Controlled drug injection monitoring system and armlet

Abstract

The invention relates to the technical field of medical monitoring information processing, in particular to a controlled medicine injection monitoring system and armlet, and the system comprises a controlled medicine injection monitoring armlet and a cloud processing module; the control drug injection monitoring armlet comprises a monitoring module, an information transmission module and a drug injection module, the monitoring module is used for collecting vital sign information of a user; the cloud processing module is used for constructing a deep learning model and determining a corresponding pain level; and the drug injection module performs estimation according to a pharmacokinetic model and a pain level through an extended unscented Kalman filter system, establishes a control model for optimization to determine a minimum-cost drug injection real-time scheme and dynamically performs drug injection. According to the method, automatic analysis of the pain condition of the user is achieved through a multi-modal analysis method, the medication scheme is dynamically adjusted in real time according to the state of the user in the injection process, and personalized analgesia optimization within the safety boundary is achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical monitoring information processing, in particular to a controlled drug injection monitoring system and an armband. Background Art

[0002] Current monitoring technology for injections of controlled drugs has several serious deficiencies, making it difficult to meet the actual needs of terminal patients for safe medication use. First, most medical institutions still rely on manual recording of injections, resulting in delayed information and prone to omissions, misrecording, or tampering, lacking real-time and accuracy. Second, existing systems lack the means to automatically identify and monitor injection behaviors, making it impossible to determine whether an injection actually occurred or whether the injection dosage is reasonable. It is also difficult to identify the identity of the injector, posing risks of "substitute injection," "misuse," or "abuse." At the same time, most systems lack an abnormal behavior warning mechanism, making it impossible to promptly detect problems such as frequent medication, overdose, or illegal use, increasing the risk of drug abuse.

[0003] Therefore, how to conduct closed-loop monitoring of the use of controlled drugs in real time, accurately and safely is an urgent problem that needs to be solved. Summary of the Invention

[0004] To solve one of the above-mentioned problems in the prior art, the present invention provides a controlled drug injection monitoring system and an armband.

[0005] To achieve the above objectives, the present invention provides, on one hand, a controlled drug injection monitoring system, which includes: a controlled drug injection monitoring armband and a cloud processing module; the controlled drug injection monitoring armband includes: a monitoring module, an information transmission module and a drug injection module; The monitoring module is used to collect vital sign information of the user wearing the controlled drug injection monitoring armband, wherein the vital sign information refers to multiple sets of data used to represent the degree of pain in the user's body; The information transmission module is used to send the vital sign information to the cloud processing module; The cloud processing module is used to construct a deep learning model, pre-train the deep learning model using the vital sign information to obtain a trained deep learning model, and obtain the user's current vital sign information data sequence in real time, using the trained deep learning model to determine the pain level corresponding to the current vital sign information data sequence based on the user's current vital sign information data sequence, and transmit the pain level to the information transmission module, wherein the current vital sign information data sequence includes the vital sign information collected at a preset period; The information transmission module is further configured to receive the pain level and transmit the pain level to the drug injection module; The drug injection module is used to obtain the pain level, estimate the real-time prediction information of the user's injection status based on the drug metabolism kinetics model and the pain level through an extended unscented Kalman filter system, use the real-time prediction information of the user's injection status to establish a control model for optimization to establish a real-time plan for drug injection with minimum cost, and dynamically perform drug injection according to the real-time plan for drug injection with minimum cost, wherein the real-time prediction information of the user's injection status includes: predicted blood drug concentration, predicted effect chamber concentration and predicted pain level.

[0006] Another aspect of the present invention provides a controlled drug injection monitoring armband, comprising: an armband body, a monitoring module, an analysis module, an information transmission module, and a drug injection module; The monitoring module is used to collect vital sign information of a user wearing the controlled drug injection monitoring armband through the armband body, wherein the vital sign information refers to multiple sets of data used to represent the degree of pain in the user's body; The analysis module is configured to construct a deep learning model, pre-train the deep learning model using the vital sign information to obtain a trained deep learning model, and acquire a current vital sign information data sequence of the user in real time, determine a pain level corresponding to the current vital sign information data sequence based on the current vital sign information data sequence of the user using the trained deep learning model, and transmit the pain level to the information transmission module, wherein the current vital sign information data sequence includes the vital sign information collected at a preset period; The information transmission module is further configured to receive the pain level, upload the pain level to a background processing center, receive drug injection authorization information returned by the background processing center, and unlock the use rights of the drug injection module according to the drug injection authorization information; The drug injection module is used to obtain the pain level, estimate the real-time prediction information of the user's injection status based on the drug metabolism kinetics model and the pain level through an extended unscented Kalman filter system, use the real-time prediction information of the user's injection status to establish a control model for optimization to establish a real-time plan for drug injection with minimum cost, and dynamically perform drug injection according to the real-time plan for drug injection with minimum cost, wherein the real-time prediction information of the user's injection status includes: predicted blood drug concentration, predicted effect chamber concentration and predicted pain level.

[0007] The beneficial effects of the present invention are reflected in the fact that a controlled drug injection monitoring system and armband are proposed, which are based on analysis of multimodal vital signs data, realize automatic analysis of the user's pain condition through multimodal analysis methods, accurately intervene in the user wearing the armband, and timely improve the user's life status. It can also accurately analyze the user's pain condition, so as to accurately set the required medication plan, and dynamically adjust the medication plan in real time according to the user's status during the injection process, so as to standardize the use of controlled drugs, prevent "misuse" or "abuse" of drugs, prevent mechanical damage and acute overdose of drugs, avoid fatal risks such as respiratory depression, and ultimately achieve personalized analgesia optimization within the safety boundary. BRIEF DESCRIPTION OF THE DRAWINGS

[0008] Figure 1 This is a schematic diagram of the structure of the controlled drug injection monitoring armband provided in Example 1 of the present invention; Figure 2 This is an example of the structural design of a controlled drug injection monitoring armband provided in Example 1 of the present invention; Figure 3 This is a schematic diagram of the structure of the controlled drug injection monitoring armband system provided in Example 2 of the present invention.

[0009] Figure numerals: 11-armband body; 12-monitoring module; 13-analysis module; 14-information transmission module; 15-drug injection module; 151-replaceable medicine stick; 152-indwelling needle interface; 16-button; 31-controlled drug injection monitoring armband; 32-cloud processing module. DETAILED DESCRIPTION

[0010] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0011] Example 1 This embodiment provides a controlled drug injection monitoring armband, the structural diagram of which is shown in FIG. Figure 1 As shown, the armband includes an armband body 11, a monitoring module 12, an analysis module 13, an information transmission module 14, and a drug injection module 15. The armband body 11 can be designed to be rolled up and wrapped around a human arm or other animal limbs, serving as a carrier for the other modules. The monitoring module 12, analysis module 13, and information transmission module 14 can be implemented using electronic components such as sensors, chips, and circuits within the armband. These components can be located on or within the armband body 11.

[0012] In an alternative embodiment, each arm ring is uniquely bound to a user id, avoiding the risk of misuse.

[0013] The arm ring body 11 is used to contact the user in a wearable manner; specifically, the arm ring body 11 can be made of flexible wearable materials, and can include components for connecting to the human body (such as a wristband, a watchband, etc.) and components for carrying other modules (such as a watch body, a watch dial, a display screen, etc.).

[0014] As Figure 2 shows a possible structural design embodiment of the arm ring of the present embodiment, in which the arm ring body 11 mainly includes a wristband and a watch body with a display screen; the watch body is internally provided with electronic components (not shown in the figure) for carrying and implementing the monitoring module 12, the analysis module 13 and the information transmission module 14; the drug injection module 15 is connected to the watch body and controlled by the other modules.

[0015] The monitoring module 12 is used to collect the vital sign information of the user wearing the arm ring body 11 of the regulated drug injection monitoring arm ring, and the vital sign information refers to a plurality of groups of data for representing the pain degree of the user's body; specifically, when the cancer end-stage patient has pain, the patient may have high blood pressure, rapid heartbeat, body tremor and other phenomena. Therefore, the monitoring module 12 needs to monitor the vital signs of the user to judge the pain condition of the patient, and the vital sign information can at least include heart rate value, blood oxygen value, body temperature value, blood pressure value and motion state, etc. which can reflect the life state of the user. The monitoring module 12 can include a heart rate monitoring unit, a blood oxygen monitoring unit, a body temperature monitoring unit, a blood pressure estimation unit, a motion state detection unit, etc. for monitoring and collecting various vital signs of the user. For example, the heart rate monitoring unit can use a PPG (Photoplethysmography, photoelectric plethysmography) sensor to obtain the average heart rate data sequence of the period , wherein represents the average heart rate data of the first period, N is an integer, indicating how many periods of data are taken; the blood oxygen monitoring unit can use dual-wavelength PPG to monitor and obtain the blood oxygen data sequence , wherein represents the average blood oxygen data of the first period; the body temperature monitoring unit can use a negative temperature coefficient thermistor (Negative Temperature Coefficient thermistor, abbreviated as NTC thermistor) to obtain the body temperature data sequence , wherein Indicates the average body temperature data of the first cycle; the blood pressure monitoring unit still uses PPG to obtain. Since PPG cannot directly obtain the user's blood pressure data, the PPG original wave signal can be directly obtained for subsequent processing; the motion state detection unit can use the three-axis accelerometer to determine the user's motion frequency data sequence ,in Indicates the average exercise frequency of the first cycle. Except for blood pressure data, the cycles of other data are the same as heart rate data.

[0016] In a specific embodiment, the specific method for obtaining blood pressure data based on the PPG raw wave signal is as follows: first, parameters including rise time, fall time, waveform area, dicrotic wave characteristics, etc. are extracted from the PPG raw waveform; then, support vector regression is used to generate a blood pressure prediction value data sequence based on the aforementioned parameters. , which can use the same period as the heart rate data (for example, also 1 minute), where Represents the average blood pressure data of the first cycle.

[0017] Analysis module 13 is used to construct a deep learning model, pre-train the deep learning model using vital sign information to obtain a trained deep learning model, and then obtain the user's current vital sign information data sequence in real time. Using the trained deep learning model, it determines the pain level corresponding to the current vital sign information data sequence based on the user's current vital sign information data sequence, and transmits the pain level to information transmission module 14. The current vital sign information data sequence includes vital sign information collected at a preset period. Specifically, analysis module 13 is used to analyze the user's pain tolerance and can be embedded in the armband body. Analysis module 13 also requires pre-training the deep learning model before performing pain analysis, which can improve the accuracy of calculating pain levels based on vital sign information.

[0018] In an optional embodiment, the analysis module 13 uses a trained deep learning model to determine the pain level corresponding to the current vital sign information data sequence based on the user's current vital sign information data sequence, specifically including: grouping the current vital sign information data sequence according to the time dimension and aligning it, and constructing a first multimodal vector; sending the first multimodal vector into the trained deep learning model for calculation to obtain first calculation data, comparing the first calculation data with the preset reference data, and determining the pain level corresponding to the current vital sign information data sequence. Specifically, the current vital sign information can represent the user's physical state, and the current vital sign information can include the aforementioned feature data sequence, such as 、 、 、 、 These data sequences may include N data, N The value of can be determined according to the needs. For example, if the data sequence acquisition period is 1 minute, N 10 can reflect the user's physical condition in the past 10 minutes. N A value of 30 can reflect the user's physical condition over the past 30 minutes. The first calculated data is obtained by running the current vital sign information data sequence through a trained deep learning model. The preset reference data is a pain level reference data defined in relevant medical science, which can be implemented in the form of a stepped threshold. By comparing the first calculated data with the values ​​of each step of the stepped threshold, the closest pain level is determined as the final result. This comparison can be implemented using Euclidean distance.

[0019] In one optional embodiment, a deep learning model includes a Transformer encoder (a deep learning architecture) and a multi-layer perceptron (MLP). Inputting a first multimodal vector into a trained deep learning model for computation includes: utilizing the Transformer encoder to perform computations on the first multimodal vector using a multi-head attention mechanism, performing residual connections and normalization, and obtaining a final layer output; and reconstructing the final layer output through the MLP. Specifically, the Transformer encoder can have multiple layers, with residual connections and normalization performed on the vector data as it passes through each layer. The final layer output is the result of the first multimodal vector data after the final layer of the Transformer encoder. Residual connections make deep networks easier to train, while normalization stabilizes data distribution and accelerates training. This embodiment utilizes the Transformer encoder's multi-head attention mechanism, allowing the model to simultaneously focus on information at different locations, while the MLP integrates and transforms this information, improving the model's generalization capability. By combining this with appropriate MLP parameter design, computational overhead can be reduced while maintaining performance.

[0020] In an optional embodiment, the analysis module 13 specifically trains the deep learning model in the following manner: the analysis module 13 receives a sequence of vital sign information samples collected by the monitoring module 12 from the user's vital sign information; groups and aligns the sequence of vital sign information samples according to the time dimension, and constructs a second multimodal vector; uses the deep learning model to operate on the second multimodal vector to obtain a second reconstruction vector; obtains a third reconstruction vector and a fourth reconstruction vector, wherein the third reconstruction vector is obtained by operating the third multimodal vector using the deep learning model, and the fourth reconstruction vector is obtained by operating the fourth multimodal vector using the deep learning model, wherein the third multimodal vector is constructed based on the user's vital sign information in a previous non-painful state, and the fourth multimodal vector is constructed based on the user's vital sign information in a previous painful state; constructs a loss function based on the second reconstruction vector, the third reconstruction vector, and the fourth reconstruction vector; updates the parameters of the deep learning model through iterative training until the loss function converges, thereby obtaining a trained deep learning model. Specifically, the sequence of vital sign information samples can include the user's real vital sign information for a certain time period. By introducing the user's vital sign information in the past non-pain state and the user's vital sign information in the past pain state to train the deep learning model, better model parameters can be provided for the calculation of subsequent specific tasks, thereby improving model performance, reducing data requirements, speeding up training, and enhancing generalization capabilities.

[0021] In a specific embodiment, the analysis module 13 can complete the pre-training of the deep learning model and determine the user's current pain level based on the current vital sign information data sequence through the following specific steps: (1) Obtain the vital signs information of users in a certain period of time as training samples, including: average heart rate data series , blood oxygen data series , body temperature data series , motion frequency data series And blood pressure prediction value data series .

[0022] (2) The training samples 、 、 、 、 Align according to the time dimension and construct a multimodal vector ,in represents the field of real numbers; Indicates the time window length ( T It can indicate how long the data is used for analysis. When the period of the data series is 1 minute, T=N ); is the number of modalities, that is, the types of vital sign information. In this embodiment, the number of modalities is 5.

[0023] (3) Sent by The Transformer encoder consists of layers, and the output of each layer is calculated as follows: ; ; in, Represents the first layer, represents the normalized calculation, Indicates in l The intermediate amount of the layer after residual connection and normalization, Indicates the l The output of the layer, Indicates the The output of the layer, Indicates the use of multi-head attention mechanism processing, Indicates processing by a feedforward neural network. When calculating, first the multi-head attention mechanism ( ) applied to the previous layer The output of , then undergoes residual connection and layer normalization, producing .Then, Through the feedforward neural network ( ), and through another residual connection and normalization, we get the The final representation of the layer The output of the last layer of Transformer (denoted as ) is reconstructed again through the multi-layer perceptron to obtain .

[0024] (4) Using the user’s vital signs information when they were pain-free After the same calculation in step (3), we can get Specifically, the user's vital sign data in the past without pain is formed into a data sequence using the same sequence format as above The data of the user in the past without pain can be the historical vital sign data collected manually from the user, or it can be the historical data collected by the armband of the present invention before, as long as the statistical caliber of the data is the same.

[0025] (5) Obtaining the user’s vital signs information under previous illness conditions After the same calculation in step (3), we can get Similarly, the vital sign data of the user in the past pain condition is formed into a data sequence in the same sequence format as described above The data of the user in the past pain condition can be historical vital sign data artificially collected for the user, or historical data collected by the arm ring of the application before this time, as long as the statistical caliber of the data is the same.

[0026] (6) Construct a variable loss function based on the results of steps (3), (4) and (5) : ; wherein, and are binary labels (one is 1 and the other is 0). and represent the labels at different inputs, and during training, the input data is , and one of which is 0 and the other is 1, the data corresponding to the value of 0 is set to a random value within the range of human vital signs, and the value of 1 is retained. The real data. Specifically, when the value is (1, 0), it means that the vital sign data under the condition of no pain is retained ; when the value is (0, 1), it means that the vital sign data under the condition of pain is retained . The binary labels and take 0 and 1 mainly to distinguish whether the historical data is in the pain state or not in the pain state, and to ensure that the historical data trained at one time is only possible in the pain state or not in the pain state.

[0027] (7) Update the parameters of the Transformer encoder and the multilayer perception iteratively until convergence.

[0028] The above completes the pre-training of the model. When the trained model is used to calculate the current pain level of the user, the following steps are also included: (8) Obtain the current vital sign information data sequence, calculate the result by calculating the current vital sign information data sequence through the trained model, and evaluate the Euclidean distance between the calculation result and the reference data to determine the pain degree (in the form of step threshold).

[0029] (9) The analysis result containing the pain degree is transmitted to the information transmission module 14.

[0030] The information transmission module 14 is used to receive the pain level, upload the pain level to the background processing center, and receive the drug injection authorization information returned by the background processing center, and unlock the use rights of the drug injection module 15 according to the drug injection authorization information. Specifically, the background processing center can be a medical institution processing center with medical qualifications. The information transmission module 14 will upload the feedback of the analysis module 13 to the medical institution processing center, and the medical institution processing center will analyze and process it. Optionally, the medical institution processing center can determine whether to grant the user injection permission based on the results of the pain analysis. The user who obtains the injection permission can press the injection button to inject the controlled drug to relieve pain. If it is in a home scenario, it can also be set so that after detecting the pain to a certain level, the device can directly grant the injection permission without the need for background processing.

[0031] The drug injection module 15 is used to obtain pain levels. It uses an extended unscented Kalman filter (UKF) to estimate real-time predictions of the user's injection status based on the pharmacokinetic model and pain level. This information is then used to establish a control model for optimization, thereby determining a minimal-cost real-time drug injection plan. The drug injection is then dynamically performed according to the minimal-cost real-time drug injection plan. The real-time predictions of the user's injection status include the predicted blood drug concentration, the predicted effect-sac concentration, and the predicted pain level. Specifically, the extended unscented Kalman filter (UKF) is a nonlinear filter estimation algorithm with high computational accuracy. The pharmacokinetic model primarily focuses on blood drug concentration and effect-sac concentration. Blood drug concentration is related to drug dosage and metabolic capacity. Excessively high blood drug concentrations can lead to drug toxicity or other side effects, while excessively low blood drug concentrations can result in decreased effect-sac concentrations. The effect-sac is the theoretical compartment where drugs exert their effects, and the effect-sac concentration is related to blood drug concentration via the blood-effect-sac transport constant. During drug injections, to balance safety and effectiveness, it is crucial to ensure drug safety while effectively alleviating pain. Therefore, the drug injection module 15 needs to first determine a safe drug injection plan based on the patient's pain condition, and estimate the patient's status information (blood drug concentration, effect chamber concentration and pain level, etc.) in real time, and adjust the medication according to the patient's real-time status.

[0032] In one optional embodiment, the predicted blood drug concentration is determined based on the drug clearance rate constant and the drug distribution volume; the predicted effect compartment concentration is determined based on the predicted blood drug concentration and the blood-effect compartment transport constant; and the predicted pain level is determined based on at least the baseline pain level and the maximum analgesic amplitude that the drug can produce. In a specific example, this can be implemented as follows: first, a state variable is constructed using an extended unscented Kalman filter: ; in, Indicates that the system is at time t The state vector of Indicates time t The pain level (the pain level can be expressed as 1 point, 2 points...10 points); Represents the blood drug concentration predicted by the filter , given by the clearance rate constant Volume of distribution of the drug drive; represents the predicted effect compartment concentration, given by the blood-effect compartment transport constant Connect to The prediction and correction equations are established through the drug metabolism kinetic model. The blood drug concentration dynamic equation is used to describe the accumulation and clearance dynamics of the drug in the blood, as shown below: ; in, Indicates the rate of change of blood drug concentration over time , represents the drug clearance rate constant , Indicates the drug infusion rate , represents the volume of distribution of the drug (L). The dynamic equation for the effect compartment concentration is shown below. The effect compartment is the theoretical compartment where the drug takes effect (such as the central nervous system). This equation quantifies the hysteresis of the drug effect. ; in, Indicates the rate of change of effect compartment concentration over time , It represents the blood-effect compartment transport constant, reflecting the speed at which the drug enters the target site; Represents the concentration gradient between the blood and the effect compartment. The pain grade-drug efficacy relationship equation (also called the Hill equation) is shown below and is used to describe the process of converting blood drug concentration into analgesic effect (VAS decrease); ; in, represents the pain level predicted by the model, It represents the actual basic pain level measured without medication. Indicates the maximum analgesic effect that the drug can produce (usually expressed as a reduction in pain scores). Half-effective concentration , that is, the effect compartment concentration when 50% of the maximum analgesic effect is achieved, Indicates the steepness of the Hill curve (entered by the drug efficacy database or doctor).

[0033] After establishing the prediction and correction equation, the residual can also be observed in real time. , used to update Taking into account individual differences among patients, it is expected that the The error converges to The residuals reflect the difference between the actual pain level and the model-predicted registration and are used to correct the error in the UKF state estimation.

[0034] In one optional embodiment, a control model is established using real-time predictions of the user's injection status to optimize and establish a cost-minimizing real-time drug infusion plan. Specifically, this includes: obtaining drug infusion constraints, which include a maximum infusion rate constraint and a 24-hour maximum infusion dose constraint; establishing an optimization function by performing a weighted sum operation based on the predicted blood drug concentration, predicted pain level, and the drug infusion constraints as cost terms; and establishing a cost-minimizing real-time drug infusion plan with minimizing the weighted sum of these three cost terms as the optimization objective. Specifically, excessive blood drug concentration may put patients at risk of poisoning, while reduced pain levels indicate analgesic efficacy. The drug infusion constraints, which include a maximum infusion rate constraint and a 24-hour maximum infusion dose constraint, can prevent mechanical damage and acute drug overdose. Using these three cost terms, the control module calculates an optimal infusion rate sequence and dynamically adjusts the infusion rate to achieve the target analgesic effect within safety constraints.

[0035] The following is a specific example. After the pharmacokinetic model is used to estimate the real-time prediction information of the user's injection status, it is fed back to the control model for iterative optimization. Specifically, the optimization can be solved at k time steps after time t. The optimization function is as follows: ; The optimization goal of the above function is to minimize the weighted sum of the three costs. The pain control goal set by the doctor (usually a pain scale of ≤ 2 points) represents the predicted pain level at time t+k; represents the square term of the predicted blood drug concentration at time t+k, represents the weight coefficient for weighing analgesia and poisoning; represents the weight coefficient of the soft penalty over-valve position, represents the drug injection rate at time t+k and is subject to the constraints: ; ; in, is the peak flow rate of the syringe pump, The upper limit of the cumulative dose over 24 hours. Constraints can be used to constrain the infusion rate at each moment and the upper limit of the cumulative dose over 24 hours (to prevent chronic poisoning), achieving dual safety guarantees.

[0036] There are three costs in the above objective optimization function. Taking the square of the deviation between the predicted VAS grade score and the target VAS grade score as the cost can ensure the analgesic effect; Taking the square of the blood drug concentration and multiplying it by the weight coefficient can penalize high blood drug concentrations (because high concentrations may lead to poisoning). By minimizing this term, the risk of poisoning can be reduced; Penalties are imposed for any portion of the infusion rate that exceeds the maximum permitted infusion rate. When , this item is 0; when When the square of the excess part is multiplied by the weight coefficient , to prevent the infusion rate from being too fast, protect the safety of the equipment and reduce the risk of side effects.

[0037] In addition, if the armband is in an offline stage, it may not be able to obtain the latest user status information. In order to realize the injection monitoring of the drug injection module 15, the reinforcement learning method can be used in the offline stage, using the state-action-reward samples generated by historical data without the need for actual interaction, thereby enhancing safety. Therefore, in an optional embodiment, when the drug injection module 15 is in an offline stage, the drug injection module 15 is also used to establish a negative reward function based on the degree to which the pain level deviates from the pain control target, whether the historical drug infusion volume exceeds the maximum threshold, and whether respiratory side effects occur; and regulate the drug injection plan according to the negative reward function. Specifically, the offline stage uses the historical trajectory and the control module to drive the simulation generated Samples are used to observe key information such as analgesic effect, real-time drug concentration, safety margin, etc. s is the set of observation vectors, a is the action set (i.e., drug infusion rate), r For reward collection. s The observation vector is as follows: ; in, Represents the action history, that is, the current moment t Previous L The historical drug infusion rates are arranged in reverse chronological order into a vector; Indicates the 24-hour dosage. s The observation vector comprehensively covers pain feedback, action history, drug metabolism status and safe drug dosage. r for: ; in, 、 and Indicates the weight of each item; The triggering conditions for the item are that the pain score deviates from the target value and the analgesia is insufficient ( >2) or excessive sedation ( =0) will be penalized, and the greater the deviation, the greater the penalty; The item is an overdose indicator, indicating the risk of acute poisoning, The trigger condition is or If the threshold is exceeded, it is recorded as 1, otherwise it is recorded as 0; The item is the respiratory side effect indicator, which indicates respiratory depression side effects. By blood oxygen saturation ( ) decrease or respiratory apnea and other respiratory side effects are triggered when respiratory side effects (such as decrease or apnea events) is 1 if the value is set, otherwise it is 0. The reward function of is negative, which means that it is regarded as a cost (minimize the cost). When the reward function takes the maximum value, the cost is minimized.

[0038] The above learning system uses state variables s Full observation, covering all safety-critical variables (pain, drug concentration, total dose), through action a Realize the dual constraints of physical speed limit and cumulative dose, and through rewards r By imposing extreme penalties for safety events (such as overdose and respiratory depression), this design ensures that the reinforcement learning strategy prioritizes patient safety over analgesia, in line with medical ethics. Therefore, offline training further avoids high-risk actions through conservative policy updates, ultimately achieving personalized analgesia optimization within safety boundaries.

[0039] In addition, the drug injection module 15 can be implemented by combining software and hardware. In an optional embodiment, Figure 2 The illustrated medication injection module 15 includes a replaceable medication stick 151 for storing medication to be injected by the user, and an indwelling needle port 152 for injecting medication into the user via an indwelling needle. Specifically, when an injection is required, the medication injection module 15 uses a built-in piezoelectric pump to inject the medication into the user's body through the indwelling needle. The vital sign data collected after the injection is specifically tagged for monitoring by medical personnel. (A copy of the data is also tagged for a period of time after the injection is completed for monitoring by medical personnel.)

[0040] This embodiment proposes a controlled drug injection monitoring armband, which analyzes multimodal vital sign data and realizes automatic analysis of the user's pain condition through multimodal analysis methods, so as to accurately intervene in the user wearing the armband and timely improve the user's vital status. It can also accurately analyze the user's pain condition, thereby accurately setting the required medication plan and dynamically adjusting the medication plan in real time according to the user's status during the injection process, thereby standardizing the use of controlled drugs, preventing drug "misuse" or "abuse", preventing mechanical damage and acute overdose of drugs, and avoiding fatal risks such as respiratory depression, ultimately achieving personalized analgesia optimization within a safe boundary.

[0041] In an optional embodiment, the controlled drug injection monitoring armband of this embodiment further includes: a button 16; the button 16 is used to receive injection instructions; the drug injection module 15 is also used to complete drug injection according to the injection instructions. Specifically, the controlled drug injection monitoring armband of this embodiment can receive injection instructions through the button 16, and can also automatically generate injection instructions according to the usage authority of the drug injection module 15 to automatically inject drugs. By setting the button 16, the user can conveniently express the intention to inject, ensuring the user's true intention expression and facilitating the user's operation. The drug injection module 15 automatically generates injection instructions, and can automatically perform injections for the user when the user is unable to express his or her intention. The combination of the two can not only ensure the user's true intention expression, but also help users who are unable to express their intention to complete automatic injections.

[0042] In one optional embodiment, after the drug injection is completed, the information transmission module 14 also uploads the injection information to the backend processing center. Specifically, the information transmission module 14 feeds back the actual operation information of the drug injection module 15 (such as injection time, injection dosage, and post-injection vital sign data) to the backend processing center, allowing the backend processing center to promptly obtain the user's medical status and prevent repeated injections. The simultaneous upload of vital sign data and injection records to the backend processing center enhances the safety of drug use.

[0043] Example 2 Different from Example 1, in this embodiment, in order to save the computing power of the controlled drug injection monitoring armband in Example 1 and reduce the size of the armband, the analysis data and processing process completed by the analysis module 13 in Example 1 are placed in the cloud processing module for processing. After obtaining the vital signs information, the controlled drug injection monitoring armband sends it to the cloud processing module and receives the pain level determined after processing by the cloud processing module. The contents already explained in the controlled drug injection monitoring armband in Example 1 will not be repeated here, and only a brief introduction to the system architecture will be given. Figure 3As shown, the controlled drug injection monitoring system of this embodiment includes: a controlled drug injection monitoring armband 31 and a cloud processing module 32; the controlled drug injection monitoring armband includes: a monitoring module, an information transmission module and a drug injection module.

[0044] The monitoring module is used to collect vital signs information from users wearing controlled drug injection monitoring armbands. Vital signs information refers to multiple sets of data used to characterize the degree of pain in the user's body. Specifically, when a terminal cancer patient experiences pain, the patient may experience increased blood pressure, increased heart rate, body tremors, and other phenomena. Therefore, the monitoring module needs to monitor the user's vital signs to determine the patient's pain condition. Vital signs information can include at least heart rate, blood oxygen, body temperature, blood pressure, and movement status, which can reflect the user's life status. The monitoring module can include detection units such as a heart rate monitoring unit, a blood oxygen monitoring unit, a body temperature monitoring unit, a blood pressure estimation unit, and a movement status detection unit, which are respectively used to monitor and collect various vital signs of the user.

[0045] The information transmission module is used to send vital sign information to the cloud processing module 32; specifically, the information transmission module in this embodiment can be used to send and receive data with the cloud processing module 32, and the data sending and receiving can be completed using the wired or wireless data interface of the armband.

[0046] The cloud processing module 32 is used to construct a deep learning model, pre-train the deep learning model using vital signs information to obtain a trained deep learning model, and obtain the user's current vital signs information data sequence in real time, and use the trained deep learning model to determine the pain level corresponding to the current vital signs information data sequence based on the user's current vital signs information data sequence, and transmit the pain level to the information transmission module, wherein the current vital signs information data sequence includes vital signs information collected at a preset period; specifically, the cloud processing module 32 is used to analyze the user's pain tolerance. The cloud processing module 32 also needs to pre-train the deep learning model before performing pain analysis, which can improve the accuracy of calculating the pain level based on the vital signs information.

[0047] The information transmission module is also used to receive the pain level and transmit the pain level to the drug injection module; specifically, the information transmission module in this embodiment receives data from the cloud processing module 32 through the external transceiver interface, and transmits the data to the drug injection module through the internal transmission path.

[0048] In an optional embodiment, the information transmission module transmits the pain level to the drug injection module after receiving the pain level; the information transmission module is also used to upload the pain level to the background processing center, and receive the drug injection authorization information returned by the background processing center, and unlock the use rights of the drug injection module according to the drug injection authorization information. Specifically, the background processing center can be a medical institution processing center with medical qualifications. The information transmission module will upload the pain level to the medical institution processing center, and the medical institution processing center will analyze and process it. Optionally, the medical institution processing center can determine whether to grant the user injection permission based on the results of the pain analysis. The user who obtains the injection permission can press the injection button to inject the controlled drug to relieve pain. If it is in a home scenario, it can also be set so that after detecting that the pain reaches a certain level, the device can directly grant injection permission without the need for background processing.

[0049] The medication injection module is used to obtain pain levels. Using an extended unscented Kalman filter system, it estimates real-time predictions of the user's injection status based on a pharmacokinetic model and pain levels. This information is then used to establish a control model for optimization, establishing a minimal-cost real-time medication injection plan. The module then dynamically delivers medication according to this minimal-cost real-time medication injection plan. This real-time prediction includes predicted blood drug concentration, predicted effect-site concentration, and predicted pain level. To balance safety and effectiveness during medication injections, ensuring medication safety while effectively alleviating pain is crucial. Therefore, the medication injection module must first determine a safe medication injection plan based on the patient's pain status and then estimate patient status information (blood drug concentration, effect-site concentration, and pain level) in real time, adjusting medication accordingly.

[0050] In an optional embodiment, the analysis function of the above-mentioned drug injection module can also be executed by the cloud processing module 32, and the cloud processing module 32 performs the corresponding calculation function, while the drug injection module only completes the physical process of drug injection according to the injection plan, so as to further save the computing power of the controlled drug injection monitoring armband 31 of this embodiment and reduce the size of the armband.

[0051] This embodiment proposes a controlled drug injection monitoring system that analyzes multimodal vital sign data. This multimodal analysis method enables automatic analysis of a user's pain status, enabling precise intervention for users wearing the armband and timely improvement of their vital status. The system also accurately analyzes the user's pain status, allowing for precise setting of the required medication regimen. This system dynamically adjusts the regimen in real time based on the user's condition during the injection process, thereby standardizing the use of controlled drugs, preventing drug misuse or abuse, mechanical injury and acute overdose, and potentially fatal risks such as respiratory depression. Ultimately, this system achieves personalized analgesia optimization within safety boundaries. Furthermore, the controlled drug injection monitoring system of this embodiment utilizes distributed computing via the armband and the cloud, which improves computing efficiency, reduces armband costs, reduces armband size, and increases armband lifespan.

[0052] In an optional embodiment, the cloud processing module 32 uses a trained deep learning model to determine the pain level corresponding to the current vital sign information data sequence based on the user's current vital sign information data sequence. Specifically, the process includes: grouping and aligning the current vital sign information data sequence according to the time dimension, and constructing a first multimodal vector; inputting the first multimodal vector into the trained deep learning model for operation to obtain first calculated data; comparing the first calculated data with preset reference data to determine the pain level corresponding to the current vital sign information data sequence. Specifically, the first calculated data is the result obtained by operating the current vital sign information data sequence through the trained deep learning model; the preset reference data is a pain level reference data defined in relevant medical science, which can be implemented in the form of a stepped threshold. By comparing the first calculated data with the values ​​of each level of the stepped threshold, the closest pain level is determined as the final result.

[0053] In an optional embodiment, the deep learning model comprises a Transformer encoder and a multi-layer perceptron; the operation of the first multi-modal vector into the trained deep learning model comprises: using the Transformer encoder to operate on the first multi-modal vector using a multi-head attention mechanism, and performing residual connection and normalization calculation to obtain the output of the last layer; and performing reconstruction operation on the output of the last layer through the multi-layer perceptron. Specifically, the Transformer encoder can have multiple layers, and the vector data is subjected to residual connection and normalization calculation when passing through each layer. The output of the last layer refers to the calculation result output by the first multi-modal vector data after passing through the last layer of the Transformer encoder. Residual connection can make it easier to train deep networks; normalization makes the data distribution more stable, accelerating the training. The multi-head attention mechanism of the Transformer encoder in this embodiment allows the model to focus on information at different positions at the same time, while the MLP can integrate and transform these information, improving the generalization ability of the model. Through the reasonable parameter design of the MLP, the performance can be maintained while reducing the computational overhead.

[0054] In an optional embodiment, the cloud processing module 32 uses the vital sign information to pre-train the deep learning model to obtain a trained deep learning model, specifically including: the cloud processing module 32 receives the vital sign information sample sequence obtained by the monitoring module through the information transmission module to collect the vital sign information of the user; the vital sign information sample sequence is grouped and aligned according to the time dimension, and a second multimodal vector is constructed; the second multimodal vector is operated on by the deep learning model to obtain a second reconstruction vector; a third reconstruction vector and a fourth reconstruction vector are obtained, the third reconstruction vector is obtained by operating the third multimodal vector by using the deep learning model, and the fourth reconstruction vector is obtained by operating the fourth multimodal vector by using the deep learning model, wherein the third multimodal vector is constructed based on the vital sign information of the user in a previous non-pain state, and the fourth multimodal vector is constructed based on the vital sign information of the user in a previous pain state; a loss function is constructed based on the second reconstruction vector, the third reconstruction vector and the fourth reconstruction vector; the parameters of the deep learning model are updated through iterative training until the loss function converges to obtain a trained deep learning model. Specifically, the vital sign information sample sequence can include the user's actual vital sign information over a specific time period. By training the deep learning model with the user's previous non-painful vital sign information and the user's previous painful vital sign information, better model parameters can be provided for the calculation of subsequent specific tasks, thereby improving model performance, reducing data requirements, accelerating training, and enhancing generalization capabilities. The specific steps and methods for cloud processing module 32 to complete pre-training of the deep learning model and determine the user's current pain level based on the current vital sign information data sequence can be found in the specific steps of the analysis module in Example 1 and will not be repeated here.

[0055] In an optional embodiment, the predicted blood drug concentration is determined based on the drug clearance rate constant and the drug distribution volume; the predicted effect compartment concentration is determined based on the predicted blood drug concentration and the blood-effect compartment transport constant; and the predicted pain level is determined based on at least the baseline pain level and the maximum analgesic amplitude that the drug can produce.

[0056] In an optional embodiment, a control model is established using real-time prediction information of the user's injection status for optimization to establish a real-time plan for drug injection with minimum cost, specifically including: obtaining a drug injection constraint formula, the drug injection constraint formula including: a maximum infusion rate constraint and a 24-hour maximum infusion dose constraint; establishing an optimization function by performing a weighted sum operation based on the predicted blood drug concentration, the predicted pain level, and the drug injection constraint formula as cost terms; establishing a real-time plan for drug injection with minimum cost with the optimization goal of minimizing the weighted sum of the three cost terms.

[0057] In an optional embodiment, the vital sign information includes at least: heart rate, blood oxygen level, body temperature, blood pressure, and exercise status. By collecting multimodal vital sign data, the user's pain condition can be more accurately determined, preventing drug misuse and abuse.

[0058] In an optional embodiment, the method of this embodiment further includes: receiving an injection instruction via a button or automatically generating an injection instruction based on the access rights of the drug injection module; and the drug injection module completing the drug injection according to the injection instruction. By setting a button, the user can conveniently express their injection intention, ensuring that the user's true intention is expressed and facilitating user operation. Automatically generating an injection instruction by the drug injection module can automatically perform an injection for the user when the user is unable to express their intention. The combination of these two methods not only ensures that the user's true intention is expressed, but also helps users who are unable to express their intention complete the automatic injection.

[0059] In an optional embodiment, the method of this embodiment further includes uploading injection information to a backend processing center after the drug injection is completed. Specifically, actual operation information of the drug injection module (such as injection time, injection dosage, and post-injection vital sign data) is fed back to the backend processing center so that the backend processing center can promptly obtain the user's medical status and prevent repeated injections. The simultaneous uploading of vital sign data and injection records to the backend processing center enhances the safety of drug use.

[0060] In the description of the embodiments of the present invention, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "vertical", "horizontal", "center", "top", "bottom", "top", "bottom", "inside", "outside", "inner side", "outer side" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. Among them, "inside" refers to an internal or enclosed area or space. "Periphery" refers to the area surrounding a specific component or specific area.

[0061] In the description of the embodiments of the present invention, the terms "first," "second," "third," and "fourth" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly specifying the number of the technical features indicated. Therefore, a feature specified as "first," "second," "third," or "fourth" may explicitly or implicitly include one or more of the features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.

[0062] In the description of the embodiments of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "installed," "connected," "connected," and "assembled" should be understood in a broad sense. For example, they may refer to fixed connections, detachable connections, or integral connections; they may refer to direct connections, indirect connections through an intermediate medium, or internal connections between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention in specific circumstances.

[0063] In the description of the embodiments of the present invention, specific features, structures, materials or characteristics may be combined in an appropriate manner in any one or more embodiments or examples.

[0064] In describing the embodiments of the present invention, it should be understood that "-" and "~" represent a range between two values, and the range includes the endpoints. For example, "AB" represents a range greater than or equal to A and less than or equal to B. "A~B" represents a range greater than or equal to A and less than or equal to B.

[0065] In describing the embodiments of the present invention, the term "and / or" is used herein to describe a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent three situations: A exists alone, A and B exist simultaneously, and B exists alone. Furthermore, the character " / " is generally used herein to indicate that the associated objects are in an "or" relationship.

[0066] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A controlled drug injection monitoring system, characterized in that: include: Controlled drug injection monitoring armband and cloud processing module; The controlled drug injection monitoring armband includes: a monitoring module, an information transmission module and a drug injection module; The monitoring module is used to collect vital sign information of the user wearing the controlled drug injection monitoring armband, wherein the vital sign information refers to multiple sets of data used to represent the degree of pain in the user's body; The information transmission module is used to send the vital sign information to the cloud processing module; The cloud processing module is used to construct a deep learning model, pre-train the deep learning model using the vital sign information to obtain a trained deep learning model, and obtain the user's current vital sign information data sequence in real time, using the trained deep learning model to determine the pain level corresponding to the current vital sign information data sequence based on the user's current vital sign information data sequence, and transmit the pain level to the information transmission module, wherein the current vital sign information data sequence includes the vital sign information collected at a preset period; The information transmission module is further configured to receive the pain level and transmit the pain level to the drug injection module; The drug injection module is used to obtain the pain level, estimate the real-time prediction information of the user's injection status based on the drug metabolism kinetics model and the pain level through an extended unscented Kalman filter system, use the real-time prediction information of the user's injection status to establish a control model for optimization to establish a real-time plan for drug injection with minimum cost, and dynamically perform drug injection according to the real-time plan for drug injection with minimum cost, wherein the real-time prediction information of the user's injection status includes: predicted blood drug concentration, predicted effect chamber concentration and predicted pain level.

2. The controlled drug injection monitoring system according to claim 1, characterized in that: The determining, using the trained deep learning model according to the user's current vital sign information data sequence, a pain level corresponding to the current vital sign information data sequence specifically includes: Grouping and aligning the current vital sign information data sequence according to the time dimension, and constructing a first multimodal vector; The first multimodal vector is sent to the trained deep learning model for calculation to obtain first calculation data, and the first calculation data is compared with preset reference data to determine the pain level corresponding to the current vital sign information data sequence.

3. The controlled drug injection monitoring system according to claim 2, characterized in that: The deep learning model includes a Transformer encoder and a multi-layer perceptron; The step of sending the first multimodal vector to the trained deep learning model for calculation includes: Using the Transformer encoder to operate on the first multimodal vector using a multi-head attention mechanism, and performing residual connection and normalization calculations to obtain the final layer output; The last layer output is reconstructed by the multi-layer perceptron.

4. The controlled drug injection monitoring system according to any one of claims 1 to 3, characterized in that: Pre-training the deep learning model using the vital sign information to obtain a trained deep learning model specifically includes: receiving, by the information transmission module, a sequence of vital sign information samples acquired by the monitoring module from the vital sign information of the user; Grouping and aligning the vital sign information sample sequences according to the time dimension, and constructing a second multimodal vector; Using the deep learning model to operate on the second multimodal vector to obtain a second reconstructed vector; Obtaining a third reconstruction vector and a fourth reconstruction vector, wherein the third reconstruction vector is obtained by applying the deep learning model to a third multimodal vector, and the fourth reconstruction vector is obtained by applying the deep learning model to a fourth multimodal vector, wherein the third multimodal vector is constructed based on vital sign information of the user in a previous non-pain state, and the fourth multimodal vector is constructed based on vital sign information of the user in a previous pain state; constructing a loss function according to the second reconstruction vector, the third reconstruction vector, and the fourth reconstruction vector; The parameters of the deep learning model are updated through iterative training until the loss function converges, thereby obtaining the trained deep learning model.

5. The controlled drug injection monitoring system according to any one of claims 1 to 3, characterized in that: The information transmission module receives the pain level and transmits the pain level to the drug injection module; The information transmission module is further used to upload the pain level to the background processing center, and receive the drug injection authorization information returned by the background processing center, and unlock the use rights of the drug injection module according to the drug injection authorization information.

6. The controlled drug injection monitoring system according to any one of claims 1 to 3, characterized in that: The predicted blood drug concentration is determined based on the drug clearance rate constant and the drug distribution volume; The predicted effect compartment concentration is determined based on the predicted blood drug concentration and the blood-effect compartment transport constant; The predicted pain level is determined based on at least the basic pain level and the maximum analgesic amplitude that the drug can produce.

7. The controlled drug injection monitoring system according to any one of claims 1 to 3, characterized in that: The method of using the real-time prediction information of the user's injection status to establish a control model for optimization to establish a real-time plan for drug injection with the minimum cost specifically includes: Obtaining a drug injection constraint formula, wherein the drug injection constraint formula includes: a maximum infusion rate constraint and a 24-hour maximum infusion dose constraint; Establishing an optimization function by performing a weighted sum operation based on the predicted blood drug concentration, the predicted pain level, and the drug injection constraint condition as cost terms; The real-time plan for minimizing the cost of drug injection is established with minimizing the weighted sum of the three cost items as the optimization goal.

8. The controlled drug injection monitoring system according to any one of claims 1 to 3, characterized in that: The vital signs information includes at least: heart rate value, blood oxygen value, body temperature value, blood pressure value and exercise status.

9. A controlled drug injection monitoring armband, characterized in that: include: Armband body, monitoring module, analysis module, information transmission module and drug injection module; The monitoring module is used to collect vital sign information of a user wearing the controlled drug injection monitoring armband through the armband body, wherein the vital sign information refers to multiple sets of data used to represent the degree of pain in the user's body; The analysis module is configured to construct a deep learning model, pre-train the deep learning model using the vital sign information to obtain a trained deep learning model, and acquire a current vital sign information data sequence of the user in real time, determine a pain level corresponding to the current vital sign information data sequence based on the current vital sign information data sequence of the user using the trained deep learning model, and transmit the pain level to the information transmission module, wherein the current vital sign information data sequence includes the vital sign information collected at a preset period; The information transmission module is further configured to receive the pain level, upload the pain level to a background processing center, receive drug injection authorization information returned by the background processing center, and unlock the use rights of the drug injection module according to the drug injection authorization information; The drug injection module is used to obtain the pain level, estimate the real-time prediction information of the user's injection status based on the drug metabolism kinetics model and the pain level through an extended unscented Kalman filter system, use the real-time prediction information of the user's injection status to establish a control model for optimization to establish a real-time plan for drug injection with minimum cost, and dynamically perform drug injection according to the real-time plan for drug injection with minimum cost, wherein the real-time prediction information of the user's injection status includes: predicted blood drug concentration, predicted effect chamber concentration and predicted pain level.

10. The controlled drug injection monitoring armband according to claim 9, characterized in that: The drug injection module also includes: A replaceable medicine stick, used to store medicine that the user needs to inject; The indwelling needle interface is used to inject medicine into the user through the indwelling needle.

Images