Percutaneous lung channel acupoint electrical stimulation optimization system based on artificial intelligence

Abstract

The invention relates to the field of intelligent medical treatment, in particular to a percutaneous lung channel acupoint electrical stimulation optimization system based on artificial intelligence, which comprises a multi-source data acquisition module (M1), a dynamic modeling module (M2) and a self-adaptive electrical stimulation module (M3). M1, collecting lung channel acupoint skin impedance, infrared thermal imaging, respiratory rhythm and cough frequency data in real time, and generating multi-source fusion features through space-time alignment; m2, based on a deep learning model, combining with a preset database to dynamically generate an electrical stimulation parameter combination, and ensuring parameter safety through impedance calibration and constraint optimization; and M3, adjusting the stimulation intensity according to the parameters, and setting an impedance threshold to trigger power-off protection. The traditional Chinese medicine meridian theory and the AI algorithm are fused, accurate acupoint positioning, personalized electrical stimulation and dynamic safety regulation are achieved, and the respiratory system disease treatment effect is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent medical technology, and specifically to an artificial intelligence-based transcutaneous lung meridian acupoint electrical stimulation optimization system. Background Art

[0002] In the field of Traditional Chinese Medicine stimulation therapy, traditional technologies rely on fixed-parameter stimulation patterns, which present three core bottlenecks: First, the electrophysiological characteristics of acupoints vary dynamically with individual differences and the treatment process, and existing equipment lacks multimodal biosensing capabilities, making it difficult to quantify acupoint status in real time; second, parameter adjustment relies on physician experience and cannot achieve adaptive coordinated regulation based on dynamic physiological signals such as respiratory rhythm and cough frequency, resulting in fluctuations in therapeutic efficacy;

[0003] In recent years, although some studies have attempted to introduce single bioelectric signals or thermal imaging technology, they have failed to solve the problem of spatiotemporal alignment and fusion of multi-source heterogeneous data, and most AI models are static parameter optimization, which cannot dynamically correct the mapping relationship between ancient book theories and real-time biological data. Existing transcutaneous electrical stimulation systems mostly rely on fixed parameters, lack dynamic response to the patient's physiological state and environmental changes, and the acupoint positioning accuracy is insufficient, which can easily lead to differences in efficacy or skin damage. Traditional methods find it difficult to fuse multi-source data (such as impedance, thermal distribution) to optimize stimulation strategies, and the safety assurance mechanism is imperfect. Therefore, there is an urgent need for an acupoint electrical stimulation system that integrates multimodal perception, dynamic knowledge-driven, and intelligent closed-loop control to break through the limitations of existing technologies. Summary of the Invention

[0004] In order to solve the technical problems mentioned in the current background technology, the purpose of the present invention is to provide an artificial intelligence-based transcutaneous lung meridian acupoint electrical stimulation optimization system and method.

[0005] To this end, the technical solution adopted in the present invention is as follows:

[0006] The artificial intelligence-based transcutaneous lung meridian acupoint electrical stimulation optimization system is characterized by including:

[0007] M1, multi-source data acquisition module, collects multi-source data including the patient's lung meridian acupoint skin impedance data, acupoint area thermal distribution image, respiratory rhythm and cough frequency;

[0008] Multi-source data is timestamped by the data transmission unit and then encapsulated and transmitted to the dynamic modeling module;

[0009] M2, a dynamic modeling module, receives the multi-source data and performs feature extraction and spatiotemporal alignment operations on the multi-source data;

[0010] calibrating the multi-source data with a preset database to correct the preset database;

[0011] Inputting multi-source data into a preset deep learning model to generate electrical stimulation parameter combinations;

[0012] M3, adaptive electrical stimulation module, adjusts the corresponding electrical stimulation parameters of the lung meridian acupoints in real time according to the electrical stimulation parameter combination,

[0013] A skin impedance threshold comparator is set, and when the amplitude of the skin impedance exceeds the dynamic safety threshold generated by the threshold comparator, power-off protection is triggered.

[0014] Furthermore, the skin impedance data of the lung meridian acupoints is collected in real time by an impedance sensor, and the electrical characteristics of the acupoints are quantified based on an equivalent circuit model:

[0015]

[0016] Where Z(ω) is the impedance (Ω) at the acquisition frequency ω, R s is the skin equivalent series resistance (Ω), C dl is the double layer capacitance (unit: F), R ct is the charge transfer resistance (Ω), j2πf is the complex frequency term, j is the imaginary unit, and represents the phase shift of the impedance.

[0017] The impedance amplitude |Z(ω=10)| at the characteristic frequency of 10 kHz is extracted and transmitted to the skin impedance threshold comparator in M3.

[0018] Furthermore, the thermal distribution image of the acupoint area is collected by infrared thermal imaging technology, and the acupoint area is locked according to a preset acupoint coordinate database.

[0019] The respiratory rhythm is quantified using chest displacement collected by an inertial sensor array arranged around the Tanzhong acupoint.

[0020] The cough frequency was quantified using a piezoelectric signal collected by a PVDF piezoelectric film placed on the sternocleidomastoid muscle.

[0021] Furthermore, the feature extraction includes:

[0022] 1) Skin impedance quantization: by extracting the impedance amplitude |Z(ω0)| of 11 lung meridian acupoints at a characteristic frequency of 10kHz, an 11-dimensional impedance dynamic vector V is constructed. z (t),

[0023]

[0024] Among them, Z i (f c =10kHz) is the characteristic frequency of the i-th lung meridian acupoint at f c= impedance amplitude at 10kHz, N is the normalization coefficient, t is the time variable,

[0025] 2) Dimensionality reduction of thermal distribution images: The thermal distribution images of the acupoint area are segmented to extract a 5×5 pixel area at the center of the acupoint, and the temperature gradient characteristics are calculated:

[0026]

[0027] in, is the temperature gradient feature (unit: °C / pixel), which indicates the temperature change rate of the acupoint area. is the spatial gradient of temperature in the x and y directions, ROI v The central area of ​​the acupuncture point.

[0028] 3) Generate respiratory cycle: Output respiratory cycle T through chest displacement and piezoelectric signal resp (t),

[0029] 4) Extract cough event marker E from PVDF piezoelectric signal by wavelet transform cough (t)={0,1},

[0030] 5) Environmental parameter embedding: Use environmental sensors to collect temperature and humidity data, and normalize the temperature and humidity data to [Temp(t), Humid(t)]∈[0,1] 2 .

[0031] Furthermore, the calibration in M2 is based on the 11-dimensional impedance dynamic vector V z (t) Modify the preset database weight matrix W ij :

[0032]

[0033] in, is the modified weight matrix, is the initial weight matrix, α is the adjustment coefficient, which controls the impact of real-time impedance difference on the weight matrix, |Z i (ω)-Z j (ω)| is the difference in impedance values ​​between acupoints i and j at the acquisition frequency ω, and max(Z)-min(Z) is the maximum and minimum impedance values ​​of all acupoints.

[0034] Furthermore, the deep learning model includes an input layer, a strategy output layer and an optimization layer.

[0035] The input layer data is fused to construct a 128-dimensional input vector X t :

[0036]

[0037] And generate a 3D multi-source tensor X through the sliding window B×T×D (B = batch size, T = time steps, D = 128),

[0038] The strategy output layer generates an electrical stimulation parameter combination [f v ,τ v ,I v ]∈R 11×3 , covering 11 acupuncture points, among which f v represents the electrical stimulation frequency, τ v represents the electrical stimulation current, τ v represents the electrical stimulation pulse width;

[0039] The optimization layer means that the combination of electrical stimulation parameters must meet the following restrictions: v with I v The product of is less than the first threshold, τ v It is within the set range, otherwise the deep learning model is retrained.

[0040] Furthermore, the skin impedance threshold comparator is implemented as follows:

[0041] Receive the current skin impedance amplitude mean Z0 from M1 every 5 seconds and calculate the dynamic safety threshold:

[0042]

[0043] Where ΔZ thres is the dynamic safety threshold, is the impedance change rate.

[0044] Compared with the prior art, the advantages of the present invention are:

[0045] 1. This invention uses multimodal biosensing technology to integrate skin impedance spectrum, thermal distribution and respiratory rhythm data in real time to achieve dynamic feature-driven electrical stimulation parameter optimization, solving the problem of efficacy deviation caused by individual differences and fluctuations in physiological status in traditional equipment.

[0046] 2. This invention uses an AI model to autonomously analyze the patient's respiratory rhythm and pathological characteristics, dynamically adjust the electrical stimulation strategy, and cover different disease courses and physical differences to achieve personalized treatment optimization;

[0047] 3. The present invention constructs an impedance threshold-electrical parameter dual closed-loop feedback mechanism to monitor abnormal skin electrical response in real time and link power-off protection, forming a multi-dimensional safety protection system to ensure treatment reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0049] Figure 1 Schematic diagram of the system workflow of the present invention;

[0050] Figure 2 This is a schematic diagram of the core process of the multimodal data acquisition module of the present invention;

[0051] Figure 3 This is a schematic diagram of the core process of the AI ​​dynamic modeling module of the present invention. DETAILED DESCRIPTION

[0052] To achieve the above objectives, the present invention is implemented through the following technical solutions. The present invention provides an artificial intelligence-based transcutaneous lung meridian acupoint electrical stimulation optimization system, which includes:

[0053] In summary, the advantage of the present invention lies in its ability to accurately analyze acupoint states through multimodal fusion, which can achieve individualized parameter optimization through dynamic knowledge graphs and deep reinforcement learning, breaking through the limitations of traditional single empirical regulation.

[0054] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. The artificial intelligence-based transcutaneous lung meridian acupoint electrical stimulation optimization system is characterized by: include; M1, multi-source data acquisition module, collects multi-source data including the patient's lung meridian acupoint skin impedance data, acupoint area thermal distribution image, respiratory rhythm and cough frequency; Multi-source data is timestamped by the data transmission unit and then encapsulated and transmitted to the dynamic modeling module; M2, a dynamic modeling module, receives the multi-source data and performs feature extraction and spatiotemporal alignment operations on the multi-source data; calibrating the multi-source data with a preset database to correct the preset database; Inputting multi-source data into a preset deep learning model to generate electrical stimulation parameter combinations; M3, adaptive electrical stimulation module, adjusts the corresponding electrical stimulation parameters of the lung meridian acupoints in real time according to the electrical stimulation parameter combination, A skin impedance threshold comparator is set, and when the amplitude of the skin impedance exceeds the dynamic safety threshold generated by the threshold comparator, power-off protection is triggered.

2. The system according to claim 1, wherein: The skin impedance data of the lung meridian acupoints is collected in real time by an impedance sensor, and the electrical characteristics of the acupoints are quantified based on an equivalent circuit model: Where Z(ω) is the impedance (Ω) at the acquisition frequency ω, R s is the skin equivalent series resistance (Ω), C dl is the double layer capacitance (unit: F), R ct is the charge transfer resistance (Ω), j2πf is the complex frequency term, j is the imaginary unit, and represents the phase shift of the impedance. The impedance amplitude |Z(ω=10)| at the characteristic frequency of 10 kHz is extracted and transmitted to the skin impedance threshold comparator in M3.

3. The system according to claim 1, wherein: The thermal distribution image of the acupoint area is collected by infrared thermal imaging technology, and the acupoint area is locked according to the preset acupoint coordinate database. The respiratory rhythm is quantified using chest displacement collected by an inertial sensor array arranged around the Tanzhong point. The cough frequency was quantified using a piezoelectric signal collected by a PVDF piezoelectric film placed on the sternocleidomastoid muscle.

4. The system according to claim 3, characterized in that The feature extraction includes: 1) Skin impedance quantization: by extracting the impedance amplitude |Z(ω0)| of 11 lung meridian acupoints at a characteristic frequency of 10kHz, an 11-dimensional impedance dynamic vector V is constructed. z (t), Among them, Z i (f c =10kHz) is the characteristic frequency of the i-th lung meridian acupoint at f c = impedance amplitude at 10kHz, N is the normalization coefficient, t is the time variable, 2) Dimensionality reduction of thermal distribution images: The thermal distribution images of the acupoint area are segmented to extract a 5×5 pixel area at the center of the acupoint, and the temperature gradient characteristics are calculated: in, is the temperature gradient feature (unit: °C / pixel), which indicates the temperature change rate of the acupoint area. is the spatial gradient of temperature in the x and y directions, ROI v The central area of ​​the acupuncture point. 3) Generate respiratory cycle: Output respiratory cycle T through chest displacement and piezoelectric signal resp (t), 4) Extract cough event marker E from PVDF piezoelectric signal by wavelet transform cough (t) = {0, 1}, 5) Environmental parameter embedding: Use environmental sensors to collect temperature and humidity data, and normalize the temperature and humidity data to [Temp(t), Humid(t)]∈[0,1] 2 .

5. The system according to claim 4, characterized in that The calibration of M2 is based on the 11-dimensional impedance dynamic vector V z (t) Modify the preset database weight matrix W ij : in, is the modified weight matrix, is the initial weight matrix, α is the adjustment coefficient, which controls the impact of real-time impedance difference on the weight matrix, |Z i (ω)-Z j (ω)| is the difference in impedance values ​​between acupoints i and j at the acquisition frequency ω, and max(Z)-min(Z) is the maximum and minimum impedance values ​​of all acupoints.

6. The system according to claim 4, characterized in that The deep learning model includes an input layer, a strategy output layer and an optimization layer. The input layer data is fused to construct a 128-dimensional input vector X t : And generate a 3D multi-source tensor X through the sliding window B×T×D (B = batch size, T = time steps, D = 128), The strategy output layer generates an electrical stimulation parameter combination [f v ,τ v ,I v ]∈R 11 ×3 , covering 11 acupuncture points, among which f v represents the electrical stimulation frequency, τ v represents the electrical stimulation current, τ v represents the electrical stimulation pulse width; The optimization layer means that the combination of electrical stimulation parameters must meet the following restrictions: v with I v The product of is less than the first threshold, τ v It is within the set range, otherwise the deep learning model is retrained.

7. The system according to claim 6, characterized in that The skin impedance threshold comparator is implemented as follows: Receive the current skin impedance amplitude mean Z0 from M1 every 5 seconds and calculate the dynamic safety threshold: Where ΔZ thres is the dynamic safety threshold, is the impedance change rate.

Images