Non-invasive blood pressure monitoring method and device based on deep learning algorithm

Abstract

The invention discloses a non-invasive blood pressure monitoring method and device based on a deep learning algorithm, the device comprises a sensor module, a control module and a display module, the sensor module is electrically connected with the control module, a digital filter is arranged between the sensor module and the control module, and the display module is electrically connected with the control module. And the control module is electrically connected with the display module. The problems that along with continuous development of artificial intelligence, traditional blood pressure monitoring equipment can only measure the blood pressure of a patient, the function is single, and the requirement for intelligence of modern medical equipment cannot be met are solved.

Description

Technical Field

[0001] The present invention relates to a non-invasive blood pressure monitoring method and device based on a deep learning algorithm. Background Art

[0002] A blood pressure monitoring device is a device used to measure blood pressure. It automatically acquires pressure, usually recording it at regular intervals, using either direct or indirect methods of determining pressure. There are two methods for measuring arterial blood pressure:

[0003] The first is direct arterial blood pressure monitoring, which uses an arterial catheter connected to a pressure transducer. This system allows for continuous monitoring of the patient's systolic, diastolic, and mean arterial pressures and also simplifies the collection of arterial blood gas samples. However, this monitoring method is not used as frequently as other methods because an arterial catheter must be placed.

[0004] The second type of blood pressure monitoring is indirect arterial blood pressure monitoring, which relies on noninvasive detection of peripheral arterial blood flow or vessel wall motion and provides intermittent arterial blood pressure measurements. This monitoring technique is the most commonly used in clinical practice, with the most commonly used indirect methods being Doppler and oscillometric blood pressure monitoring. To obtain indirect arterial blood pressure readings, a cuff is inflated over the artery until blood flow is blocked.

[0005] With the continuous development of artificial intelligence, traditional blood pressure monitoring equipment can only measure the patient's blood pressure, with relatively simple functions, and cannot meet the needs of intelligent modern medical equipment. Summary of the Invention

[0006] The purpose of the present invention is to overcome the existing defects and provide a non-invasive blood pressure monitoring method and device based on a deep learning algorithm. Through the machine learning algorithm, the patient's subsequent blood pressure can be analyzed and predicted while measuring blood pressure, thereby improving the level of intelligence.

[0007] The technical solution for achieving the above object is: a non-invasive blood pressure monitoring device based on a deep learning algorithm, comprising a sensor module, a control module, and a display module, wherein the sensor module is electrically connected to the control module, a digital filter is provided between the sensor module and the control module, and the control module is electrically connected to the display module;

[0008] The monitoring method includes the above-mentioned monitoring device, and the monitoring method includes the following steps:

[0009] S1: The sensor module is placed close to the measurement person's collection position, and the sensor module transmits the collected data signal to the digital filter;

[0010] S2 data signal is input into the digital filter, the digital filter processes the signal, extracts frequency domain or time domain features from the blood pressure waveform data, and transmits the data to the control module;

[0011] S3: After the pre-processing is completed, the information is processed by the control module to obtain a blood pressure measurement curve and an analysis and prediction of the blood pressure curve, and then the control module transmits the data to the display module;

[0012] S4 The display module displays the blood pressure curve and related data information.

[0013] The frequency domain or time domain features include systolic pressure, diastolic pressure, pulse pressure, heart rate, pulse pressure and cardiac index, etc.

[0014] The control module has two built-in algorithms, namely a deep learning algorithm and an artificial intelligence algorithm. After the preprocessing is completed, the information is processed by the deep learning algorithm model and the artificial intelligence algorithm model respectively to obtain a blood pressure measurement curve and predicted data.

[0015] The algorithm model establishment of the deep learning algorithm includes the following steps:

[0016] S11 Data Collection and Preparation: Collect a large-scale blood pressure dataset, including blood pressure measurement characteristics and corresponding blood pressure value labels, and preprocess the collected data, including data cleaning, standardization, and removal of outliers;

[0017] S12 Model Selection and Design: Use multiple layers of convolutional and pooling layers to extract features and reduce dimensionality, and then use fully connected layers to predict the final blood pressure value;

[0018] S13 Model Training and Optimization: Divide the dataset into training and test sets for model training and evaluation. Use an appropriate loss function as the objective function for training and use the gradient descent algorithm and its variants to optimize model parameters. Regularization methods and random dropout techniques can also be used to prevent overfitting.

[0019] S14 Model Evaluation and Tuning: Use the test set to evaluate the performance of the model. Based on the evaluation results, tune the model. You can consider adjusting the network structure, changing hyperparameters, or trying other optimization algorithms.

[0020] S15 Cross-validation and model fusion: To more accurately evaluate the performance of the model, you can use cross-validation to divide the dataset into multiple subsets for training and validation. In addition, consider using model fusion technology to combine multiple different models or model training processes.

[0021] S16 Result Interpretation and Explainability: In the medical field, model interpretability is crucial. Consider using attention mechanisms or visualization techniques to understand how the model uses input features to predict blood pressure.

[0022] S17 Continuous Improvement and Update: As data accumulates and the problem is better understood, the model is updated and adjusted based on new data and feedback to improve the accuracy and generalization ability of the blood pressure measurement model.

[0023] The deep learning algorithm model includes a vector machine model, and the creation steps are as follows:

[0024] S21 Data Preparation: First, a set of labeled blood pressure curve datasets is prepared as a training set. Each sample should contain the characteristic data of the blood pressure curve and the corresponding label, such as normal blood pressure, hypertension, or hypotension.

[0025] S22 Feature Extraction: In the blood pressure curve, some features can be extracted to represent important information such as the shape and volatility of the curve, for example, maximum value, minimum value, mean value, standard deviation, slope, etc.

[0026] S23 Data preprocessing: For machine learning algorithms, data preprocessing is usually required to ensure its quality and usability, which includes operations such as removing outliers and normalizing or standardizing data;

[0027] S24 model training: Use preprocessed training data as input and train the model using the support vector machine algorithm. In the support vector machine, a decision boundary is found to separate data points of different categories and to maximize the distance from the boundary to the nearest data point. This decision boundary is the hyperplane of the support vector machine model.

[0028] S25 Model Evaluation: Use the trained model to predict the test set data and compare it with the true labels in the test set to evaluate the model performance and prediction accuracy;

[0029] S26 Model Optimization and Parameter Adjustment: Optimize the model through parameter adjustment, feature selection, and other methods. Common optimization methods include grid search and cross-validation.

[0030] S27 Prediction and Classification: When the model training is completed and passes the evaluation, new blood pressure curve samples can be input into the trained model for prediction and classification. The model will determine the blood pressure category to which it belongs based on the characteristics of the input sample.

[0031] The deep learning algorithm model includes a decision tree model, and the decision tree model is established in the following steps:

[0032] S31 Data Preparation: Prepare a set of labeled blood pressure curve datasets as a training set. Each sample should contain the characteristic data of the blood pressure curve and the corresponding label, such as normal blood pressure, hypertension, or hypotension.

[0033] S32 Feature Extraction: Extract features from the blood pressure curve to build a decision tree model. These features may include maximum value, minimum value, mean, standard deviation, slope, etc., which are used to describe the morphological characteristics of the blood pressure curve;

[0034] S33 model construction: Use the feature data and corresponding labels of the training set to build a decision tree model. The decision tree is a classification algorithm based on a tree structure that gradually divides data into different categories through a series of decision rules.

[0035] S34 Feature Selection: When building a decision tree, you need to select features for segmenting data. Common feature selection methods include information gain, information gain ratio, and Gini coefficient. These methods can evaluate the contribution of features to classification and select the optimal features for segmentation.

[0036] S35 Decision Tree Construction: Based on the selected features, a decision tree is recursively constructed. Each node represents a feature, and each edge represents the value of the feature. By judging the value of the sample on the feature, the data is further segmented and child nodes are created until a leaf node is reached or the purity requirement is met.

[0037] S36 Model Evaluation: Use the trained decision tree model to predict the test set data and compare it with the true labels in the test set to evaluate the model performance and prediction accuracy;

[0038] S37 Visualization and Interpretation: Decision tree models are usually interpretable and can be visualized as a tree structure to facilitate understanding and interpretation of the model's decision-making process. By observing the branches and decision rules of the decision tree, insights and tips for blood pressure classification can be obtained;

[0039] S38 Prediction result interpretation: After using the support vector machine to classify and predict the blood pressure curve, the results need to be interpreted. According to the prediction results of the model, the blood pressure curve can be divided into normal blood pressure, hypertension or hypotension. For each curve, the patient's blood pressure status can be evaluated by judging its category.

[0040] S39 Model update and iteration: As data accumulates and new research progresses, the model can be updated and iterated periodically to add new training data and improve the performance and accuracy of the model through retraining.

[0041] The beneficial effects of the present invention are: by simultaneously processing the blood pressure data obtained by the sensor module through the deep learning algorithm model and the artificial intelligence algorithm model, a blood pressure measurement curve and prediction data are obtained, and the patient's hypertension risk factors and disease change trends are discovered, thereby providing the patient with more accurate treatment to achieve better blood pressure lowering effects, making the blood pressure monitoring device more intelligent. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 It is the overall flow chart of blood pressure monitoring of the present invention;

[0043] Figure 2 It is a flow chart for constructing the deep learning algorithm model of the present invention;

[0044] Figure 3 It is a flow chart of constructing the vector machine model of the present invention;

[0045] Figure 4 It is a flow chart for constructing the decision tree model filtering component of the present invention. DETAILED DESCRIPTION

[0046] The technical solution of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. In the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside" 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, and do not indicate or imply 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. In addition, the terms "first", "second", and "third" are used for descriptive purposes only and cannot be understood as indicating or implying relative importance.

[0047] The present invention will be further described below with reference to the accompanying drawings.

[0048] Implementation Case 1, combined with Figure 1-Figure 3 Analysis: A non-invasive blood pressure monitoring device based on a deep learning algorithm includes a sensor module, a control module and a display module. The sensor module is electrically connected to the control module, a digital filter is provided between the sensor module and the control module, and the control module is electrically connected to the display module. The device also includes a power supply module, which provides energy for the sensor module, the control module and the display module.

[0049] The monitoring method includes the above-mentioned monitoring device, and the monitoring method includes the following steps:

[0050] S1 places the sensor module close to the measurement person's collection position, and the sensor module transmits the collected data signal to the digital filter;

[0051] The S2 data signal is input into the digital filter, which processes the signal and extracts frequency domain or time domain features from the blood pressure waveform data, and transmits the data to the control module;

[0052] After the information is processed by the control module, the blood pressure measurement curve and the analysis and prediction of the blood pressure curve are obtained, and then the control module transmits the data to the display module;

[0053] The S4 display module shows the blood pressure curve and related data information.

[0054] Frequency domain or time domain features include systolic blood pressure, diastolic blood pressure, pulse pressure, heart rate, pulse pressure and cardiac index.

[0055] It should be noted that in practical applications, the classification and prediction of blood pressure curves involves medical expertise and experience. Therefore, in addition to machine learning algorithms, the participation and guidance of medical professionals are also required to ensure the accuracy and reliability of the prediction results.

[0056] Furthermore, to protect personal privacy and data security, appropriate measures should be taken to safeguard the confidentiality and integrity of blood pressure curve data and comply with relevant laws, regulations, and privacy policies. Furthermore, model outputs should be verified and supplemented, rather than relying solely on the judgment of machine learning algorithms.

[0057] The performance of a deep learning model depends not only on its structure but also on factors such as the quality, quantity, and diversity of the dataset. Furthermore, for medical applications, data privacy and security are crucial. A multi-layer convolutional neural network model can be constructed using the following steps:

[0058] S11 Data Collection and Preparation: Collect a large-scale blood pressure dataset, including the characteristics of blood pressure measurement and the corresponding blood pressure value labels, to ensure that the dataset covers different populations, different measurement conditions, and various factors that may affect blood pressure. Preprocess the data, including data cleaning, standardization, and removal of outliers.

[0059] S12 Model selection and design: Select an appropriate model based on the complexity of the problem and the size of the dataset. Convolutional neural network is a commonly used deep learning model that performs well in image processing problems. You can try using multiple layers of convolutional layers and pooling layers to extract features and reduce dimensions, and then use fully connected layers to predict the final blood pressure value.

[0060] S13 Model Training and Optimization: Divide the dataset into training and test sets for model training and evaluation. Use an appropriate loss function (such as mean squared error) as the training objective function and optimize model parameters using the gradient descent algorithm or its variants. Regularization methods and random dropout techniques can also be used to prevent overfitting.

[0061] S14 Model evaluation and tuning: Use the test set to evaluate the performance of the model and calculate various indicators (such as root mean square error, mean absolute error, etc.) to measure the gap between the predicted results and the true values; based on the evaluation results, tune the model. You can consider adjusting the network structure, changing hyperparameters (such as learning rate, batch size, etc.), or trying other optimization algorithms.

[0062] S15 Cross-validation and model fusion: In order to more accurately evaluate the performance of the model, a cross-validation method can be used to divide the dataset into multiple subsets for training and validation; in addition, consider using model fusion technology to combine multiple different models or model training processes to further improve the accuracy of blood pressure prediction.

[0063] S16 Result interpretation and explainability: In the medical field, model interpretability is crucial, so it is necessary to explain the decision-making process and key features of the model; you can consider using attention mechanisms or visualization techniques to understand how the model uses input features to predict blood pressure.

[0064] S17 Continuous Improvement and Update: As data accumulates and the problem is understood more deeply, it is important to continuously improve the model; update and adjust the model based on new data and feedback to improve the accuracy and generalization ability of the blood pressure measurement model.

[0065] First, the neural network multiplies the input feature vector x with the weight matrix W and the bias vector b, and then passes it through the activation function f to obtain the output of the hidden layer.

[0066] The output of the hidden layer can be expressed as:

[0067] z=Wx+b

[0068] a=f(z)

[0069] Where z represents the weighted input of the hidden layer and a represents the output of the hidden layer.

[0070] Then, the output of the hidden layer is used as the input of the next layer, and the above steps are repeated until the last output layer.

[0071] The input to the output layer can be expressed as:

[0072] z_out=W_out*a+b_out

[0073] a_out=f_out(z_out)

[0074] Among them, z_out represents the weighted input of the output layer, and a_out represents the output of the output layer.

[0075] Loss function:

[0076] In neural networks, a loss function is often used to measure the difference between the predicted value and the true value.

[0077] Assuming the true value is y and the predicted value is y_hat, the loss function can be expressed as L(y,y_hat).

[0078] Backward Propagation:

[0079] Backpropagation is the process used to update the weights and biases in a neural network so that the loss function is minimized.

[0080] First, calculate the gradient of the loss function with respect to the output layer input:

[0081] dL / dz_out=dL / da_out*df_out(z_out)

[0082] Then, the gradients in the hidden and input layers are calculated layer by layer using the chain rule:

[0083] dL / da=W_out^T*dL / dz_out

[0084] dL / dz=dL / da*df(z)

[0085] dL / dW=dL / dz*x^T

[0086] dL / db=dL / dz

[0087] Finally, the weights and biases are updated according to the gradient descent rule:

[0088] W_new=W-learning_rate*dL / dW

[0089] b_new=b-learning_rate*dL / db

[0090] Among them, dL / dW represents the gradient of the loss function with respect to the weight, dL / db represents the gradient of the loss function with respect to the bias, and learning_rate represents the learning rate used to control the update speed of the weight and bias.

[0091] By repeatedly iterating the forward and backpropagation steps, the neural network can continuously optimize the weights and biases, thereby improving the model's performance and prediction accuracy. It is important to note that factors such as the specific neural network structure and the choice of activation function will affect the specific calculation formula.

[0092] At the same time, the artificial intelligence algorithm uses a vector machine model, and its creation steps are:

[0093] S21 Data preparation: First, a set of labeled blood pressure curve datasets needs to be prepared as a training set; each sample should contain the characteristic data of the blood pressure curve and the corresponding label, such as normal blood pressure, hypertension or hypotension and other classification information.

[0094] S22 Feature Extraction: From the blood pressure curve, some features can be extracted to represent important information such as the shape and volatility of the curve; for example, maximum value, minimum value, mean value, standard deviation, slope, etc. These features will be used to describe the morphological characteristics of the blood pressure curve.

[0095] S23 Data preprocessing: For machine learning algorithms, data usually needs to be preprocessed to ensure its quality and usability; this includes operations such as removing outliers and normalizing or standardizing data.

[0096] S24 model training: Using preprocessed training data as input, the model is trained using the support vector machine algorithm. In the support vector machine, we find a decision boundary that separates data points of different categories and maximizes the distance from the boundary to the nearest data point. This decision boundary is the hyperplane of the support vector machine model.

[0097] S25 Model Evaluation: Use the trained model to predict the test set data and compare it with the true labels in the test set to evaluate the model performance and prediction accuracy; commonly used evaluation indicators include accuracy, precision, recall, F1 value, etc.

[0098] S26 Model optimization and parameter adjustment: If the model performance is not ideal, the model can be optimized through parameter adjustment, feature selection and other methods; common optimization methods include grid search, cross-validation, etc.

[0099] S27 Prediction and Classification: Once the model training is complete and passes the evaluation, new blood pressure curve samples can be input into the trained model for prediction and classification. The model will determine the blood pressure category to which it belongs based on the characteristics of the input sample.

[0100] If support vector regression is used, the calculation formula is as follows:

[0101] y=w1x1+w2x2+...+w x+b

[0102] Where y represents the predicted value of blood pressure; x1, x2, ..., x are the features of the feature vector X; w1, w2, ..., w are the weights of the model; and b is the bias term.

[0103] The key to introducing support vector regression into the linear regression model lies in the concept of support vector machines, which performs regression by finding an optimal hyperplane. In support vector regression, kernel functions are also introduced to handle nonlinear problems and map low-dimensional features into a high-dimensional feature space.

[0104] The goal of support vector regression is to optimize the model parameters to minimize the prediction error and maintain the largest margin when fitting the data. Different loss functions and regularization terms are usually used to define the optimization problem, such as ε-insensitive loss, L1 regularization, etc.

[0105] It is important to note that the specific support vector regression formula will vary depending on the kernel function used and the model parameters. The above is the basic linear formula for support vector regression. In actual applications, kernel functions (such as linear kernels, Gaussian kernels, etc.) may be used to handle nonlinear relationships and perform more complex model training and prediction.

[0106] By building a blood pressure measurement model using deep learning algorithms, we can monitor and analyze patients' blood pressure data in real time and adjust medication regimens based on the results to achieve better blood pressure-lowering effects. For example, if a patient's blood pressure exceeds the target value, we can appropriately increase the medication dosage or change the type of medication to quickly lower the patient's blood pressure.

[0107] Implementation Case 2, combined with Figure 4 The analytical artificial intelligence algorithm uses a vector machine model, and its creation steps are:

[0108] S31 Data preparation: Prepare a set of labeled blood pressure curve datasets as a training set; each sample should contain the characteristic data of the blood pressure curve and the corresponding label, such as normal blood pressure, hypertension, or hypotension.

[0109] S32 Feature Extraction: Extract features from the blood pressure curve to build a decision tree model; these features can include maximum value, minimum value, mean, standard deviation, slope, etc., which are used to describe the morphological characteristics of the blood pressure curve.

[0110] S33 model construction: Use the feature data and corresponding labels of the training set to build a decision tree model; the decision tree is a classification algorithm based on a tree structure that gradually divides data into different categories through a series of decision rules.

[0111] S34 Feature Selection: When building a decision tree, you need to select features for data segmentation. Common feature selection methods include information gain, information gain ratio, and Gini coefficient. These methods can evaluate the contribution of features to classification and select the optimal features for segmentation.

[0112] S35 Decision Tree Construction: Based on the selected features, a decision tree is recursively constructed. Each node represents a feature, and each edge represents the value of that feature. By determining the sample's value on that feature, the data is further segmented and child nodes are created until a stopping condition is met (such as reaching a leaf node or achieving the required purity).

[0113] S36 Model Evaluation:

[0114] Use the trained decision tree model to make predictions on the test set data and compare them with the true labels in the test set to evaluate the model performance and prediction accuracy. Common evaluation metrics include accuracy, precision, recall, and F1 score.

[0115] S37 Visualization and Interpretation: Decision tree models are usually interpretable and can be visualized as a tree structure to facilitate understanding and interpretation of the model's decision-making process. By observing the branches and decision rules of the decision tree, insights and hints about blood pressure classification can be obtained.

[0116] S38 prediction results interpretation:

[0117] After using a support vector machine to classify and predict blood pressure curves, the results need to be interpreted. Based on the model's predictions, blood pressure curves can be classified as normal blood pressure, hypertension, or hypotension. For each curve, the patient's blood pressure status can be assessed by determining its category.

[0118] S39 model updates and iterations:

[0119] As data accumulates and new research progresses, the model can be periodically updated and iterated to add new training data and improve the performance and accuracy of the model through retraining; this helps the model better adapt to changing blood pressure data and diagnostic requirements.

[0120] Decision tree regression is a regression method based on the decision tree algorithm. Its calculation formula is as follows:

[0121] For a given input feature vector x, decision tree regression will make a prediction through a decision tree model. The basic idea of decision tree regression is to divide the input space into multiple regions and fit a local linear regression model in each region.

[0122] Specifically, decision tree regression constructs a decision tree, where each internal node represents a judgment about an input feature and each leaf node represents a predicted value. The decision tree construction process involves selecting the best features and partitioning points so that the samples in each partitioned subset have a small squared error or other metric.

[0123] The output of decision tree regression can be expressed as:

[0124] y=Σ(ci) / |R|

[0125] Where y represents the predicted value, Σ(ci) represents the sum of the true values of all training samples falling into the leaf node where the sample is located, and |R| represents the number of samples in the leaf node where the sample is located.

[0126] When performing decision tree regression, depending on different algorithms and strategies, parameters such as pruning, setting the minimum number of samples for leaf nodes, and depth limits may be introduced to avoid overfitting or improve the generalization ability of the model.

[0127] It’s important to note that the decision tree regression formula is based on the partitioning of the input space and local linear fitting, and does not have a clear mathematical expression like linear regression. The specific judgment conditions and branching decisions of each decision tree node are automatically learned by the decision tree algorithm based on the training data.

[0128] Implementation Case 3: Artificial intelligence algorithms can also use linear regression models for calculations. Example of the linear regression model calculation formula

[0129] Assume that the feature vector is X = [x1, x2, ..., x], the corresponding weight vector is W = [w1, w2, ..., w], and the bias term is b. The linear regression model is calculated as follows:

[0130] y=w1x1+w2x2+...+w x+b

[0131] Where y represents the predicted value of blood pressure.

[0132] In this example, the feature vector X can include features extracted from the blood pressure waveform data, such as frequency domain features obtained by Fourier transform or wavelet transform, as well as other blood pressure-related features. The weight vector W is the parameter that the model needs to learn, which controls the contribution of each feature to blood pressure prediction. The bias term b is the offset of the model.

[0133] It's important to note that different models and algorithms may have different operational formulas. This example uses a simple linear regression model. In practice, more complex models and features may be used to improve the accuracy of blood pressure prediction. Furthermore, data preprocessing, model training, and evaluation are required to achieve better prediction results.

[0134] Implementation Case 4: Artificial intelligence algorithms can also use random forest models for calculations. Random forest regression is a regression method based on the random forest algorithm. Its calculation formula can be divided into two parts: random forest model construction and prediction.

[0135] Random forest model construction:

[0136] First, Random Forest constructs multiple decision trees. Each decision tree is generated by random selection of features and bootstrap sampling of the training set.

[0137] Each decision tree in a random forest performs full feature selection and partitioning to build a decision tree model with a certain depth. The specific partitioning method may be to select the optimal partition point based on Gini impurity or mean square error.

[0138] Random Forest Model Prediction:

[0139] For a given input feature vector x, random forest regression obtains the final prediction value by averaging the prediction results of each decision tree.

[0140] The prediction formula is as follows:

[0141] y=(y1+y2+...+y) / n

[0142] Where y represents the final predicted value, y1, y2, ..., y represent the predicted values of each decision tree respectively, and n represents the number of decision trees in the random forest.

[0143] It's important to note that random forest regression makes predictions by ensembling multiple decision trees, with each decision tree's predictions assigned a certain weight. In the simplest case, the weights are equal, i.e., the mean. However, in practice, the weights of individual decision trees may be adjusted based on factors such as model performance or sample weight.

[0144] In summary, the calculation formula of random forest regression can be regarded as a weighted average of the prediction results of multiple decision trees.

[0145] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A non-invasive blood pressure monitoring device based on a deep learning algorithm, characterized in that: The system comprises a sensor module, a control module and a display module, wherein the sensor module is electrically connected to the control module, a digital filter is provided between the sensor module and the control module, and the control module is electrically connected to the display module; The monitoring method includes the above-mentioned monitoring device, and the monitoring method includes the following steps: S1: The sensor module is placed close to the measurement person's collection position, and the sensor module transmits the collected data signal to the digital filter; S2 data signal is input into the digital filter, the digital filter processes the signal, extracts frequency domain or time domain features from the blood pressure waveform data, and transmits the data to the control module; S3: After the pre-processing is completed, the information is processed by the control module to obtain a blood pressure measurement curve and an analysis and prediction of the blood pressure curve, and then the control module transmits the data to the display module; S4 The display module displays the blood pressure curve and related data information.

2. A non-invasive blood pressure monitoring method and device based on deep learning algorithm according to claim 1, characterized in that: The frequency domain or time domain features include systolic pressure, diastolic pressure, pulse pressure, heart rate, pulse pressure and cardiac index, etc.

3. The non-invasive blood pressure monitoring method and device based on deep learning algorithm according to claim 1, characterized in that: The control module has two built-in algorithms, namely a deep learning algorithm and an artificial intelligence algorithm. After the preprocessing is completed, the information is processed by the deep learning algorithm model and the artificial intelligence algorithm model respectively to obtain a blood pressure measurement curve and predicted data.

4. The non-invasive blood pressure monitoring method and device based on deep learning algorithm according to claim 2, characterized in that: The algorithm model establishment of the deep learning algorithm includes the following steps: S11 Data Collection and Preparation: Collect a large-scale blood pressure dataset, including blood pressure measurement characteristics and corresponding blood pressure value labels, and preprocess the collected data, including data cleaning, standardization, and removal of outliers; S12 Model Selection and Design: Use multiple layers of convolutional and pooling layers to extract features and reduce dimensionality, and then use fully connected layers to predict the final blood pressure value; S13 Model Training and Optimization: Divide the dataset into training and test sets for model training and evaluation. Use an appropriate loss function as the objective function for training and use the gradient descent algorithm and its variants to optimize model parameters. Regularization methods and random dropout techniques can also be used to prevent overfitting. S14 Model Evaluation and Tuning: Use the test set to evaluate the performance of the model. Based on the evaluation results, tune the model. You can consider adjusting the network structure, changing hyperparameters, or trying other optimization algorithms. S15 Cross-validation and model fusion: To more accurately evaluate the performance of the model, you can use cross-validation to divide the dataset into multiple subsets for training and validation. In addition, consider using model fusion technology to combine multiple different models or model training processes. S16 Result Interpretation and Explainability: In the medical field, model interpretability is crucial. Consider using attention mechanisms or visualization techniques to understand how the model uses input features to predict blood pressure. S17 Continuous Improvement and Update: As data accumulates and the problem is better understood, the model is updated and adjusted based on new data and feedback to improve the accuracy and generalization ability of the blood pressure measurement model.

5. The non-invasive blood pressure monitoring method and device based on deep learning algorithm according to claim 1, characterized in that: The artificial intelligence algorithm model includes a vector machine model, and the creation steps are as follows: S21 Data Preparation: First, a set of labeled blood pressure curve datasets is prepared as a training set. Each sample should contain the characteristic data of the blood pressure curve and the corresponding label, such as normal blood pressure, hypertension, or hypotension. S22 Feature Extraction: In the blood pressure curve, some features can be extracted to represent important information such as the shape and volatility of the curve, for example, maximum value, minimum value, mean value, standard deviation, slope, etc. S23 Data preprocessing: For machine learning algorithms, data preprocessing is usually required to ensure its quality and usability, which includes operations such as removing outliers and normalizing or standardizing data; S24 model training: Use preprocessed training data as input and train the model using the support vector machine algorithm. In the support vector machine, a decision boundary is found to separate data points of different categories and to maximize the distance from the boundary to the nearest data point. This decision boundary is the hyperplane of the support vector machine model. S25 Model Evaluation: Use the trained model to predict the test set data and compare it with the true labels in the test set to evaluate the model performance and prediction accuracy; S26 Model Optimization and Parameter Adjustment: Optimize the model through parameter adjustment, feature selection, and other methods. Common optimization methods include grid search and cross-validation. S27 Prediction and Classification: When the model training is completed and passes the evaluation, new blood pressure curve samples can be input into the trained model for prediction and classification. The model will determine the blood pressure category to which it belongs based on the characteristics of the input sample.

6. The non-invasive blood pressure monitoring method and device based on deep learning algorithm according to claim 1, characterized in that: The artificial intelligence algorithm model includes a decision tree model, and the decision tree model is established in the following steps: S31 Data Preparation: Prepare a set of labeled blood pressure curve datasets as a training set. Each sample should contain the characteristic data of the blood pressure curve and the corresponding label, such as normal blood pressure, hypertension, or hypotension. S32 Feature Extraction: Extract features from the blood pressure curve to build a decision tree model. These features may include maximum value, minimum value, mean, standard deviation, slope, etc., which are used to describe the morphological characteristics of the blood pressure curve; S33 model construction: Use the feature data and corresponding labels of the training set to build a decision tree model. The decision tree is a classification algorithm based on a tree structure that gradually divides data into different categories through a series of decision rules. S34 Feature Selection: When building a decision tree, you need to select features for segmenting data. Common feature selection methods include information gain, information gain ratio, and Gini coefficient. These methods can evaluate the contribution of features to classification and select the optimal features for segmentation. S35 Decision Tree Construction: Based on the selected features, a decision tree is recursively constructed. Each node represents a feature, and each edge represents the value of the feature. By judging the value of the sample on the feature, the data is further segmented and child nodes are created until a leaf node is reached or the purity requirement is met. S36 Model Evaluation: Use the trained decision tree model to predict the test set data and compare it with the true labels in the test set to evaluate the model performance and prediction accuracy; S37 Visualization and Interpretation: Decision tree models are usually interpretable and can be visualized as a tree structure to facilitate understanding and interpretation of the model's decision-making process. By observing the branches and decision rules of the decision tree, insights and tips for blood pressure classification can be obtained; S38 Prediction result interpretation: After using the support vector machine to classify and predict the blood pressure curve, the results need to be interpreted. According to the prediction results of the model, the blood pressure curve can be divided into normal blood pressure, hypertension or hypotension. For each curve, the patient's blood pressure status can be evaluated by judging its category. S39 Model update and iteration: As data accumulates and new research progresses, the model can be updated and iterated periodically to add new training data and improve the performance and accuracy of the model through retraining.

Images