse current to the metal main needle according to the electrical stimulation parameters of the acupoint predicted by the AI ​​control module. The metal main needle is used to insert into the predicted acupoints on the patient's body for electrical stimulation, thereby stimulating the acupoints.

[0169] In this embodiment, the electrical stimulation pulse current is a low-frequency pulse current with a frequency of 20±2 Hz and an intensity of ≤5 mA.

[0170] In this embodiment, the electrical stimulation pulse current is conducted to the corresponding acupoint through the metal main needle to activate the nerve pathway.

[0171] In another embodiment, the electrical stimulation device has a built-in PWM pulse signal generator that outputs a low-frequency pulse current (electrical stimulation pulse current).

[0172] In another embodiment, the electrical stimulation device further includes a housing, a current-limiting resistor, a lithium battery pack, and an LED interactive screen; the PWM pulse signal generator, the current-limiting resistor, and the lithium battery pack are integrated inside the housing; and the LED interactive screen is integrated on the surface of the housing.

[0173] In this embodiment, to ensure the safe use of the device, the current-limiting resistor (resistance range 1-5kΩ) is set to ensure that the current intensity meets human safety standards (IEC 60601-1).

[0174] In this embodiment, the electrical stimulation device is a portable device with an overall size of 2 cm in diameter and 1 cm in thickness.

[0175] In this implementation method, clinical validation was performed: A multicenter trial (n=120) compared the results of portable devices (electric stimulation devices) with those of standard PSG monitoring, and the Kappa concordance coefficient reached 0.86.

[0176] In this embodiment, the multimodal closed-loop feedback press needle electrostimulation antidepressant system can generate a 20Hz low-frequency pulse current through a PWM (pulse width modulation) signal generator. The precisely controlled current intensity (usually controlled within 5mA) is conducted to the acupoint through the metal main needle, thereby activating the relevant neural pathways and achieving the effect of regulating the central nervous system.

[0177] Implementation method 3: The synergistic treatment module includes an acupoint stimulation module and a skin drug delivery module; Acupoint stimulation module: Based on the predicted acupoints and electrical stimulation parameters, perform acupoint electrical stimulation on the patient; Skin delivery module: Administers drugs to patients through the skin based on predicted drug release rates.

[0178] Implementation method 4: The skin drug delivery module includes a microneedle triggering device and hydrogel microneedles; The hydrogel microneedles are equipped with electrodes that are electrically connected to a microneedle triggering device. The microneedle triggering device is signal-connected to an AI control module and is used to conduct a drug release triggering current to the hydrogel microneedles based on the drug release rate predicted by the AI ​​control module for skin administration. The drug release triggering current is used to regulate the pH value of the skin surface through electrolysis. The tips of the hydrogel microneedles are uniformly distributed with drug-loaded microspheres; the drug-loaded microspheres are pH-responsive microspheres, and their drug release rate is regulated by changes in the pH value of the skin surface.

[0179] In this embodiment, the pH response mechanism of the drug-loaded microspheres is achieved by regulating the local tissue pH value through electrical stimulation (trigger current): Electrical stimulation can induce electrochemical reactions in the skin, generating H⁺ or OH⁻ ions, thereby regulating tissue pH. Using a given pH adjustment curve (the relationship between drug release rate and pH value), the AI ​​control module dynamically optimizes the current parameters (frequency 20Hz, intensity ≤5mA) to achieve high-precision pH control. By applying electrical stimulation through hydrogel microneedles, the pH environment of the local tissue at the acupoint is altered, thereby accelerating or slowing down the drug release rate at the acupoint.

[0180] In this embodiment, the hydrogel microneedles are used to be inserted near the predicted acupoints for electrical stimulation on the patient's body.

[0181] Implementation method 5: The acupoint stimulation module includes an electrical stimulation device and a group of medical stainless steel main needles; The electrical stimulation device is connected to the AI ​​control module and is used to transmit electrical stimulation pulse current to the metal main needle according to the electrical stimulation parameters of the acupoint predicted by the AI ​​control module. The medical stainless steel main needle group is used to insert into the predicted acupuncture points on the patient's body for electrical stimulation, thereby providing acupuncture point electrical stimulation. Implementation method 6: The synergistic treatment module includes an electrical stimulation device, a microneedle triggering device, and a drug-loaded microneedle patch; The drug-loaded microneedle patch includes: a group of medical stainless steel main needles, a group of hydrogel drug-loaded auxiliary needles, and adhesive tape. The adhesive tape is thicker in the middle and thinner around the edges; the medical stainless steel main needle group and the hydrogel drug-loaded auxiliary needle group are fixed in the middle area of ​​the adhesive tape. The medical stainless steel main needle group includes 4 medical stainless steel main needles; the 4 medical stainless steel main needles are symmetrically distributed around the center of the adhesive tape; The hydrogel drug-loaded needle array is made of hydrogel material; the hydrogel drug-loaded needle array includes multiple drug-loaded needles, which are drug-loaded microneedles; the tips of the drug-loaded microneedles are uniformly distributed with drug-loaded microspheres; the drug-loaded microspheres are pH-responsive microspheres, and their drug release rate is regulated by changes in the pH value of the skin surface. The hydrogel drug-loaded auxiliary needle group is divided into 4 regions, each region is arranged in an octagonal array; the 4 regions are distributed around the periphery of the medical stainless steel main needle group, forming the shape of 4 petals of a lilac flower; the needle height of the medical stainless steel main needle group is greater than the needle height of the hydrogel drug-loaded auxiliary needle group. The medical stainless steel main needle group is electrically connected to the electrical stimulation device and receives electrical stimulation pulse current; The hydrogel drug-loaded accessory needle group is equipped with electrodes for electrical connection with the microneedle triggering device; The electrical stimulation device is signal-connected to the AI ​​control module and is used to transmit electrical stimulation pulse current to the medical stainless steel main needle group according to the electrical stimulation parameters of the acupoints predicted by the AI ​​control module. The medical stainless steel main needle group is used to insert into the predicted acupoints on the patient's body for acupoint electrical stimulation, thereby performing acupoint electrical stimulation on the patient. The microneedle triggering device is signal-connected to the AI ​​control module and is used to conduct a drug release triggering current to the hydrogel microneedles according to the drug release rate predicted by the AI ​​control module for skin administration. The drug release triggering current is used to regulate the pH value of the skin surface through electrolysis, thereby regulating the drug release rate of the drug-loaded microspheres.

[0182] In this embodiment, the needle body of the medical stainless steel main needle is made of austenitic stainless steel wire, specifically: The needle body of the medical stainless steel main needle is made of 06Cr19Ni10 (SUS304) austenitic stainless steel wire as specified in GB / T4240-2019 "Stainless Steel Wire".

[0183] In this embodiment, the medical stainless steel main needle is conductive to provide electrical stimulation to acupoints.

[0184] In this embodiment, the four regions of the hydrogel drug-loaded accessory needle group are closely distributed around the periphery of the medical stainless steel main needle group.

[0185] In this embodiment, the height of the medical stainless steel main needle group is greater than the height of the hydrogel drug-loaded auxiliary needle group, that is, the medical stainless steel main needle is slightly longer, which can effectively stimulate the target acupoint.

[0186] In this embodiment, the hydrogel-loaded drug-eluting needle group occupies approximately 2 / 3 of the area on the adhesive tape, resulting in a wide drug delivery range.

[0187] In this embodiment, the patch uses a main needle group plus a secondary needle group in the center. The secondary needles in each area are located behind the main needles and are distributed in an octagon. The peripheral secondary needles can work together with the main needles to provide auxiliary stimulation to specific acupoints, thereby enhancing the therapeutic effect. The secondary needle area is adjacent to the main needle area, and there is a certain height difference between the main needle group and the secondary needle group. The metal needle group (i.e., the main needle) is slightly longer to effectively stimulate the target acupoints.

[0188] In this embodiment, because the metal main needle group and the hydrogel secondary needle tip have different heights, the height difference can ensure that the depth of the metal main needle group piercing the skin is not affected by the dense secondary needle group. At the same time, even if the patch is slightly deviated from its position, the four main needles can still effectively stimulate specific acupoints, thereby improving the fault tolerance rate of the patch and reducing the difficulty of use.

[0189] In this embodiment, the secondary needles in the four areas form a lilac shape around the central main needle area, enhancing the visual appeal.

[0190] Implementation method 7: The diameter of the medical stainless steel main needle is 0.16mm-0.30mm, the needle body length is 0.8mm-3mm, and the needle body is made of austenitic stainless steel wire; The drug-loaded accessory needle is conical in shape, with a needle length of 0.8 mm and a bottom diameter of 0.3 mm.

[0191] In this embodiment, the dense hydrogel needle cluster is 0.8 mm long and can penetrate the epidermal layer for sustained drug release.

[0192] Implementation method 8: In each region of the hydrogel drug-loaded needle group, the needle tip spacing between the drug-loaded needles is 0.65 mm, the number of drug-loaded needles is 10 × 10 (i.e., 100), and the distribution area of ​​the drug-loaded needles is 6.5 mm × 6.5 mm.

[0193] Implementation method 9: The adhesive tape includes a waterproof layer, a limiting and fixing layer, and an adhesive; The waterproof layer is a PU film, located on the outermost layer; The limiting layer is made of non-woven fabric and is located below the waterproof layer; The portion of the adhesive tape that comes into contact with the skin is an adhesive; the adhesive is polysilicon.

[0194] In this embodiment, the waterproof layer is made of medical PU film, which is extremely thin and has the characteristics of high waterproofness, low allergy, and breathability.

[0195] In this embodiment, the material of the fixing layer is non-woven fabric, which is moisture-proof, breathable, flexible, lightweight, and inexpensive.

[0196] In this embodiment, the adhesive material is: The adhesive that comes into contact with the skin is polysiloxane, also known as silicone gel. It is a polymer of organic compounds containing siloxanes. In this polymer, it is directly bonded to Si atoms through Si-O bonds and Si-C bonds. It usually exists in a gel-like form. Polyurethane dressings containing silicone gel can quickly adhere to the skin, have a soft texture, and their multi-layered structure can relieve pressure injuries. They are also easy to separate from the skin when removed.

[0197] In this embodiment, the size of the adhesive tape is 8.9mm × 8.9mm.

[0198] In this embodiment, the adhesive used to fix the needle body and the limiting fixation layer is a hydrogel.

[0199] In this embodiment, the limiting and fixing layer and the waterproof layer are fixedly connected by hydrocolloid.

[0200] In this embodiment, the medical stainless steel main needle is fixedly connected to the adhesive tape via a hydrocolloid: The needle tip of the medical stainless steel main needle passes through the limiting and fixing layer; The bottom of the medical stainless steel main needle is wrapped between the limiting and fixing layer and the waterproof layer; The bottom surface of the medical stainless steel main needle is fixedly connected to the upper surface of the waterproof layer using a hydrocolloid.

[0201] A limiting and fixing layer is used to limit and fix the bottom of the medical stainless steel main needle, which strengthens the fixation of the main needle and enhances its stability.

[0202] In this embodiment, the hydrogel drug-loaded accessory needle group is fixedly connected to the upper surface of the limiting and fixing layer via hydrocolloid.

[0203] It should be noted that the structure and composition of a traditional snap button are as follows: (1) Structure: It is made of medical stainless steel wire.

[0204] (2) Composition: The needle body of the press needle is made of 06Cr19Ni10 (sus304) austenitic stainless steel wire as specified in GB / T4240-2019 "Stainless Steel Wire".

[0205] (3) Number of needles: 1.

[0206] (4) Specifications: needle diameter 0.16mm-0.30mm, needle length 0.6mm-30mm.

[0207] (5) Adhesive tape: Skin-colored hypoallergenic adhesive tape.

[0208] In this embodiment, compared with traditional press needles, the hydrogel press needle (drug-loaded microneedle patch) has the following advantages: (1) Increase the number of main needles and the number of auxiliary needles: The patch adopts a main needle group plus an auxiliary needle group in the center. When using this acupoint patch, the main needle is aligned with the acupoint to achieve targeted stimulation of the specific acupoint. By setting auxiliary needles around the main needle, the auxiliary needles can work together with the main needle to achieve auxiliary stimulation of the specific acupoint, thereby enhancing the therapeutic effect. In addition, considering that users of this patch (medicated microneedle patch) may not have very professional acupoint identification ability, this technical solution, through the layout of the main needle and the auxiliary needles, can still achieve effective stimulation of the specific acupoint even if the patch is slightly deviated from its position, thereby improving the error tolerance of the patch and reducing the difficulty of use.

[0209] (2) The auxiliary needle is a drug-loaded sustained-release needle: The auxiliary needle is a minimally invasive device that can overcome the stratum corneum barrier to achieve transdermal drug delivery. It uses a single-emulsion method to create sustained-release microspheres that encapsulate the drug, allowing it to enter the bloodstream through the skin to produce local or systemic therapeutic effects. This effectively solves the problem of large molecule drugs being unable to penetrate the stratum corneum barrier and can improve the transdermal penetration rate of the drug. This product avoids drug degradation by enzymes in the stomach and metabolism in the liver, thus reducing stomach irritation, and also avoids the pain caused by injection.

[0210] (3) Improved adhesive material: The back of the patch is made of PU film combined with silicone gel, which is more waterproof than traditional medical tape and has low allergenicity, making it suitable for long-term use.

[0211] (4) Improved backing structure: The backing structure is optimized to be thick in the middle and thin around the edges to ensure that the backing gel adheres tightly to the skin. This not only allows water to flow quickly through the patch to effectively prevent water penetration, but also allows the patch to adapt to different user environments. Patients can move freely without being restricted by the treatment method, thus improving their sense of well-being during the treatment.

[0212] It should be noted that the drug-loaded microspheres are pH-responsive microspheres, and their drug release rate can be controlled by changes in environmental pH.

[0213] In this embodiment, the drug-loaded auxiliary needle is equipped with microelectrodes, which form a weak electric field on the surface of human skin through electrolysis, thereby changing the pH value and ion concentration of the skin surface and stimulating skin cells to produce electrochemical reactions; by changing the pH value, the drug release rate of the auxiliary needle is controlled.

[0214] Implementation method 10: Drug-loaded microspheres, wherein the drug-loaded microspheres are W / O / W microspheres, comprising a liposome core, a PLGA intermediate layer and a PVA shell; The liposome core encapsulates a lipophilic drug component; The PLGA intermediate layer encapsulates the liposome core; The PVA shell encapsulates the PLGA intermediate layer to form a W / O / W structure.

[0215] In this embodiment, the liposome core comprises Bupleurum-White Peony Extract and 5% Trehalose Stabilizer; The PLGA intermediate layer is formed by dissolving amino-terminated PLGA in a mixed solvent of benzyl benzoate and benzyl alcohol. The PVA shell is formed by curing a polyvinyl alcohol-sucrose matrix solution.

[0216] In this embodiment, W / O / W: water-in-oil-in-water.

[0217] W / O / W microspheres: water-in-oil-in-water (drug) microspheres, water-in-oil-in-water double emulsion microspheres.

[0218] In this embodiment, the liposome core is an inner aqueous phase; the PLGA intermediate layer is an oil phase; and the PVA shell is an outer aqueous phase.

[0219] In this embodiment, polyvinyl alcohol (PVA) is used.

[0220] In this embodiment, the liposome core comprises Bupleurum-White Peony Extract and 5% Trehalose Stabilizer: Adding 5% trehalose stabilizer protects the drug's activity during emulsification, solidification, and freeze-drying processes, and reduces burst release.

[0221] In this embodiment, the drug-loaded microspheres have a particle size of 10-20 μm, an encapsulation efficiency of ≥77.8%, and a drug loading of ≥4.28%.

[0222] In this embodiment, the structure of liposome core + PLGA intermediate layer realizes dual controlled release (dual drug delivery controlled release system): (1) Core components:

[0223] (2) Dual controlled release mechanism: Liposome core release: Initial rapid release is achieved after the phospholipid bilayer breaks down, i.e., the rapid release phase, during which the lipophilic components diffuse into the cell; PLGA intermediate layer degradation: slow hydrolysis of the polymer achieves long-lasting sustained release, i.e., the sustained release stage, which controls the release rate of liposomes; Liposome core + PLGA intermediate layer: synergistically maintains blood drug concentration.

[0224] (3) Special design of the internal water phase: Key structures: The liposome core encapsulates a lipophilic drug component; the PLGA intermediate layer encapsulates the liposome core; the liposome core is an internal aqueous phase. This structure breaks through the traditional drug delivery structure: Traditional drug delivery structures: lipophilic components are directly dispersed in the oil phase of PLGA, resulting in low encapsulation efficiency and high burst release rate; The drug delivery structure of this technology is a triple structure of "liposome (core) - PLGA (intermediate layer) - PVA (shell)".

[0225] (4) Technological advantages: Dual controlled release synergistic mechanism: Improve encapsulation rate: The liposome barrier prevents lipophilic components from leaking into the oil phase during emulsification.

[0226] Reduce burst release effect: Liposomes and PLGA microspheres form a dual carrier: The PLGA layer prevents the drug from directly contacting the external environment; the liposomes must first overcome the PLGA barrier to release the drug; together they reduce the burst release rate.

[0227] Co-regulation release kinetics: Immediate-release phase: After microneedle puncture, the liposomes rapidly release the drug, resulting in a rapid onset of action (relieving acute symptoms). Sustained-release phase: PLGA controls the sustained release of liposomes, maintaining long-term blood drug concentration (reducing relapse rate).

[0228] In this embodiment, the PVA shell is solidified using a polyvinyl alcohol (PVA)-sucrose matrix solution to form water-soluble microneedle tips: (1) Dissolution mechanism: PVA contains a large number of hydroxyl groups (-OH), while sucrose contains multiple hydroxyl groups and forms hydrogen bonds when it comes into contact with water. Upon contact with skin tissue fluid, the hydrogen bond network breaks down, leading to rapid dissolution.

[0229] (2) The functional positioning of "cutting-edge": Core region: 500–800 μm segment at the tip of the microneedle; Drug-loaded focusing: Drug-loaded microspheres are enriched at the tip (accounting for ≥80%), and the substrate is mainly blank PVA-sucrose matrix.

[0230] (3) Structural realization of water-soluble tips: Polyvinyl alcohol (PVA): Function: Film-forming framework, providing mechanical strength; Contribution to solubility: Water-soluble polymers; sucrose: Functions: Pore-forming agent and plasticizer, improving the dissolution rate after puncture; Solubility contribution: small molecule sugars.

[0231] Tip dissolution kinetics: Puncture of the stratum corneum → infiltration of tissue fluid → hydrogen bond breakage → matrix dissolution → microsphere release.

[0232] Time threshold: swelling begins within 5 seconds → complete dissolution within 60 seconds → release of drug-loaded microspheres into the dermis.

[0233] Synergistic effect: Sucrose accelerates water penetration, while PVA controls the dissolution rate to prevent premature burst release of microspheres.

[0234] (4) Technical advantages compared to traditional water-insoluble microneedles:

[0235] The core problem to be solved: Precise acupoint delivery: After dissolution, the microspheres are positioned in the dermal layer of the Neiguan acupoint (depth ≈ 1 mm) to activate nerve endings; Avoid drug waste: Tip dissolution and local release reduce systemic circulation losses and increase drug concentration at the target site.

[0236] Implementation method 11: Microfluidic chip, wherein the microfluidic chip is used to prepare the above-mentioned drug-loaded microspheres; The microfluidic chip includes a main channel, a first channel, and a second channel; Main pipeline: Its diameter gradually increases from upstream to downstream, and it is used to inject the internal water phase; First pipe: intersects with the main pipe at an angle, used to inject the oil phase to cut the inner aqueous phase and form the W / O emulsion core; The second pipe is located downstream of the first pipe and intersects with the main pipe at a cross shape. Its diameter differs from that of the first pipe by ≤5%. It is used to inject the external aqueous phase to encapsulate the W / O emulsion core and form W / O / W microspheres.

[0237] In this embodiment, the microfluidic chip is designed using PR drawing software and fabricated using integrated 3D printing technology.

[0238] In this embodiment, the microfluidic chip further includes a substrate: It measures 4cm x 5cm, is 12mm thick, and is made of PDMS. The substrate is used to encapsulate microchannels (i.e., a network of microfluidic channels fabricated on the substrate, including a main channel, a first channel, and a second channel).

[0239] The substrate serves as a three-dimensional carrier, and the microchannel structure is encapsulated inside it using integrated 3D printing technology to form a complete flow path system.

[0240] The core functions of a substrate: (1) Physical support and packaging: Carrying channels: The designed cross (intersecting) channel structure is materialized inside the PDMS substrate through photolithography or 3D printing process; Leak-proof sealing: The elastic PDMS material automatically seals the flow channels after bonding, preventing fluid leakage.

[0241] (2) Fluid control assistance: Wettability regulation: The hydrophobicity of PDMS surface promotes oil / water phase separation and supports laminar flow at low Reynolds number; Optical transparency: Facilitates observation of the droplet formation process (such as the oil phase cutting through the internal aqueous phase).

[0242] (3) System Integration Fundamentals: The substrate serves as a mechanical interface, supporting the docking of the microfluidic chip with external devices (such as piezoelectric droplet generators and pressure pumps).

[0243] In this embodiment, the diameters of the first pipe and the second pipe are equal or similar (difference ≤ 5%).

[0244] In this embodiment, under laminar flow conditions with low Reynolds number (e.g., Re < 100), the interface between the inner aqueous phase and the oil phase remains stable within the cross-shaped channel; the shear force generated by the vertical injection of the oil phase cuts the inner aqueous phase into monodisperse water-in-oil (W / O) emulsion cores.

[0245] In this embodiment, the microfluidic chip is used to integrate with a piezoelectric droplet generator to cut the internal aqueous phase into monodisperse droplets through piezoelectric vibration; at the same time, the diameter of the main channel gradually increases from upstream to downstream to suppress droplet aggregation, and finally obtains uniform microspheres with a small coefficient of variation (CV) (e.g., <5%).

[0246] Implementation Method 12: Drug-loaded microsphere preparation system, the system being used to prepare the above-mentioned drug-loaded microspheres; The system includes: a microfluidic pressure pump, a piezoelectric droplet generator, and a microfluidic chip; the microfluidic chip is the microfluidic chip described above. The piezoelectric droplet generator is integrated into the main channel inlet of the microfluidic chip and is used to cut the internal aqueous phase into monodisperse droplets; The microfluidic pressure pump is used to inject internal aqueous phase, oil phase, and external aqueous phase into the microfluidic chip at a given flow ratio: Main pipeline: Injects internal water phase; First conduit: Injecting oil phase to cut the internal aqueous phase droplets; Second channel: Inject external aqueous phase to encapsulate oil phase, forming W / O / W microspheres.

[0247] In this embodiment, the piezoelectric microdroplet generator has a built-in piezoelectric ceramic and is connected to the excitation power supply signal; the excitation power supply sends a pulse voltage, and the piezoelectric ceramic generates high-frequency vibration (>1 kHz) under the pulse voltage, cutting the continuous internal water phase flow into uniform droplets.

[0248] In this embodiment, the drug-loaded microsphere preparation system: Piezoelectric microdroplet generator: Enables on-demand generation of monodisperse droplets, significantly improving particle size uniformity and controllability.

[0249] Microfluidic chip cross (intersecting) channel: orderly three-phase assembly to ensure the integrity of W / O / W structure.

[0250] Precise flow control of microfluidic pressure pumps: Optimized oil / water ratio for high encapsulation efficiency.

[0251] Particle size controllability: The active cutting of the piezoelectric module (piezoelectric microdroplet generator) significantly reduces the standard deviation of microsphere size compared to traditional methods.

[0252] Drug release stability: Uniform particle size and double encapsulation structure significantly reduce burst release effect.

[0253] Implementation Method 13: A method for preparing drug-loaded microspheres, the method comprising the following steps: Drug extraction steps: The effective components of Bupleurum chinense and Paeonia lactiflora were extracted using the ethanol reflux method to obtain the extract. Preparation steps for PLGA solution: Under constant temperature of 40℃, 180 mg of amino-terminated PLGA was added to a mixed solvent of benzyl benzoate and benzyl alcohol, and the mixture was stirred magnetically until completely dissolved. The resulting PLGA solution was used as the oil phase. Preparation steps of drug solution: The extract was prepared into liposomes, and the lipophilic components were encapsulated in the liposomes and then used as the inner aqueous phase to improve the encapsulation efficiency and achieve dual controlled release. 5% trehalose stabilizer was added to the inner aqueous phase to protect the activity of the drug during emulsification, solidification and freeze-drying processes and reduce burst release. Steps for forming a water-in-oil-in-water double emulsion: Polyvinyl alcohol solution was used as the external aqueous phase; Using the drug-loaded microsphere preparation system described above, W / O / W microspheres are prepared based on an internal aqueous phase, an oil phase, and an external aqueous phase. Microsphere solidification and purification steps: Photocuring of W / O / W microspheres in the presence of a photoinitiator enhances the mechanical strength and stability of the microspheres and reduces burst release. The photocured microspheres are then irradiated with 365 nm blue light for 120 seconds to achieve cross-linking and curing. The cross-linked and cured microspheres were washed and purified to remove residual solvents and surfactants, and purified microspheres were obtained. Microsphere post-processing steps: The purified microspheres were mixed with deionized water at a mass ratio of 2:1 and incubated in a constant temperature shaker for 1 hour. The incubated microspheres were treated with 50 kHz ultrasound for 5 minutes to promote uniform dispersion of the microspheres and obtain an ultrasound-treated suspension. Mannitol was added to the sonicated suspension as a freeze-drying protectant to improve resolubility and reduce microsphere aggregation and structural collapse during freeze-drying. The suspension containing the lyophilization protectant was transferred to a lyophilizer and freeze-dried at -60°C for 72 hours to obtain PLGA sustained-release microsphere powder as drug-loaded microspheres.

[0254] In this embodiment, the ethanol reflux method is also known as the alcohol extraction method, and the effective components of Bupleurum-White Peony are the active components in the Bupleurum-White Peony raw materials.

[0255] In this embodiment, the steps of the ethanol reflux method are as follows: Raw material processing: Bupleurum and white peony coarse powder (1:1, total mass 9g) were heated and refluxed three times (2h / time) with ethanol (material-liquid ratio 1:10). Concentration process: Combine the extracts and concentrate them to a semi-fluid state (approximately 5 mL volume) using a rotary evaporator (55℃ water bath, 60 Pa, 30 r / min) to obtain the extract.

[0256] In this embodiment, during the preparation step of the PLGA solution: 180 mg amino-terminated PLGA:PLA-PGA, monomer ratio 25:75, molecular weight range 5k-30k; Benzyl benzoate and benzyl alcohol mixed solvent: 1 mL, ratio 1:1; Magnetic stirring: 300 rpm; Oil phase (PLGA solution): PLGA concentration is 10%-20% (w / v).

[0257] In this embodiment, a drug-loaded microsphere preparation system is used to prepare W / O / W microspheres based on an internal aqueous phase, an oil phase, and an external aqueous phase, including the following steps: Three-phase fluid drive and flow control steps: The microfluidic pressure pump delivers the (drug) liquid phase to the microfluidic chip according to the following flow parameters: Main pipeline: Injection of an internal aqueous phase solution containing drug liposomes, with a flow rate range of 5–10 μL / min; First pipeline: Inject oil phase solution (a mixed solvent of benzyl benzoate / benzyl alcohol containing PLGA, or simply PLGA solution), with a flow rate range of 3–5 μL / min; Second pipeline: Inject external aqueous phase solution (polyvinyl alcohol, PVA) at a flow rate of 3–5 μL / min; Steps for piezoelectric on-demand droplet generation: When the internal aqueous phase flows through the piezoelectric microdroplet generator integrated into the main pipeline inlet, monodisperse droplet control is achieved through the following mechanism: Active cutting: Piezoelectric ceramics generate high-frequency vibrations (>1 kHz) under pulsed voltage, cutting the continuous internal water phase flow into uniform droplets (diameter deviation <±0.5μm). On-demand start / stop: Instantly starts and stops droplet generation in response to electronic control signals to avoid wasting ineffective fluid; Microfluidic channel emulsion assembly steps: After a monodisperse aqueous droplet enters the cross channel (i.e., the channel formed by the intersection of the first and second pipes with the main pipe), it sequentially undergoes the following: Oil phase cutting: The oil phase in the first channel encapsulates the water phase droplets inside, forming a water-in-oil (W / O) primary emulsion; External aqueous phase encapsulation: The external aqueous phase (PVA) of the second conduit (c) further encapsulates the W / O droplets, forming water-in-oil-in-water (W / O / W) double emulsion microspheres.

[0258] The generated W / O / W emulsion is exported through the downstream channel (of the main pipeline) and its structure is fixed by photocuring (365nm blue light irradiation for 120 seconds), finally obtaining drug-loaded (sustained-release) microspheres with a particle size of 10–20 μm and an encapsulation efficiency of "≥77.8%".

[0259] In this embodiment, the cross-linked and cured microspheres are washed and purified to remove residual solvents and surfactants: The cross-linked and cured microspheres were transferred to centrifuge tubes and washed three times each with acetone and 75% ethanol (v / v) (10 mL each time, centrifugation conditions: 3000 rpm, 10 min) to remove residual solvent and surfactant.

[0260] In this embodiment, the purified microspheres were mixed with deionized water at a mass ratio of 2:1 and incubated in a constant temperature shaker for 1 hour at an incubation temperature of 25±2℃. Purpose: To fully hydrate the microspheres and restore them to their hydrated state before freeze-drying; Eliminate the damage to the microsphere structure caused by residual organic solvents (acetone / ethanol) during the purification process.

[0261] In this embodiment, the incubated microspheres are treated with 50 kHz ultrasound for 5 minutes: power density 0.5 W / cm², pulse mode: 2 seconds on and 1 second off.

[0262] Purpose: Break up microsphere aggregates to ensure monodispersity (reduce particle size distribution width); To avoid structural inhomogeneity caused by aggregation during subsequent freeze-drying.

[0263] In this embodiment, mannitol is added to the ultrasonically treated suspension as a freeze-drying protectant: 10% (w / v) mannitol, and the mixture is magnetically stirred (300 rpm) until completely dissolved.

[0264] Purpose: Improved solubility: Mannitol forms a glassy structure, reducing microsphere clumping after freeze-drying; Preventing structural collapse: A crystalline scaffold is formed during freezing to maintain the porous morphology of the microspheres.

[0265] In this embodiment, PLGA sustained-release microsphere powder: The particle size is 10-20 μm, the encapsulation efficiency is ≥77.8%, and the drug loading is ≥4.28%.

[0266] Implementation Method 14: Method for preparing drug-loaded microneedles, the method comprising the following steps: Vacuum casting loading steps: The drug-loaded microspheres described above are mixed with a polyvinyl alcohol-sucrose matrix solution at a mass ratio of 1:9 to form a drug-loaded suspension. The drug-loaded suspension was injected into the cavity of the microneedle mold, and a vacuum was applied for 20 minutes under a vacuum of -0.1 MPa to allow the drug-loaded microspheres to fully penetrate into the gap at the tip of the microneedle mold. Centrifugal filling and shaping steps: The vacuum-cast microneedle mold was centrifuged at 4200 rpm for 5 minutes. This process was repeated 3 times to remove excess matrix solution, so that the drug-loaded microspheres were evenly distributed at the tip of the microneedle mold to form drug-loaded microneedles.

[0267] In this embodiment, a polyvinyl alcohol (PVA)-sucrose matrix solution is used to form water-soluble microneedle tips: (1) Dissolution mechanism: PVA contains a large number of hydroxyl groups (-OH), while sucrose contains multiple hydroxyl groups and forms hydrogen bonds when it comes into contact with water. Upon contact with skin tissue fluid, the hydrogen bond network breaks down, leading to rapid dissolution.

[0268] (2) The functional positioning of "cutting-edge": Core region: 500–800 μm segment at the tip of the microneedle; Drug-loaded focusing: Through vacuum casting and centrifugal filling, drug-loaded microspheres are enriched at the tip (accounting for ≥80%), and the substrate is mainly blank PVA-sucrose matrix.

[0269] (3) Process implementation of water-soluble tips: Material selection: Polyvinyl alcohol: Function: Film-forming framework, providing mechanical strength; Contribution to solubility: Water-soluble polymers; sucrose: Functions: Pore-forming agent and plasticizer, improving the dissolution rate after puncture; Solubility contribution: small molecule sugars.

[0270] Tip dissolution kinetics: Puncture of the stratum corneum → infiltration of tissue fluid → hydrogen bond breakage → matrix dissolution → microsphere release.

[0271] Time threshold: swelling begins within 5 seconds → complete dissolution within 60 seconds → release of drug-loaded microspheres into the dermis.

[0272] Synergistic effect: Sucrose accelerates water penetration, while PVA controls the dissolution rate to prevent premature burst release of microspheres.

[0273] (4) Technical advantages compared to traditional water-insoluble microneedles:

[0274] The core problem to be solved: Precise acupoint delivery: After dissolution, the microspheres are positioned in the dermal layer of the Neiguan acupoint (depth ≈ 1 mm) to activate nerve endings; Avoid drug waste: Tip dissolution and local release reduce systemic circulation losses and increase drug concentration at the target site.

[0275] In this embodiment, the microneedle mold corresponds to the microneedle (mechanical) structure, the needle shape is conical, the needle length is 0.8mm, the bottom diameter of the needle body is 0.3mm, and the needle body is made of hydrogel material.

[0276] Implementation Method 15: A specific embodiment is provided for the determination of drug content in drug-loaded microspheres: (1) Liquid chromatography conditions: Column: Aglient C18 (4.6 × 250 mm, 5 µm); Mobile phase: 0-5 min: 10% acetonitrile; 5-20 min: 10%→35%; 20-25 min: 35%→50%; Column temperature: Low polar phase: 35℃ → High polar phase: 45℃; Flow rate: 1.0 mL⋅min⁻¹; Injection volume: 10.00; Injection volume: 20 μL; Detection wavelength: 190-400 nm; (2) Microsphere determination method: Dissolve an appropriate amount of alcohol-extracted drug in ultrapure water to prepare a 10% standard solution as a reference solution; dissolve drug microspheres (i.e., drug-loaded microspheres) in acetonitrile, perform high-pressure microfluidic jet for 15 min (to completely destroy the microsphere structure and eliminate drug encapsulation residue), add PBS to rinse the solution, centrifuge and collect the supernatant as the sample solution; PLGA blank microspheres were processed in the same way and used as blank solutions.

[0277] The determination was performed according to the chromatographic method, and the chromatographic peaks were recorded.

[0278] The formula for calculating drug component concentration is as follows: (Peak area in Ax sample; peak area in As standard; concentration of Cs standard) Data analysis system, constructing a triple validation model: ① Convolutional Neural Network - Peak Recognition: Automatically separates overlapping peaks with an AUC recognition rate of >99%.

[0279] ②PLS chemometrics eliminates matrix background interference, with a recovery rate of 98-102%.

[0280] ③ Blockchain data storage ensures that experimental data is uploaded to the blockchain in real time, guaranteeing that the original data cannot be tampered with.

[0281] By combining advanced separation technology, smart materials, artificial intelligence and blockchain, the first traditional Chinese medicine microsphere analysis platform that conforms to the FDAALCOA+ principle has been established, providing a new paradigm for quality research of complex formulations.

[0282] (3) Results: ① Liquid chromatography results: The extract, paeoniflorin, and paeoniflorin lactone were prepared into a 10% standard solution using ultrapure water. The peak areas were determined by liquid chromatography, and a standard curve was plotted with concentration and peak area as variables.

[0283] As shown in the high-performance liquid chromatogram, the retention time of paeoniflorin in the extract was 12 min. Therefore, under these chromatographic conditions, the determination of drug concentration was not affected by the excipients used in the preparation of the microspheres or the reagents used in the analysis, indicating that this chromatographic method has good specificity and can be used for the determination of drug content.

[0284] ② Results of microsphere encapsulation efficiency measurement: The concentration of the standard drug was 0.1 mg / ml. Accurately weigh 10 mg of drug-loaded microspheres into a 5 ml volumetric flask, add an appropriate amount of acetonitrile, dissolve the microspheres using a high-pressure microfluidic jet, dilute with PBS buffer, and bring to volume. After sonication to mix thoroughly, centrifuge 1 ml (3000 rpm, 15 min), and determine the drug content in the supernatant using HPLC.

[0285] The formulas for calculating drug loading and encapsulation efficiency are as follows: Drug loading (%) = Drug mass in microspheres / Total mass of microspheres × 100%.

[0286] Encapsulation efficiency (%) = Actual drug loading of microspheres / Theoretical drug loading of microspheres × 100%.

[0287]

[0288] Implementation method 16: A specific embodiment is provided for pharmacodynamic experimental research.

[0289] (1) Laboratory animals: Sixty SD rats, weighing approximately 180-220 g, were selected based on the ease of operation, high success rate, good repeatability, and low feeding cost. Half were male and half female. The rats had free access to water and food. They were kept at a temperature of 22-24 ℃ and humidity of 45%-55% for one week using a day-night cycle. Fine hairs on the rats' forelimbs and parietal region were removed with a depilatory cream, and the rats were allowed one day to acclimatize.

[0290] (2) Experimental method: ①Grouping and Dosing: Before the experiment, 60 rats were acclimatized for 7 days. To eliminate individual differences, after measuring body weight and conducting open field tests, 60 rats with similar indicators were selected and divided into 6 groups (n=10), with half males and half females. These were: blank group, model group, positive control drug fluoxetine hydrochloride hydrogel microneedle group, Bupleurum-Paeonia lactiflora drug-loaded microneedle group (including only the auxiliary needle), single press needle group, and Bupleurum-Paeonia lactiflora hydrogel press needle group (i.e., drug-loaded microneedle patch, including the main needle and auxiliary needle). Drug administration was administered while the model was being established. The control group received no stimulation, while the other groups received continuous transdermal drug administration combined with electrical stimulation for 56 days. The model group received blank hydrogel press needles combined with electrical stimulation to establish the model.

[0291] During the experiment, the inner side of the rat's forelimb was routinely disinfected with 75% ethanol at the Neiguan acupoint (located approximately 3 mm from the wrist joint between the radius and ulna) and the Baihui acupoint (located at the midline of the parietal bone). The product was then applied to the skin corresponding to the acupoints using tweezers, and an electrical stimulation device (20Hz, ≤5mA) was placed on top.

[0292] ②CUMS Depression Model: After 7 days of acclimatization, rats were individually housed in all groups except the control group (which was kept together). The model group, the positive control group (fluoxetine hydrochloride hydrogel microneedle group), the Bupleurum-Paeonia lactiflora drug-loaded microneedle group, the single-needle press group, and the Bupleurum-Paeonia lactiflora drug-loaded hydrogel press group were all subjected to 56 days of chronic unpredictable mild stress (CUMS). Nine stimulation methods were used: fasting, water deprivation, ice water swimming, day-night reversal, tail clamping, damp bedding, restraint, noise stimulation, and electric shock to the paws. One to two stimulation methods were used daily without repetition, ensuring the rats could not predict the next stimulation method. Each 7-day cycle lasted for 8 weeks (56 days). The rats' condition was observed daily, and their weight was measured weekly.

[0293] Fasting: Deprive the animal of feed for 24 hours.

[0294] Water restriction: Deprivation of drinking water for 24 hours.

[0295] Ice water swimming: Rats were placed in a large bucket with a diameter of 20 cm, a height of 50 cm, and a water depth of 30 cm, and swam in ice water at 4 ℃ for 5 minutes.

[0296] Day and night reversed: 12 hours of darkness during the day, 12 hours of light at night.

[0297] Tail clamping: Clamp the rat's tail 1 cm from the base with tweezers for 2 minutes.

[0298] Moist bedding: Mice were placed in bedding that had been moistened with water for 24 hours.

[0299] Restraint: Rats were placed inside mineral water bottles to restrict their free movement for 3 hours.

[0300] Noise stimulation: Create noise for 3 hours (40 MHz) using an ultrasonic cleaner.

[0301] Electric shock to the soles of the feet: 36 V voltage, one shock every 10 seconds, each shock lasting 2 seconds, for a total of 10 shocks.

[0302] The specific stimulus schedule for the CUMS model is shown in the table below.

[0303] Stimulation schedule of the CUMS model

[0304] ③ Behavioral tests: (a) Body weight test; (b) Food intake test; (c) Open field test (OFT); (d) Sugar water preference test (SPT); (e) Forced swimming test (FST).

[0305] ④ Statistical processing: Statistical analysis and graphing were performed using GraphPad Prism 8.0 software (GraphPad Software, USA). All data are expressed as mean ± standard deviation (±s). Independent samples t-tests were used for comparisons between two groups, and one-way ANOVA with LSD test was used for pairwise comparisons among multiple groups.

[0306] (3) Experimental results: ①Weight: In week 1, there were no statistically significant differences among the groups of rats (P>0.05), and their body weight remained relatively consistent, as shown in the table below. After 8 weeks of modeling, the body weight of the model group rats was significantly lower than that of the control group (P<0.01). Compared with the depressed rats in the model group, there were significant differences in the Bupleurum-Paeonia lactiflora drug-loaded microneedle group, the single press needle group, and the Bupleurum-Paeonia lactiflora drug-loaded hydrogel press needle group (P<0.05).

[0307] Table of weight changes of rats in each group (g) ± s , n =10)

[0308]

[0309] ②Food intake: Compared with the control group, the food intake of the CUMS-induced model rats was significantly lower than that of the control group, and the difference was statistically significant (P<0.01). Compared with the model group, the food intake of rats in the Bupleurum-White Peony hydrogel press needle group, the Bupleurum-White Peony drug-loaded microneedle group, and the single press needle group was significantly increased (P<0.05), as shown in the table below. This indicates that the combined treatment of Bupleurum-White Peony extract and press needles can, to some extent, improve the decrease in food intake induced by CUMS in rats.

[0310] Food intake table of rats in each group at week 8 ( ± s , n =10)

[0311]

[0312] ③Open field experiment: The open field test is mainly used to assess the rats' ability to move autonomously and explore in unfamiliar environments, thereby determining their depressive status. As shown in the table below, the number of times the model group rats stood upright was significantly less than that of the control group rats, with a statistically significant difference (P<0.01), indicating that the autonomous exploration ability of depressed rats was reduced. Compared with the model group, the number of times the rats stood upright was significantly increased in the Bupleurum-White Peony hydrogel press needle group, the Bupleurum-White Peony drug-loaded microneedle group, and the single press needle group (P<0.05), indicating that the combined treatment of Bupleurum-White Peony extract and press needles can significantly improve the reduced activity level in rats caused by CUMS.

[0313] Table of the number of times rats stood upright in open field experiment for each group ( ± s , n =10)

[0314]

[0315] ④ Sugar water preference experiment: The sucrose preference test is mainly used to assess anhedonia in depressed rats. Anhedonia leads to a decreased preference for sucrose in rats. Before modeling, there was no significant difference in sucrose preference rates among the different groups. After 8 weeks of CUMS modeling, compared with the control group, the sucrose preference rate in the model group was significantly lower (P<0.01), indicating anhedonia in the rats. Compared with the model group, the sucrose preference rate was higher in the Bupleurum-Paeonia lactiflora hydrogel press needle group, the Bupleurum-Paeonia lactiflora drug-loaded microneedle group, and the single press needle group (P<0.05), indicating that Bupleurum-Paeonia lactiflora extract combined with press needle therapy has a certain ameliorative effect on anhedonia induced by CUMS in depressed rats, and the combined treatment of Bupleurum-Paeonia lactiflora extract with press needle therapy is more effective than the single press needle group and the single drug administration group. (See table below.) Table of sucrose preference rates in each group of rats ( ± s , n =10)

[0316]

[0317] ⑤ Forced swimming: The forced swimming test was used to represent the despair behavior of depressed rats, with the immobility time reflecting the rats' despair state. As shown in the table below: Compared with the control group, the immobility time of the model group rats was significantly longer (P<0.01), indicating that the model group rats clearly exhibited despair behavior. Compared with the model group, the immobility time of rats in the Bupleurum-Paeonia lactiflora hydrogel press needle group, the Bupleurum-Paeonia lactiflora drug-loaded microneedle group, and the single press needle group was significantly shorter (P<0.05), indicating a stronger will to live.

[0318] The timetable for forced swimming immobility in each group of rats ( ± s , n =10)

[0319]

[0320] (4) Summary: Using CUMS-depressed rats as the research subjects, and taking rat body weight, food intake, open field test, sucrose preference test, and forced swimming test as behavioral indicators, this study observed and compared the effects of the Bupleurum-White Peony hydrogel press-fit group (drug-loaded microneedle patch, including main needle and auxiliary needle), the Bupleurum-White Peony drug-loaded microneedle group (only auxiliary needle), and the single press-fit group on improving the behavioral characteristics of CUMS-depressed rats. The results showed that, compared with the control group, the model rats had smaller body weight, lower food intake, fewer open field standing times, lower sucrose preference rate, and longer immobile time during forced swimming. This indicated that the model group rats experienced decreased appetite, loss of novelty in free exploration, a gradually decreasing preference for sucrose, and poor mental state, demonstrating the successful establishment of the CUMS-depressed rat model. Compared with the model group, the Bupleurum-White Peony hydrogel press needle group, the Bupleurum-White Peony drug-loaded microneedle group, and the single press needle group showed increased uprighting frequency, sugar water preference rate, and food intake, and shortened forced swimming immobility time. This indicates that the Bupleurum-White Peony combined with press needle and the positive drug group can significantly improve the rats' autonomous exploration activity, deepen their preference for sugar water, and improve their mental state, thus exhibiting a good antidepressant effect.

[0321] Implementation Method 17: A specific embodiment is provided to verify the safety and physiological response of electrical stimulation.

[0322] Test conditions: 30 healthy subjects (male-female ratio 1:1, age 25–45 years) wore the device on the Neiguan acupoint. Stimulation parameters: frequency 20±2 Hz, intensity stepped increase (1–5 mA).

[0323] Key results: (1) Safety: The current intensity of all subjects was stable within the set value ±0.2 mA (current limiting resistor 1–5 kΩ), and there were no skin burns or muscle rigidity (compliant with IEC 60601-1 standard).

[0324] (2) Neural activation effect: When the intensity was ≥3 mA, fMRI showed that the activity inhibition rate of the default mode network (DMN) reached 32.7±5.1% (p<0.01 vs. baseline), confirming the regulatory effect on the central nervous pathway.

[0325] Implementation Method 18: A specific embodiment is provided for the accuracy evaluation of multimodal physiological monitoring (multi-source physiological data acquisition module).

[0326] (1) Heart rate detection module: PPG accuracy: Compared with medical ECG (n=120 cases), the mean absolute error (MAE) of real-time heart rate was 1.2±0.3 bpm, and the RMSE of HRV was ≤5 ms (r=0.96, p<0.001).

[0327] Association with depressive state: The low-frequency / high-frequency power ratio (LF / HF) of HRV was significantly increased in patients with major depressive disorder (PHQ-9>20) (2.8±0.6 vs. 1.5±0.4 in the healthy group, p<0.001).

[0328] (2) Sleep data acquisition module and sleep quality analysis module:

[0329] Note: The algorithm based on random forest + LSTM was validated in a multi-center trial in three tertiary hospitals and met the clinical diagnostic needs (Kappa>0.8 indicates high consistency).

[0330] Implementation method 19: A specific embodiment is provided to perform performance testing on the AI ​​control module.

[0331] (1) Optimization of dynamic treatment parameters Reinforcement learning response speed: After the system detects a decrease in HRV >15%, it automatically increases the intensity of electrical stimulation (in 0.5 mA increments) within 200 ms, and the patient's depressive mood score (PHQ-9) decreases by 18.3±4.7% within 7 days (p=0.003).

[0332] Acupoint recommendation accuracy: The 3D CNN+GNN model achieved a 93.6% accuracy rate in recommending acupoints (compared to expert acupuncturist assessments).

[0333] (2) Effectiveness of personalized treatment Grouped trial: 60 patients with moderate depression were randomly divided into two groups (Group A: conventional electrical stimulation; Group B: AI closed-loop system). Results after 4 weeks:

[0334]

[0335] Implementation Method 20: A specific embodiment is provided to verify the system integration performance.

[0336] Edge computing latency: TensorFlow Lite models have an inference latency of <15ms on the neuromorphic chip (Loihi 2) with a power consumption of 2.8 mAh / 72h.

[0337] User compliance: Data visualization on the APP (patient client and doctor client) increased patient treatment compliance rate to 89.5% (compared to 67.2% with traditional devices).

[0338] Privacy protection: Cross-institutional model training under the federated learning framework with zero leakage of raw data (AES-256 encryption + differential privacy, ε=0.5).

[0339] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, reasonable combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A multimodal closed-loop feedback press needle electrical stimulation antidepressant system, characterized in that, The system includes the following modules: Patient profiling module: Collects patient personal information and builds patient profiles; Multi-source physiological data acquisition module: Collects multi-source physiological data from patients; Sleep data acquisition module: Employs multimodal sensor fusion technology to collect patient sleep data through a non-invasive wearable device; Sleep quality analysis module: Obtains sleep quality data based on patient sleep data analysis; Effect feedback information collection module: Collects patient effect feedback information for each acupoint electrical stimulation and skin drug administration; AI regulation module: Generates a synergistic antidepressant plan of acupoint electrical stimulation and skin drug delivery based on patient profile, multi-source physiological data, sleep quality data and treatment effect feedback information, and predicts treatment parameters; the treatment parameters include acupuncture points and electrical stimulation parameters for acupoint electrical stimulation and drug release rate for skin drug delivery; Synergistic Therapy Module: Based on predicted treatment parameters, patients receive synergistic therapy involving acupoint electrical stimulation and transdermal drug delivery.

2. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 1, characterized in that, The multi-source physiological data includes the patient's electroencephalogram (EEG) data, electrocardiogram (ECG) data, skin conductance data, and heart rate data.

3. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 1, characterized in that, The patient sleep data includes patient body movement frequency during sleep, patient blood oxygen saturation, patient pulse rate variability, ambient temperature and humidity data, and ambient noise data.

4. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 1, characterized in that, The sleep quality data includes: sleep stage data, sleep apnea event data, and sleep statistics; the sleep statistics include sleep efficiency, sleep latency, REM sleep percentage, and number of awakenings.

5. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 4, characterized in that, The sleep quality analysis module includes: a sleep staging model, a sleep apnea identification model, and a sleep data statistics unit; The sleep staging model is a random forest classifier: it distinguishes sleep stages based on body movement frequency analysis to obtain sleep stage data; The sleep apnea recognition model is an LSTM network: based on the synchronous monitoring results of the patient's blood oxygen saturation and pulse rate variability during sleep, sleep apnea events are identified and sleep apnea event data is obtained; The sleep data statistics unit calculates sleep efficiency, sleep latency, REM sleep percentage, and number of awakenings based on the patient's sleep data.

6. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 1, characterized in that, The AI ​​control module includes control software, edge computing devices, a cloud platform, and a user interaction terminal; The control software device includes an AI model, a storage unit, a training unit, and an interactive analysis unit; The AI ​​model is integrated into an edge computing device and is used to predict treatment parameters; The storage unit, training unit, and interactive analysis unit are integrated in the cloud platform; the storage unit is used to store data; the training unit is used to train AI models; and the interactive analysis unit is used to interact with users through user interaction terminals.

7. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 6, characterized in that, The AI ​​model uses reinforcement learning and online learning mechanisms to predict treatment parameters; The reinforcement learning involves learning the optimal treatment parameters by continuously adjusting treatment parameters based on patient profiles, multi-source physiological data, sleep quality data, and treatment effect feedback. The online learning involves continuously updating and learning the AI ​​model based on multi-source physiological data, sleep quality data, and treatment effect feedback.

8. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 6, characterized in that, The training unit employs few-shot learning, transfer learning, federated learning, and encrypted learning mechanisms to train the AI ​​model.

9. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 6, characterized in that, The interactive analysis unit includes an AI interpretability component; the user interaction terminal includes a doctor's mobile terminal. The AI ​​explainability component integrates SHAP and LIME technologies to provide explanations for the AI ​​model's decisions and pushes the explanation results to doctors via their mobile devices.

10. The multimodal closed-loop feedback press needle electrical stimulation antidepressant system according to claim 1, characterized in that, The synergistic treatment module includes an electrical stimulation device, a microneedle triggering device, and a drug-loaded microneedle patch. The drug-loaded microneedle patch includes: a group of medical stainless steel main needles, a group of hydrogel drug-loaded auxiliary needles, and adhesive tape. The adhesive tape is thicker in the middle and thinner around the edges; the medical stainless steel main needle group and the hydrogel drug-loaded auxiliary needle group are fixed in the middle area of ​​the adhesive tape. The medical stainless steel main needle group includes 4 medical stainless steel main needles; the 4 medical stainless steel main needles are symmetrically distributed around the center of the adhesive tape; The hydrogel drug-loaded needle array is made of hydrogel material; the hydrogel drug-loaded needle array includes multiple drug-loaded needles, which are drug-loaded microneedles; the tips of the drug-loaded microneedles are uniformly distributed with drug-loaded microspheres; the drug-loaded microspheres are pH-responsive microspheres, and their drug release rate is regulated by changes in the pH value of the skin surface. The hydrogel drug-loaded auxiliary needle group is divided into 4 regions, each region is arranged in an octagonal array; the 4 regions are distributed around the periphery of the medical stainless steel main needle group, forming the shape of 4 petals of a lilac flower; the needle height of the medical stainless steel main needle group is greater than the needle height of the hydrogel drug-loaded auxiliary needle group. The medical stainless steel main needle group is electrically connected to the electrical stimulation device and receives electrical stimulation pulse current; The hydrogel drug-loaded accessory needle group is equipped with electrodes for electrical connection with the microneedle triggering device; The electrical stimulation device is signal-connected to the AI ​​control module and is used to transmit electrical stimulation pulse current to the medical stainless steel main needle group according to the electrical stimulation parameters of the acupoints predicted by the AI ​​control module. The medical stainless steel main needle group is used to insert into the predicted acupoints on the patient's body for acupoint electrical stimulation, thereby performing acupoint electrical stimulation on the patient. The microneedle triggering device is signal-connected to the AI ​​control module and is used to conduct a drug release triggering current to the hydrogel microneedles according to the drug release rate predicted by the AI ​​control module for skin administration. The drug release triggering current is used to regulate the pH value of the skin surface through electrolysis, thereby regulating the drug release rate of the drug-loaded microspheres.

Images