Vasoactive drug injection system based on AI and invasive blood pressure monitoring

Abstract

The present invention discloses a vasoactive drug injection system based on AI and invasive blood pressure monitoring, comprising: a data acquisition module for collecting physiological data; a core computing module for constructing a hemodynamic prediction model and a vasoactive drug dosage optimization model to obtain a prediction curve and vasoactive drug infusion rate adjustment; an injection control module for receiving the vasoactive drug infusion rate adjustment and generating a flow rate instruction; a flow sensor for monitoring the flow rate of the vasoactive drug injected by the injection control module and generating a control signal for the injection drive unit in the injection control module; and a human-computer interaction module for displaying necessary parameters. Thus, by real-time acquisition of arterial blood pressure data and electrocardiogram signals, combined with the cumulative dose of the vasoactive drug, a more accurate and personalized drug infusion rate adjustment can be provided, ensuring the patient's hemodynamic stability and reducing delays and errors caused by manual adjustments.

Description

Technical Field

[0001] The present invention relates to the field of medical devices, and in particular to a vasoactive drug injection system based on AI and invasive blood pressure monitoring. Background Art

[0002] In modern intensive care and anesthesia, precise administration of vasoactive drugs is crucial for maintaining the patient's hemodynamic stability. Traditional vasoactive drug injection methods rely primarily on the experience and manual adjustments of medical staff. This method suffers from delayed response and insufficient adjustment accuracy, especially in the face of rapidly changing clinical conditions. It is difficult to adjust the drug dosage in a timely and accurate manner to adapt to the patient's rapidly changing physiological state. In addition, manual operations are susceptible to human factors, such as fatigue or misjudgment, which increases the risk and uncertainty of treatment.

[0003] In recent years, with the development of artificial intelligence (AI) technology and the improvement of the intelligence level of medical equipment, AI-based automation systems have begun to be applied in the medical field, especially in drug management and infusion control, showing great potential. By combining advanced sensing technology and machine learning algorithms, real-time monitoring and analysis of patients' vital signs data can be achieved, and the drug infusion rate can be automatically adjusted accordingly, thereby providing more personalized and accurate treatment plans. However, most smart infusion pumps on the market can only deliver drugs at a fixed rate according to preset programs and lack dynamic adjustment capabilities. Some products that attempt to introduce AI-assisted decision-making often have low prediction accuracy due to insufficient model training or failure to effectively integrate multi-source heterogeneous data. Summary of the Invention

[0004] The present invention aims to solve the technical problems in the above-mentioned technologies at least to some extent.

[0005] To this end, the present invention discloses a vasoactive drug injection system based on AI and invasive blood pressure monitoring, comprising:

[0006] The data acquisition module is used to collect the systolic pressure P of the arterial blood pressure at a frequency not less than the preset frequency through the invasive blood pressure sensor. s , diastolic blood pressure P d , mean arterial pressure and continuous blood pressure waveform function B(t), synchronously obtain electrocardiogram signal E(t) with multi-lead electrocardiograph, record cumulative dose D(t) of vasoactive drugs, and generate the mean arterial pressure P in time series form. m (t);

[0007] The core computing module is used to construct the hemodynamic prediction model M1 and the vasoactive drug dosage optimization model M2, where:

[0008] The mean arterial pressure P in time series form is input to the hemodynamic prediction model M1. m (t) to obtain the prediction curve for the future preset time

[0009] The mean arterial pressure P in time series form is input to the vasoactive drug dosage optimization model M2. m (t), prediction deviation and the patient's basic physiological parameter vector To obtain the adjustment amount of vasoactive drug infusion rate at the next moment

[0010] An injection control module is used to receive the vasoactive drug infusion rate adjustment amount ΔR(t) at the next moment and generate a flow rate instruction R(t+1)=R(t)+ΔR(t);

[0011] Flow sensor, used to monitor the flow rate R of the vasoactive drug injected by the injection control module a (t), and generating a control signal for the injection drive unit in the injection control module in,

[0012] Error signal e(t) = R(t) - R a (t);

[0013] K p ,K i ,K d To tune the parameters;

[0014] The human-computer interaction module is used to display the arterial blood pressure waveform and vasoactive drug injection rate curve, as well as to pop up an alarm and voice prompt medical staff when the predicted blood pressure deviates from the preset safety range.

[0015] The vasoactive drug injection system based on AI and invasive blood pressure monitoring disclosed in the present invention can provide more accurate and personalized drug infusion rate adjustment by collecting arterial blood pressure data and electrocardiogram signals in real time and combining the cumulative dose of vasoactive drugs, thereby ensuring the patient's hemodynamic stability and reducing delays and errors caused by manual adjustments.

[0016] In addition, the vasoactive drug injection system based on AI and invasive blood pressure monitoring disclosed in the present invention may also have the following additional technical features:

[0017] In one embodiment of the present invention, in the data acquisition module, the preset frequency is not less than 200 Hz.

[0018] In one embodiment of the present invention, in the core operation module, the future preset time ranges from 5 to 8 minutes.

[0019] In one embodiment of the present invention, in the core operation module, the patient's basic physiological parameter vector include:

[0020] The patient's age, gender, weight, height, basal heart rate, basal blood pressure, and whether he or she has cardiovascular disease.

[0021] In one embodiment of the present invention, in the human-computer interaction module, the preset safety range is 5 to 10 mmHg.

[0022] In one embodiment of the present invention, in the human-computer interaction module, the data acquisition module is calibrated according to a standard pressure gas cylinder through an operation interface, and the initial flow rate R(0) of the injection control module is set.

[0023] In one embodiment of the present invention, it further comprises:

[0024] The emergency stop module is used to be pressed when the predicted blood pressure deviates from the preset safety range value to stop the injection control module from injecting the vasoactive drug.

[0025] In one embodiment of the present invention, the hemodynamic prediction model M1 updates the rules once every 5 to 10 seconds. After each update of the hemodynamic prediction model M1, the vasoactive drug dosage optimization model M2 immediately generates the vasoactive drug infusion rate adjustment amount at the next moment.

[0026] Additional contents and advantages of the present invention will be given in the following description or can be understood through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The technical solutions and beneficial effects of the present invention will become apparent and easily understood from the following contents in conjunction with the accompanying drawings, in which:

[0028] Figure 1 This is a system block diagram of the vasoactive drug injection system based on AI and invasive blood pressure monitoring of the present invention;

[0029] Figure 2 This is a workflow diagram of the vasoactive drug injection system based on AI and invasive blood pressure monitoring of the present invention. DETAILED DESCRIPTION

[0030] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0031] The vasoactive drug injection system based on AI and invasive blood pressure monitoring disclosed in the present invention will be described below with reference to the accompanying drawings.

[0032] like Figure 1 and Figure 2 As shown, a vasoactive drug injection system 100 based on AI and invasive blood pressure monitoring includes:

[0033] The data acquisition module 101 is used to collect the systolic pressure P of the arterial blood pressure at a frequency not less than a preset frequency through an invasive blood pressure sensor. s , diastolic blood pressure P d , mean arterial pressure The continuous blood pressure waveform function B(t) and the multi-lead electrocardiograph are used to synchronously obtain the ECG signal E(t), record the cumulative dose of vasoactive drugs D(t), and generate the mean arterial pressure P in the form of a time series. m (t);

[0034] It should be noted that the preset frequency is not less than 200Hz;

[0035] The core computing module 102 is used to construct a hemodynamic prediction model M1 and a vasoactive drug dosage optimization model M2, wherein:

[0036] The mean arterial pressure P in the form of a time series is input to the hemodynamic prediction model M1. m (t) to obtain the prediction curve for the future preset time

[0037] The mean arterial pressure P in the form of a time series is input to the vasoactive drug dosage optimization model M2. m (t), prediction deviation and the patient's basic physiological parameter vector To obtain the adjustment amount of vasoactive drug infusion rate at the next moment

[0038] It should be noted that the future preset time ranges from 5 to 8 minutes;

[0039] It should also be noted that the patient's basic physiological parameter vector include:

[0040] The patient's age, gender, weight, height, baseline heart rate, baseline blood pressure, and whether he or she has cardiovascular disease;

[0041] Moreover, the hemodynamic prediction model M1 updates its rules once every 5 to 10 seconds. After each update of the hemodynamic prediction model M1, the vasoactive drug dosage optimization model M2 immediately generates the vasoactive drug infusion rate adjustment amount for the next moment.

[0042] The injection control module 103 is configured to receive the vasoactive drug infusion rate adjustment amount ΔR(t) at the next moment and generate a flow rate instruction R(t+1)=R(t)+ΔR(t);

[0043] It should be noted that the preset safety range is 5 to 10 mmHg;

[0044] Flow sensor 104 is used to monitor the flow rate R of the vasoactive drug injected by the injection control module. a (t), and generates a control signal for the injection drive unit in the injection control module in,

[0045] Error signal e(t) = R(t) - R a (t);

[0046] K p ,K i ,K d To tune the parameters;

[0047] Human-computer interaction module 105, used to display the arterial blood pressure waveform and vasoactive drug injection rate curve, and to pop up an alarm and voice prompt medical staff when the predicted blood pressure deviates from the preset safe range;

[0048] It should be noted that the data acquisition module is calibrated according to the standard pressure cylinder through the operation interface, and the initial flow rate R(0) of the injection control module is set;

[0049] In addition, the vasoactive drug injection system based on AI and invasive blood pressure monitoring also includes:

[0050] The emergency stop module is used to be pressed when the predicted blood pressure deviates from the preset safety range value to stop the injection of vasoactive drugs by the injection control module.

[0051] In the present invention, medical personnel calibrate the data acquisition module according to the standard pressure cylinder through the operation interface of the human-computer interaction module to ensure that the collected blood pressure data is accurate and reliable, and at the same time set the initial flow rate R(0)=5ml / h of the injection control module to start injecting vasoactive drugs for the patient;

[0052] In the data acquisition module 101, the invasive blood pressure sensor collects the systolic pressure P of the arterial blood pressure at a frequency of 200 Hz. s , diastolic blood pressure Pd , mean arterial pressure The continuous blood pressure waveform function B(t) and the multi-lead electrocardiograph are used to synchronously obtain the ECG signal E(t), record the cumulative dose of vasoactive drugs D(t), and generate the mean arterial pressure P in the form of a time series. m (t);

[0053] The hemodynamic prediction model M1 in the core computing module 102 receives the mean arterial pressure P in the form of a time series. m (t), as well as the ECG signals E(t-30),...,E(t) and the cumulative drug doses d(t-30),...,D(t) in the past 30 seconds;

[0054] After the hemodynamic prediction model M1 completes the update at time t1, the vasoactive drug dosage optimization model M2 immediately obtains the current mean arterial pressure P m (t1), prediction deviation And the patient's basic physiological parameter vector Examples include: patients aged 65 years, male, weighing 70 kg, 175 cm tall, with a basal heart rate of 70 times / min, a basal blood pressure of 120 / 80 mmHg, and cardiovascular disease;

[0055] The vasoactive drug dosage optimization model M2 calculates the vasoactive drug infusion rate adjustment amount ΔR(t1) = 0.5 ml / h at the next moment (t1+1) based on these data;

[0056] After receiving ΔR(t1), the injection control module 103 generates a new flow rate instruction R(t1+1)=R(t1)+ΔR(t1)=5+0.5=5.5 ml / h, and controls the injection drive unit to inject the vasoactive drug according to the new flow rate;

[0057] The flow sensor 104 monitors the actual flow rate R of the vasoactive drug injected by the injection control module in real time. a (t1), assuming the actual flow rate is 5.4 ml / h, then the error signal e(t1) = R(t1)-R a (t1) = 5.5-5.4 = 0.1 ml / h, the flow sensor is based on the error signal e(t1) and the tuning parameter K p ,K i ,K d (Assuming K p =1,K i =0.1,K d =0.05) generates the control signal u(t1) of the injection drive unit to fine-tune the injection flow rate to reduce the error and make the actual flow rate closer to the command flow rate;

[0058] The human-computer interaction module displays the arterial blood pressure waveform and vasoactive drug injection rate curve in real time, allowing medical staff to intuitively observe the patient's blood pressure changes and drug infusion status.

[0059] Assume that during the subsequent monitoring process, the blood pressure is predicted to deviate from the preset safety range by 5 to 10 mmHg. For example, the mean arterial pressure is predicted to be lower than 55 mmHg. The human-computer interaction module will immediately pop up an alarm and provide voice prompts to medical staff. At the same time, the emergency stop module can be pressed by medical staff in an emergency to stop the injection of vasoactive drugs by the injection control module to ensure patient safety.

[0060] Specifically, regarding the two models used in the embodiments of the present invention,

[0061] The hemodynamic prediction model M1 is built based on the long short-term memory network (LSTM) and aims to predict the patient's future mean arterial pressure (MAP) change trend based on historical data.

[0062] For the collected arterial blood pressure data (systolic pressure P s , diastolic blood pressure P d , mean arterial pressure The continuous blood pressure waveform function B(t) and the multi-lead electrocardiograph are used to synchronously obtain the ECG signal E(t), record the cumulative dose of vasoactive drugs D(t), and generate the mean arterial pressure P in the form of a time series. m (t)), first, data cleaning is performed to remove outliers and noise interference. For example, if the blood pressure value at a certain moment is significantly deviated from the normal physiological range and lacks continuity with the previous and subsequent data, it is considered an outlier and corrected or eliminated. Then, the cleaned data is normalized so that data with different features have similar dimensions to facilitate model training.

[0063] The LSTM network is composed of multiple stacked LSTM layers, each containing multiple memory cells. By setting the appropriate number of hidden layers and memory cells, the model can learn long-term dependencies in the data. For example, after experimental comparison, it was chosen to set up three hidden layers, each containing 64 memory cells. This can effectively capture the time series characteristics of data such as blood pressure and electrocardiogram signals while ensuring computational efficiency.

[0064] After the LSTM layer, a fully connected layer is connected to map the output of the LSTM layer to the predicted MAP value. The number of neurons in the fully connected layer is determined by the dimension of the predicted target MAP, which is 1 here.

[0065] The preprocessed time series data (including the mean arterial pressure P in the past 30 seconds) m (t-30),…,P m(t), ECG signals E(t-30),…,E(t) and cumulative drug doses D(t-30),…,D(t) are divided into training set, validation set and test set, which can be divided according to the ratio of 70%, 15%, and 15%;

[0066] The model is trained using the training set, and the weights and biases of the model are adjusted through the back propagation algorithm to minimize the mean square error (MSE) between the predicted value and the actual value. The calculation formula of MSE is: Where n is the number of samples, y i is the actual value, is the predicted value;

[0067] During training, early stopping is used to prevent overfitting. That is, training is stopped when the loss function on the validation set stops decreasing. At the same time, a learning rate decay strategy is used to gradually reduce the learning rate as the number of training rounds increases to improve the convergence of the model. For example, the initial learning rate is set to 0.001, and after every 10 epochs, the learning rate decays to 0.9 times the original value.

[0068] Use the test set to evaluate the trained model and calculate evaluation indicators such as root mean square error (RMSE) and mean absolute error (MAE) to measure the prediction performance of the model. The calculation formula of RMSE is: The calculation formula of MAE is,

[0069] Vasoactive drug dosage optimization model M2, built based on the deep Q-network (DQN) in reinforcement learning, aims to determine the optimal vasoactive drug infusion rate adjustment based on the current hemodynamic state and the patient's basic physiological parameters;

[0070] The state space includes the mean arterial pressure P at the current moment m (t), prediction deviation output by hemodynamic prediction model M1 And the patient's basic physiological parameter vector (Including information such as the patient's age, gender, weight, height, basal heart rate, basal blood pressure, and whether or not they have cardiovascular disease). These state variables can fully reflect the patient's current physiological state and hemodynamic change trends, providing a basis for drug dosage adjustments;

[0071] Discretize continuous state variables (such as blood pressure and heart rate) into several intervals to facilitate DQN model processing. For example, divide mean arterial pressure into intervals of 5 mmHg and heart rate into intervals of 10 beats per minute. For categorical variables (such as gender and cardiovascular disease status), use one-hot encoding.

[0072] The action space is the adjustment amount ΔR(t) of the vasoactive drug infusion rate. Considering the safety and effectiveness in actual clinical applications, the adjustment amount is set to a finite set of discrete values, for example, {-0.5ml / h, -0.25ml / h, 0ml / h, 0.25ml / h, 0.5ml / h}, indicating that the infusion rate can be reduced, maintained, or increased by a certain amount.

[0073] The reward function is designed to guide the model to select actions that can bring the patient's hemodynamic state closer to the target range (i.e., maintain stability). If the predicted mean arterial pressure at the next moment is closer to the preset target range (for example, the normal mean arterial pressure range is 70-105 mmHg), a positive reward is given. Conversely, if the predicted blood pressure deviates further from the target range, a negative reward is given.

[0074] The specific form of the reward function can be adjusted and optimized according to the actual situation. For example, the reward value can be calculated using the following formula in, is the mean arterial pressure at the next moment predicted based on the current action, P target is the target mean arterial pressure (the middle value of the target range can be taken, such as 87.5 mmHg);

[0075] The DQN model learns the optimal strategy by interacting with the environment (here the patient's hemodynamic system). At each time step t, the model learns the optimal strategy based on the current state s. t Select an action t That is, the drug infusion rate adjustment amount), after executing this action, the environment will feedback the next state s t+1 and the corresponding reward r t ;

[0076] The experience tuple (s t ,a t ,r t ,s t+1 ) is stored in the experience replay buffer. During the training process, a batch of experience data is randomly sampled from the buffer to update the Q-value function of the model. The Q-value function represents the expected cumulative reward for taking an action in a given state;

[0077] A deep neural network is used to approximate the Q-value function. The input of the network is the state vector, and the output is the Q-value estimate of each action. The network parameters are updated by minimizing the mean square error between the target Q-value and the predicted Q-value. The target Q-value is calculated as follows: Among them, γ is a discount factor, which is used to weigh the importance of future rewards and current rewards, and is usually between 0.9 and 0.99.

[0078] In summary, the vasoactive drug injection system based on AI and invasive blood pressure monitoring disclosed in the present invention can provide more accurate and personalized drug infusion rate adjustment by real-time collection of arterial blood pressure data and electrocardiogram signals, combined with the cumulative dose of vasoactive drugs, to ensure the patient's hemodynamic stability and reduce the delays and errors caused by manual adjustments.

[0079] In the description of the present invention, it should be understood that the terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential" and the like to 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 should not be understood as limiting the present invention.

[0080] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature identified as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present invention, "plurality" means two or more, unless otherwise specifically defined.

[0081] In the present invention, unless otherwise expressly specified or limited, the terms "mounted," "connected," "connect," "fixed," etc. should be understood broadly. For example, they may refer to fixed connection, detachable connection, or integration; mechanical connection or electrical connection; direct connection or indirect connection through an intermediate medium; internal communication between two components or interaction between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on specific circumstances.

[0082] In the present invention, unless otherwise expressly specified or limited, when a first feature is "above" or "below" a second feature, it may mean that the first and second features are in direct contact, or that the first and second features are in indirect contact through an intermediary. Furthermore, when a first feature is "above," "above," or "above" a second feature, it may mean that the first feature is directly above or diagonally above the second feature, or simply means that the first feature is at a higher level than the second feature. When a first feature is "below," "below," or "below" a second feature, it may mean that the first feature is directly below or diagonally below the second feature, or simply means that the first feature is at a lower level than the second feature.

[0083] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.

[0084] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A vasoactive drug injection system based on AI and invasive blood pressure monitoring, characterized in that: include: The data acquisition module is used to collect the systolic pressure P of the arterial blood pressure at a frequency not less than the preset frequency through the invasive blood pressure sensor. s , diastolic blood pressure P d , mean arterial pressure and continuous blood pressure waveform function B(t), synchronously obtain electrocardiogram signal E(t) with multi-lead electrocardiograph, record cumulative dose D(t) of vasoactive drugs, and generate the mean arterial pressure P in time series form. m (t); The core computing module is used to construct the hemodynamic prediction model M1 and the vasoactive drug dosage optimization model M2, where: The mean arterial pressure P in time series form is input to the hemodynamic prediction model M1. m (t) to obtain the prediction curve for the future preset time The mean arterial pressure P in time series form is input to the vasoactive drug dosage optimization model M2. m (t), prediction deviation and the patient's basic physiological parameter vector To obtain the adjustment amount of vasoactive drug infusion rate at the next moment An injection control module is used to receive the vasoactive drug infusion rate adjustment amount ΔR(t) at the next moment and generate a flow rate instruction R(t+1)=R(t)+ΔR(t); Flow sensor, used to monitor the flow rate R of the vasoactive drug injected by the injection control module a (t), and generating a control signal for the injection drive unit in the injection control module in, Error signal e(t) = R(t) - R a (t); K p ,K i ,K d To tune the parameters; The human-computer interaction module is used to display the arterial blood pressure waveform and vasoactive drug injection rate curve, as well as to pop up an alarm and voice prompt medical staff when the predicted blood pressure deviates from the preset safety range.

2. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: In the data acquisition module, the preset frequency is not less than 200 Hz.

3. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: In the core operation module, the future preset time ranges from 5 to 8 minutes.

4. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: In the core operation module, the patient's basic physiological parameter vector include: The patient's age, gender, weight, height, basal heart rate, basal blood pressure, and whether he or she has cardiovascular disease.

5. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: In the human-computer interaction module, the preset safety range is 5 to 10 mmHg.

6. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: In the human-computer interaction module, the data acquisition module is calibrated according to the standard pressure cylinder through the operation interface, and the initial flow rate R(0) of the injection control module is set.

7. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: Also includes: The emergency stop module is used to be pressed when the predicted blood pressure deviates from the preset safety range value to stop the injection control module from injecting the vasoactive drug.

8. The vasoactive drug injection system based on AI and invasive blood pressure monitoring according to claim 1, characterized in that: The hemodynamic prediction model M1 updates the rules once every 5 to 10 seconds. After each update of the hemodynamic prediction model M1, the vasoactive drug dosage optimization model M2 immediately generates the vasoactive drug infusion rate adjustment amount at the next moment.

Images