Atrial fibrillation measurement method based on low-pressure pulse measurement method

Abstract

The invention provides an atrial fibrillation measurement method based on a low-pressure pulse measurement method, and relates to the field of medical instruments. The atrial fibrillation measurement method based on the low-pressure pulse measurement method specifically comprises the following steps that S1, a sensor is assembled; s2, data acquisition and transmission; s3, signal processing and feature extraction; s4, carrying out algorithm development and machine learning; and S5, verifying and optimizing. The atrial fibrillation measuring method based on the low-pressure pulse measuring method has the advantages of being non-invasive, portable and continuous in monitoring, heart rhythms can be monitored in real time in a seamless integration mode in daily life, heart rhythm data are captured through a portable device, analysis is carried out in combination with an intelligent algorithm, feedback is provided, early diagnosis and timely treatment are facilitated, and the atrial fibrillation measuring method is suitable for being used in daily life. Compared with a traditional heart rhythm monitoring means, the method is more economical and user-friendly, meanwhile, data safety and privacy protection are ensured, and the method is widely applied to different living environments.

Description

Technical Field

[0001] The invention relates to the field of medical technology, and in particular to an atrial fibrillation measurement method based on a low-pressure pulse measurement method. Background Art

[0002] Atrial fibrillation is a heart rhythm disorder that causes the upper chambers of the heart to beat irregularly and rapidly. In atrial fibrillation, the upper part of the heart (the atria) no longer contracts in a coordinated manner, but instead beats in a rapid and irregular pattern, which can cause the heart to pump less efficiently. Atrial fibrillation is a common heart rhythm disorder that can lead to a range of health problems, including heart failure, stroke, and other vascular events. Atrial fibrillation can be associated with conditions such as heart disease, high blood pressure, hyperthyroidism, or can be caused by other factors. The purpose of taking an atrial fibrillation measurement is to determine if atrial fibrillation is present in the heart and to assess its severity. Atrial fibrillation measurements are usually taken with an electrocardiogram (ECG) or other heart rhythm monitoring methods. After atrial fibrillation is detected, doctors can take steps to manage it, including medication, cardiac ablation procedures, or implanting a pacemaker. Early identification and treatment of atrial fibrillation can help reduce the risk of associated complications and improve the patient's quality of life.

[0003] The existing technologies used for atrial fibrillation (AF) detection mainly include electrocardiogram (ECG) monitoring. ECG is a method of monitoring heart health by recording the electrical activity of the heart. It can provide detailed information on the electrical activity of the heart in each cycle and is the gold standard for diagnosing arrhythmias such as atrial fibrillation. However, its traditional ECG equipment is usually large and not suitable for long-term wear or carrying outside. The installation and data interpretation of ECG usually require professional medical personnel, which limits its popularity in the home. Heart sound analysis, which evaluates the health of the heart by analyzing the sounds (heart sounds) emitted by the heart. This method usually uses a stethoscope or a special heart sound monitoring device. The heart sound signal may be affected by environmental noise, and its accuracy is lower than that of the ECG, especially in non-ideal environments. Blood pressure measurement, which indirectly reflects the heart state by measuring blood pressure. Some blood pressure monitors are equipped with the function of detecting arrhythmias, which can be used to indicate atrial fibrillation. Blood pressure measurement mainly provides blood pressure information, and has limited direct diagnostic capabilities for arrhythmias. Wearable device monitoring, using wearable devices (such as smart watches or health trackers) to monitor heart rate and other physiological parameters for daily health management, compared with traditional medical devices, the data accuracy and reliability of wearable devices are often lower, especially in extreme exercise or non-standard use conditions. Although these existing technologies have their own advantages, they still have certain limitations in terms of portability, user-friendliness, continuous monitoring capabilities and accuracy. Therefore, the present invention proposes a method for measuring atrial fibrillation based on low-voltage pulse measurement to solve the above-mentioned problems. Summary of the invention

[0004] 1. Technical issues to be resolved In view of the deficiencies of the prior art, the present invention provides an atrial fibrillation measurement method based on low-voltage pulse measurement, which solves the problem that the prior art still has certain limitations in terms of portability, user-friendliness, continuous monitoring capability and accuracy.

[0005] (II) Technical solution To achieve the above objectives, the present invention is implemented by the following technical solutions: a method for measuring atrial fibrillation based on a low-pressure pulse measurement method, specifically comprising the following steps: S1. Sensor Assembly Attach the sensor to the wrist, fingertip or earlobe while ensuring that the sensor can be worn comfortably without interfering with daily activities; S2. Data collection and transmission Utilize installed sensors to continuously monitor and record, and design systems to continuously monitor and record low-voltage pulse signals, including key parameters of pulse waveform and frequency; S3. Signal processing and feature extraction Preprocess the signal data and use digital filtering technology to remove noise and interference in the signal, including removing artifacts and electromagnetic interference caused by movement. Then perform feature extraction to extract features related to arrhythmia in the pulse signal, such as irregularity of the RR interval and waveform variability, to extract useful information from the original pulse signal and convert it into structured data that can be used for subsequent machine learning model training. S4. Algorithm Development and Machine Learning Develop algorithms for analyzing low-voltage pulse signal characteristics, using time series analysis, frequency analysis methods, and then using machine learning models to identify signal patterns that are unique to atrial fibrillation, and continuously use new data to train and optimize the model to improve accuracy; S5. Verification and Optimization Cooperate with professional medical institutions to conduct clinical trials to verify the accuracy and reliability of the system, adjust algorithm parameters based on clinical feedback, optimize sensor design, and improve user experience and detection accuracy.

[0006] Preferably, in the step S1, the sensor assembly preferably comprises any one or more combinations of a photoplethysmographic (PPG) sensor or a microrheological pressure sensor.

[0007] Preferably, in the step S2, data collection and transmission is performed by wireless connection to achieve data transmission, cache and store the data, and encrypted transmission is also required during the data transmission.

[0008] Preferably, the signal processing and feature extraction in step S3 specifically includes signal denoising, pulse waveform detection, feature extraction of the transmitted model, and then standardization and normalization and data set construction, and finally preliminary graphic analysis.

[0009] Preferably, the algorithm development in step S4 and the machine learning model in the machine learning use any one or more combinations of support vector machines, random forests, and deep learning networks.

[0010] Preferably, after selecting the machine learning model in step S4 algorithm development and machine learning, model training is also required, specifically including cross-validation, evaluation indicators and model persistence.

[0011] Preferably, the verification and optimization in step S5 specifically includes verifying the degree of freedom of the model, using independent data for verification, algorithm performance and accuracy testing, model optimization, and then performing result feedback, model updating and retesting.

[0012] (III) Beneficial effects The present invention provides an atrial fibrillation measurement method based on a low-pressure pulse measurement method. It has the following beneficial effects: The present invention provides an atrial fibrillation measurement method based on a low-voltage pulse measurement method. The method of the present invention significantly reduces the physiological interference and discomfort to the user by simply measuring the pulse waveform of the wrist, finger or other peripheral parts without skin puncture or implantation of devices. The technology of the present invention can be integrated into smart watches and health tracker portable devices, so that users can easily monitor in daily life without special settings or professional knowledge. Users can perform daily heart rhythm monitoring like wearing a watch. Unlike traditional electrocardiograms or periodic heart rhythm monitoring, the low-voltage pulse measurement method can achieve all-weather and continuous data collection, which helps to capture occasional atrial fibrillation events that are easily missed in clinical practice and provide users with more comprehensive heart health monitoring. The device integrated with the intelligent algorithm of the present invention can analyze pulse data in real time, detect abnormal heart rhythm in time and remind users through smart phones or other devices. This real-time feedback mechanism is crucial for early diagnosis and timely treatment of atrial fibrillation. Compared with traditional diagnostic tools (such as electrocardiographs), low-voltage pulse measurement devices are usually lower in cost and simpler to maintain, which not only reduces the economic burden of users, but also contributes to the widespread promotion and application of this technology. The devices implemented by the method of the present invention are usually equipped with encryption and security protection measures to ensure the security and privacy of user data and avoid the leakage of personal health information. The measurement method of the present invention is not restricted by the environment and can be used at home, in the office or during exercise, which makes it very suitable for active modern lifestyles and individuals with continuous needs for health monitoring. Therefore, the atrial fibrillation measurement method based on low-voltage pulse measurement of the present invention provides a user-friendly, efficient and economical solution, which can improve the monitoring and management of arrhythmias, especially for atrial fibrillation, a common heart rhythm problem that may cause serious complications. DETAILED DESCRIPTION

[0013] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. 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 work are within the scope of protection of the present invention.

[0014] Example: The embodiment of the present invention provides an atrial fibrillation measurement method based on a low-voltage pulse measurement method, which specifically includes the following steps: S1. Sensor Assembly Attach the sensor to your wrist, fingertip, or earlobe, while ensuring that the sensor can be worn comfortably without interfering with daily activities.

[0015] Photoplethysmography (PPG) sensor: The PPG sensor measures pulse signals optically. It obtains the pulse waveform by emitting light of a specific wavelength to the skin and detecting the intensity change of the reflected light. Usually a green light or infrared light source is used, this sensor is suitable for the wrist, fingertips and earlobes.

[0016] Microrheological pressure sensors: These sensors detect tiny pressure changes caused by blood flow based on the piezoresistive effect, piezoelectric effect, or capacitive effect. They are sensitive to minute pressure changes and can measure the arterial blood pressure waveform caused by the pulse.

[0017] Selection Criteria: Compare the sensitivity, accuracy, size, durability, and energy consumption characteristics of different sensors to determine the type that is most suitable for continuous monitoring and data collection.

[0018] Sensor Location and Comfort: Location selection: Generally, suitable monitoring locations include the wrist, fingertips, and earlobes. Clinical trials are conducted to evaluate the quality of pulse waveform signals at different locations and find the optimal location.

[0019] Wearing comfort: The wearing position should avoid friction and pressure, ensuring that the user does not easily fall off or feel uncomfortable during daily activities.

[0020] Sensor design and development: Modularity: Choose a flexibly replaceable module design to facilitate future upgrades and maintenance.

[0021] Anti-interference design: Design a durable housing to resist environmental interference and enable the sensor to fit the skin stably.

[0022] Power consumption control: Use low power consumption design to reduce the impact on battery life and ensure continuity of long-term monitoring.

[0023] Experimental verification of the sensor: Performance testing: Experimental testing of the sensitivity, noise suppression and response speed indicators of the selected sensors.

[0024] Simulated environmental testing: actual application scenarios in various environments such as human motion, temperature and humidity are simulated in the laboratory to evaluate the performance of sensors in reality.

[0025] Comparison with other detection methods: The collected pulse data is compared with clinical standard electrocardiogram and Holter monitoring to ensure the accuracy of the sensor data.

[0026] These steps allow you to select and optimize a low-voltage pulse sensor suitable for atrial fibrillation measurement, laying the foundation for accurate detection.

[0027] S2. Data collection and transmission Utilize installed sensors to continuously monitor and record, and design a system to continuously monitor and record low-voltage pulse signals, including key parameters of pulse waveform and frequency.

[0028] Acquisition hardware configuration: Data acquisition module: Equipped with a high-performance data acquisition module to ensure that there is no loss or error in the process of data collection from the sensor to the central processor.

[0029] Sampling frequency and resolution: Set the sampling frequency (such as 100 Hz or above) and resolution to ensure accurate capture of the pulse signal and provide high-quality raw data for subsequent data analysis.

[0030] Data preprocessing: Signal amplification and filtering: Through analog or digital signal processing technology, the amplitude of the pulse signal is adjusted to an appropriate range, and electromagnetic interference and motion artifact noise are removed.

[0031] Signal segmentation and labeling: Segment the pulse signal according to its characteristics and add time stamps to the data for analysis and alignment.

[0032] Data transmission method: Wireless connection: Use wireless technologies such as Bluetooth, Wi-Fi or Zigbee for real-time data transmission. Bluetooth or Wi-Fi is preferred because these two methods support portable devices and smartphones and are suitable for mobile application scenarios.

[0033] Data format: Encode data into a standardized format (such as CSV, JSON, or XML) to improve transmission compatibility and ensure that different platforms can decode and parse it smoothly.

[0034] Central processing unit configuration: Data caching and storage: Set up a data cache on a central processing unit (such as a smartphone or server) to avoid data loss due to transmission interruptions or delays. At the same time, use persistent storage (such as a cloud database) to back up data.

[0035] Real-time data synchronization: Real-time data synchronization and backup between multiple devices through wireless networks or the Internet.

[0036] Data Security and Privacy: Encrypted transmission: Use TLS / SSL protocol or other encryption methods to ensure that data is not stolen or tampered with during transmission.

[0037] Access control: Set user permissions to manage the reading, writing, and deletion of data to ensure that private information is not leaked.

[0038] Audit and Tracking: Record data access logs to facilitate tracking of abnormal operations and strengthen data protection.

[0039] By implementing the above steps, high-quality low-voltage pulse data acquisition and reliable transmission can be effectively achieved, providing a solid foundation for subsequent analysis of atrial fibrillation detection.

[0040] S3. Signal processing and feature extraction The signal data is preprocessed, and digital filtering technology is used to remove noise and interference in the signal, including removing artifacts and electromagnetic interference caused by movement. Feature extraction is then performed to extract features related to arrhythmias in the pulse signal, such as irregularity of the RR interval and waveform variability, to extract useful information from the original pulse signal and convert it into structured data that can be used for subsequent machine learning model training.

[0041] Signal denoising: Apply filters: Use low-pass, high-pass, or band-pass filters to remove noise from the signal. For example, use Butterworth, Chebyshev, or moving average filters to remove non-heartbeat related high-frequency noise and low-frequency drift.

[0042] Adaptive filtering: In more complex or uncertain noise environments, adaptive filters, such as the LMS algorithm, are used to automatically adjust the filter parameters to match the characteristics of the signal.

[0043] Pulse waveform detection: Peak identification: Identify the peak by finding the local maximum point in the pulse signal. This process can use threshold method or derivative-based method.

[0044] Waveform segmentation: Using the peak as a marker, the continuous pulse signal is segmented into individual pulse cycles for further waveform analysis and feature extraction.

[0045] Feature extraction: Time domain characteristics: measure the duration of each pulse cycle, the height of the peak and trough, and the slope of rise and fall.

[0046] Frequency domain features: The pulse signal is converted to the frequency domain through Fourier transform (FFT) to analyze the spectral components of the signal, such as the main frequency components and power spectrum density.

[0047] Nonlinear Features: Apply nonlinear dynamics tools such as Poincaré plots, approximate entropy (AppEn), and sample entropy (SampEn) to reveal the nonlinear characteristics of signals.

[0048] Standardization and Normalization: The extracted features are normalized to eliminate the effects of different scales and dimensions so that the feature values ​​are limited to a uniform range, such as [0, 1] or [-1, 1].

[0049] Dataset construction: The extracted feature values ​​are integrated into a dataset for machine learning. Each sample contains multiple feature values ​​and corresponding labels (such as normal or atrial fibrillation).

[0050] Graphics and Statistical Analysis: Preliminary graphical analyses (e.g., scatter plots, box plots) and statistical tests (e.g., T-tests) were performed to assess the effectiveness and significance of different features in distinguishing normal from abnormal heartbeats.

[0051] Through the above steps, useful information is extracted from the original pulse signal and converted into structured data that can be used for subsequent machine learning model training. This is an important basis for automatically identifying atrial fibrillation.

[0052] S4. Algorithm Development and Machine Learning Develop an algorithm for analyzing the characteristics of low-voltage pulse signals, using time series analysis and frequency analysis methods, and then using a machine learning model to identify signal patterns unique to atrial fibrillation. Continuously use new data to train and optimize the model to improve accuracy.

[0053] Select a machine learning model: Supervised learning models: Commonly used ones include support vector machines (SVM), random forests, gradient boosting machines (GBM), and neural networks. When choosing a model, consider the nature of the problem (classification), the amount of data, the type of features, and the expected performance.

[0054] Deep learning models: For complex pattern recognition and large data sets, consider using convolutional neural networks (CNN) or recurrent neural networks (RNN).

[0055] Data preprocessing: Standardization: Ensure that all numerical features have similar scales, usually using Z-score standardization or Min-Max standardization.

[0056] Class imbalance processing: Use oversampling (such as SMOTE technology) or undersampling technology to solve the imbalance in the number of normal heartbeat and atrial fibrillation categories.

[0057] Divide the dataset: Training set, validation set and test set: usually divided into 70% (training set), 15% (validation set) and 15% (test set). The training set is used for model training, the validation set is used for model selection and parameter adjustment, and the test set is used for final model evaluation.

[0058] Feature Engineering: Feature selection: Use feature selection algorithms (such as recursive feature elimination (RFE), model-based feature selection, etc.) to remove redundant or irrelevant features.

[0059] Feature construction: Develop new features, such as interaction features or polynomial features derived from original features, to improve model performance.

[0060] Model training and parameter adjustment: Cross-validation: K-fold cross-validation method is used to evaluate the generalization ability of the model.

[0061] Hyperparameter optimization: Find the best model parameters using grid search, random search, or Bayesian optimization methods.

[0062] Model Evaluation: Evaluation indicators: Accuracy, precision, recall, F1 score and ROC-AUC indicators are used to comprehensively evaluate model performance.

[0063] Confusion Matrix: Construct a confusion matrix to visualize the performance of the model on each category.

[0064] Model deployment: Model persistence: Save the trained model using a suitable format (such as Pickle, HDF5, ONNX).

[0065] Integrate into applications: Integrate the model into the final application or system to ensure that the model can run stably in the production environment.

[0066] Continuous monitoring and optimization: Performance monitoring: Continuously monitor the performance and stability of the model in the production environment.

[0067] Retraining mechanism: Regularly update and optimize the model based on newly collected data to address the concept drift problem.

[0068] Through these steps, a robust machine learning system can be established to effectively identify heart diseases such as atrial fibrillation and improve the accuracy and efficiency of diagnosis.

[0069] S5. Verification and Optimization Cooperate with professional medical institutions to conduct clinical trials to verify the accuracy and reliability of the system, adjust algorithm parameters based on clinical feedback, optimize sensor design, and improve user experience and detection accuracy.

[0070] Verify the degrees of freedom of the model: Check the model for overfitting and underfitting. If the model appears to be overfitting, you can simplify the model, add more data, or use regularization. If it appears to be underfitting, you can try using a more complex model or adding more features.

[0071] Verification using independent data: Collect new, independent data from real-world situations for model validation. This can provide information about how the model performs with real-world data, which is critical to understanding the effectiveness of the model.

[0072] Algorithm performance and accuracy test: Calculate common performance metrics for the models, such as precision, recall, F1 score, and ROC-AUC, and compare them. This can help determine the accuracy and robustness of the model for scene recognition.

[0073] Model optimization: Hyperparameter tuning: Systematically adjusting the parameters of a machine learning algorithm to find an optimal solution. Changing the values ​​of parameters such as the learning rate and the number of batches used for gradient descent can often improve model performance while keeping the model structure unchanged.

[0074] Feature optimization: During feature selection, see if you can add or remove some features to determine if they have an impact on the model results. If a feature has little impact on the results, then you can try removing it.

[0075] Result feedback and model update: Based on the results of the validation phase, the algorithm model is adaptively optimized and updated. For example, if the model performs poorly in a specific category or scenario, this needs to be focused on and improved.

[0076] Keep monitoring your model to ensure that its performance remains stable over time. If performance degrades, diagnose the problem and take appropriate action (such as retraining the model or updating hyperparameters) to fix it.

[0077] Retest: All optimization processes need to be retested to confirm whether the modified algorithm or model has improved the prediction accuracy and whether it can make more accurate judgments on categories such as atrial fibrillation.

[0078] Through the above steps, the accuracy and reliability of the algorithm can be guaranteed, and it can be optimized to improve the diagnosis rate of atrial fibrillation, which is extremely important for early warning and treatment effects in practical applications.

[0079] Identify user needs: Before designing the UI, you must first conduct user research to understand the needs and habits of end users, including the convenience of user operations, the clarity of data presentation, and the core functions that users care about.

[0080] Designing the User Interface: Design a simple, intuitive interface based on user needs. Ensure that users can easily access and operate core functions, such as starting scans, viewing diagnostic results and history.

[0081] Use modern UI design software tools such as Sketch, Adobe XD, or Figma to design interface prototypes and collect feedback from your team and potential users for iterative optimization.

[0082] Develop interactive features: Front-end development: Develop the front-end interface according to the UI design drawings, and use HTML, CSS and JavaScript frameworks (such as React or Vue.js) to implement responsive design and adapt to different devices and screen sizes.

[0083] Backend integration: Ensure that the UI can communicate seamlessly with the backend server, transmit user requests and display processing results. The backend service needs to handle data storage, algorithm calls and result feedback.

[0084] Implementing a feedback system: Collect user feedback: Integrate feedback functionality into the interface so that users can easily report issues, make suggestions, or rate the app. This can be a simple feedback form or a more advanced interactive dialog box.

[0085] Data tracking and analysis: Use data analysis tools, such as Google Analytics, to track user behavior and understand data such as frequency of feature usage and user preferences, so as to continuously adjust and optimize the UI.

[0086] Test and Iterate: Conduct extensive user testing, including unit testing, integration testing, and user acceptance testing (UAT) to ensure the interface is easy to use, efficient, and bug-free.

[0087] Based on user feedback and test results, we continuously iterate products and optimize user experience.

[0088] Release and maintenance: After ensuring that the user interface and interactions meet high quality standards, the product can be released to the target user group.

[0089] The performance of the application must be monitored in real time, while user support services are provided to resolve any issues that users may encounter. The application must be updated regularly to introduce new features, improve the UI, and fix known issues.

[0090] By integrating a professional and friendly user interface with an effective feedback system, the final product can be made more in line with user needs, while continuous optimization through user feedback can maintain the vitality and competitiveness of the application.

[0091] For the key codes implemented in each step, the present invention will mainly use Python, and also include some commonly used libraries and frameworks, such as NumPy, Pandas, scikit-learn, and React and Node.js for front-end development. The following are the key code details for the mentioned steps: Model training, validation, and optimization (Python) 1. Checking Overfitting and Underfitting (Python) from sklearn.metrics import mean_squared_error from sklearn.model_selection import train_test_split from sklearn.ensemble import RandomForestRegressor # Data preparation X_train, X_val, y_train, y_val = train_test_split(X, y, test_size=0.2, random_state=42) # Model training model = RandomForestRegressor(n_estimators=100) model.fit(X_train, y_train) # predict train_predictions = model.predict(X_train) val_predictions = model.predict(X_val) # Evaluate train_error = mean_squared_error(y_train, train_predictions) val_error = mean_squared_error(y_val, val_predictions) print("Training error:", train_error) print("Validation error:", val_error) 2. Hyperparameter Tuning (Python using scikit-learn) from sklearn.model_selection import GridSearchCV # Set up the hyperparameter grid param_grid = { 'n_estimators': [50, 100, 200], 'max_features': ['auto', 'sqrt', 'log2'], 'max_depth' : [4,5,6,7,8] } # Initialize the GridSearchCV object CV_rfc = GridSearchCV(estimator=RandomForestRegressor(), param_grid=param_grid, cv= 5) # Perform a grid search CV_rfc.fit(X_train, y_train) # Best parameters and model print("Best parameters:", CV_rfc.best_params_) Step 6: User Interface and Feedback System (React + Node.js) 1. Front-end code example (React) import React, { useState} from 'react'; function App() { const [input, setInput] = useState(''); const handleInputChange = (event) =>{ setInput(event.target.value); }; const handleSubmit = () => { console.log('Submitted:', input); / / Here you can add the code sent to the server }; return ( <input type="text" value="{input}" onChange="{handleInputChange}"> <button onclick="{handleSubmit}">Submit< / button> ; } export default App; 2. Backend code example (Node.js) const express = require('express'); const bodyParser = require('body-parser'); const app = express(); app.use(bodyParser.json()); app.post(' / submit', (req, res) =>{ console.log('Received data:', req.body); res.send({ message: 'Data received successfully!'}); }); const PORT = 3000; app.listen(PORT, () => { console.log(`Server is running on http: / / localhost:${PORT}`); }); The above codes provide an implementation framework for the core functions of each stage, ensuring that the present invention can adjust and improve these functions according to the actual project needs.

[0092] In algorithm development and machine learning, there are many models to choose from for implementation, depending on the nature of the problem (such as classification, regression, or clustering) and the characteristics of the data. The following are several commonly used machine learning models, each of which is suitable for different types of machine learning tasks: 1. Linear Regression Suitable for regression problems, such as predicting numerical values ​​(such as price or sales).

[0093] Code example (Python using scikit-learn): from sklearn.linear_model import LinearRegression from sklearn.model_selection import train_test_split from sklearn.metrics import mean_squared_error # Data preparation X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2) # Model training model = LinearRegression() model.fit(X_train, y_train) # predict predictions = model.predict(X_test) # Evaluate mse = mean_squared_error(y_test, predictions) print('Mean Squared Error:', mse) 2. Decision Tree Applicable to classification and regression problems. It builds models by learning decision rules from data features to outputs.

[0094] Code example (Python using scikit-learn): from sklearn.tree import DecisionTreeClassifier from sklearn.metrics import accuracy_score # Data preparation and segmentation are the same as above # Model training tree_model = DecisionTreeClassifier() tree_model.fit(X_train, y_train) # predict predictions = tree_model.predict(X_test) # Evaluate accuracy = accuracy_score(y_test, predictions) print('Accuracy:', accuracy) 3. Random Forest It is an ensemble learning method suitable for classification and regression problems. It creates multiple decision trees and improves the accuracy and stability of the model through voting or averaging prediction methods.

[0095] Code example (Python using scikit-learn): from sklearn.ensemble import RandomForestClassifier # Data preparation and segmentation are the same as above # Model training forest_model = RandomForestClassifier(n_estimators=100) forest_model.fit(X_train, y_train) # predict predictions = forest_model.predict(X_test) # Evaluate accuracy = accuracy_score(y_test, predictions) print('Accuracy:', accuracy) 4. Support Vector Machine (SVM) Suitable for classification problems, especially when the data dimension is high.

[0096] Code example (Python using scikit-learn): from sklearn.svm import SVC # Data preparation and segmentation are the same as above # Model training svm_model = SVC(kernel='linear')# Different kernel functions can be selected svm_model.fit(X_train, y_train) # predict predictions = svm_model.predict(X_test) # Evaluate accuracy = accuracy_score(y_test, predictions) print('Accuracy:', accuracy) The choice of model is usually based on the understanding of the data and problem, as well as the model's performance and predictive accuracy. Typically, multiple models are tried at the beginning of implementation to verify which one is best suited to the current data and business needs. Cross-validation and hyperparameter tuning can also be applied to further improve model performance.

[0097] Although 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 the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for measuring atrial fibrillation based on low-voltage pulse measurement, It is characterized in that The specific steps include: S1. Sensor Assembly Attach the sensor to the wrist, fingertip or earlobe while ensuring that the sensor can be worn comfortably without interfering with daily activities; S2. Data collection and transmission Utilize installed sensors to continuously monitor and record, and design systems to continuously monitor and record low-voltage pulse signals, including key parameters of pulse waveform and frequency; S3. Signal processing and feature extraction Preprocess the signal data and use digital filtering technology to remove noise and interference in the signal, including removing artifacts and electromagnetic interference caused by movement. Then perform feature extraction to extract features related to arrhythmia in the pulse signal, such as irregularity of the RR interval and waveform variability, to extract useful information from the original pulse signal and convert it into structured data that can be used for subsequent machine learning model training. S4. Algorithm Development and Machine Learning Develop algorithms for analyzing low-voltage pulse signal characteristics, using time series analysis, frequency analysis methods, and then using machine learning models to identify signal patterns that are unique to atrial fibrillation, and continuously use new data to train and optimize the model to improve accuracy; S5. Verification and Optimization Cooperate with professional medical institutions to conduct clinical trials to verify the accuracy and reliability of the system, adjust algorithm parameters based on clinical feedback, optimize sensor design, and improve user experience and detection accuracy.

2. The method for measuring atrial fibrillation based on low-voltage pulse measurement according to claim 1, Features: In the step S1, the sensor assembly is preferably composed of any one or more combinations of photoplethysmography (PPG) sensors or microrheological pressure sensors.

3. The method for measuring atrial fibrillation based on low-voltage pulse measurement according to claim 1, Features: In the step S2 of data collection and transmission, data transmission is achieved through wireless connection, and the data is cached and stored, and encrypted transmission is also required during the data transmission.

4. The method for measuring atrial fibrillation based on low-voltage pulse measurement according to claim 1, Features: The signal processing and feature extraction in step S3 specifically includes signal denoising, pulse waveform detection, feature extraction of the transmitted model, and then standardization and normalization and data set construction, and finally preliminary graphic analysis.

5. The method for measuring atrial fibrillation based on low-voltage pulse measurement according to claim 1, Features: The machine learning model in the algorithm development and machine learning in step S4 is composed of any one or more combinations of support vector machines, random forests, and deep learning networks.

6. The method for measuring atrial fibrillation based on low-voltage pulse measurement according to claim 1, Features: After selecting the machine learning model in step S4 algorithm development and machine learning, model training is required, which specifically includes cross-validation, evaluation indicators and model persistence.

7. The method for measuring atrial fibrillation based on low-voltage pulse measurement according to claim 1, Features: The verification and optimization in step S5 specifically includes verifying the degree of freedom of the model, using independent data for verification, algorithm performance and accuracy testing, model optimization, and then providing result feedback, model updating, and retesting.