Multi-parameter fusion intelligent medical electronic application training platform

Abstract

The invention discloses a multi-parameter fusion intelligent medical electronic application training platform, which comprises a medical terminal device connected with a data acquisition and processing unit, and the data acquisition and processing unit is connected with a wireless communication module; the medical terminal equipment comprises a main controller, a multi-sensor module is connected with the main controller through a signal conditioning circuit, and the main controller is connected with a wireless communication module; the main controller receives data collected by the multi-sensor module and processes the collected data, and processing tasks comprise real-time signal filtering, simple feature extraction and data encryption; the load balancer receives data of the wireless communication module and distributes the data to the local monitoring server or the cloud server cluster; the Web service platform and the mobile terminal APP communicate with the cloud server cluster and are used for real-time data monitoring and data analysis. The modular design is adopted, multi-invention configuration and function extension are achieved, and the requirements of different levels of teaching and scientific research are met.

Description

Technical Field

[0001] The present invention relates to an intelligent medical electronic application innovation training platform, belonging to the technical field of the combination of medical electronics, embedded systems and health management education. Background Art

[0002] The present invention relates to an intelligent medical electronic application innovation training platform, belonging to the technical field of the combination of medical electronics, embedded systems and health management education. Summary of the invention

[0003] The technical problem to be solved by the present invention is: how to construct a multi-parameter integrated intelligent medical electronic application innovation training platform to achieve a comprehensive teaching and scientific research experimental environment.

[0004] The technical solution adopted by the present invention is: A multi-parameter fusion intelligent medical electronic application training platform, including: A medical terminal device, wherein the medical terminal device is connected to a data acquisition processing unit, and the acquisition processing unit is connected to a wireless communication module; The medical terminal device includes a main controller, the multi-sensor module is connected to the main controller through a signal conditioning circuit, and the main controller is connected to the wireless communication module; The main controller receives the data collected by the multi-sensor module and processes the collected data. The main controller acts as an edge computing node, and its processing tasks include real-time signal filtering, simple feature extraction, and data encryption. The load balancer receives data from the wireless communication module and distributes it to the local monitoring server or cloud server cluster; The Web service platform and mobile terminal APP communicate with the cloud server cluster respectively for real-time data monitoring and data analysis.

[0005] The aforementioned multi-parameter fusion intelligent medical electronic application training platform, the multi-sensor module includes a body temperature sensor, a blood pressure sensor, a blood oxygen sensor, a skin electrical sensor, and a motion posture sensor, which are connected to the main controller through a standard interface.

[0006] The aforementioned multi-parameter fusion intelligent medical electronic application training platform, the wireless communication module includes WiFi / 4G / 5G and Bluetooth modules, the main controller interacts with the local monitoring server and the mobile terminal through the WiFi or Bluetooth module; the main controller directly interacts with the mobile terminal through the WiFi or Bluetooth module, and the local monitoring server communicates wirelessly with the Web service platform and the mobile terminal APP through WiFi / 4G / 5G.

[0007] The aforementioned multi-parameter fusion intelligent medical electronic application training platform, the medical terminal equipment, the local monitoring server, and the Web service platform are all provided with a human-computer interaction module; In the human-computer interaction module of the medical terminal equipment, the numerical values ​​and dynamic curves of various physiological parameters are displayed in real time through a high-definition display screen using a variety of charts, and operation buttons or touch interfaces are set for parameter configuration, mode switching and system control; The human-computer interaction module of the local monitoring server includes a Web console for users to view patients' physiological data and health reports, and to display teaching experiment data and analysis results; In the human-computer interaction module of the Web service platform, various display contents of the remotely accessed local monitoring server are displayed.

[0008] The aforementioned multi-parameter fusion intelligent medical electronic application training platform includes the following modules in the main controller: Signal preprocessing module, used to perform noise filtering, baseline correction, and signal standardization on data collected by multiple sensors; Simple feature extraction module: used to extract simple features of different physiological parameters, including heart rate variability, blood oxygen saturation trend, and skin electrical response.

[0009] In the aforementioned multi-parameter fusion intelligent medical electronic application training platform, the following functional modules are set in the local monitoring server: Physiological parameter abnormality detection module: Set safety thresholds for various physiological parameters, identify and alarm abnormal physiological signals based on threshold judgment or machine learning models, and use trained machine learning models to identify abnormal physiological signals; Complex feature extraction module: used to extract complex features of different physiological parameters, including heart rate variability, blood oxygen saturation trend, and skin electrical response; Complex features include: Complex features of heart rate variability: features obtained using Lomb-Scargle periodogram analysis; Complex features of blood oxygen saturation: high-order features extracted by convolutional neural networks; fusion features combining time domain feature vectors and wavelet packet energy entropy; Complex characteristics of galvanic skin response: frequency domain bandwidth, maximum bandwidth energy ratio, and multi-scale nonlinear characteristics.

[0010] The aforementioned multi-parameter fusion intelligent medical electronic application training platform uses a dynamic frequency band division method to extract heart rate variability features and realizes dynamic reorganization of frequency bands through an adaptive power spectrum density threshold method, including: First, the RR interval sequence was resampled to 4 Hz and divided into frames, with 256 points per frame and 50% overlap, and then FFT transformation was performed after adding Hanning window; the RR interval sequence is the time limit between two R waves on the electrocardiogram; Adaptively merge adjacent frequency bands with power spectrum density ratio greater than 0.15 to achieve dynamic optimization of frequency bands; Use multi-scale sample entropy algorithm to enhance feature expression, including: (1) Coarse-graining the original time series, the scale factor From 3 to 7, each scale factor The coarse-grained sequence calculation formula is as follows:

[0011] Represents the first data points; is the index variable in the summation process, and its value range is arrive , used to traverse all data points in the current coarse-grained window; represents the first data points, The value range is ,in is the length of the coarse-grained sequence; Represents the total number of data points in the original time series; is the new sequence obtained after coarse-graining processing; (2) For each scale, the coarse-grained sequence obtained after the coarse-grained processing in step (1) is

[0012] According to the following structure Dimensional vector:

[0013] in, is the first data points; (3) Calculate each scale Sample entropy under :

[0014] is the value constructed in step (2) In the vector set, for each vector, all vectors whose distance to this vector is less than the similarity tolerance The number of vectors; The calculation formula of sample entropy is:

[0015] is the pattern dimension, is a similar tolerance.

[0016] In the aforementioned multi-parameter fusion intelligent medical electronic application training platform, in the dynamic frequency band division method, the power spectrum density ratio formula in the dynamic frequency band division is:

[0017] Frequency point The power spectral density at is the maximum value of the power spectral density, is the frequency point.

[0018] The aforementioned multi-parameter fusion intelligent medical electronic application training platform, in the process of extracting blood oxygen saturation trend features, includes: (1) Perform time domain feature extraction to obtain a time domain feature vector. Time domain feature extraction methods include the descending slope method and the area under the curve method. (2) Extract wavelet packet energy entropy features to obtain wavelet packet energy entropy; (3) The time domain feature vector and wavelet packet energy entropy are weighted feature fused to obtain the blood oxygen saturation trend feature. The feature fusion formula is expressed as:

[0019] is the time domain feature vector, is the wavelet packet energy entropy, is the weight coefficient, ranging from 0 to 1, To integrate the blood oxygen saturation trend characteristics.

[0020] The aforementioned multi-parameter fusion intelligent medical electronic application training platform, in the process of extracting skin electrical response features, includes: Extracting an optimal feature subset from the 28-dimensional feature pool, including time domain features, frequency domain features, and nonlinear features; the time domain features include SCR slope and SCL mean, the frequency domain features include 0.02-0.2 Hz energy ratio, and the nonlinear features include Katz fractal dimension and multi-scale entropy; The exponential cooling method was used to optimize the optimal feature subset, with an initial temperature of 100 °C, a cooling coefficient of 0.95, and a taboo table length of 15; The feature classification accuracy and feature dimension are then evaluated through the fitness function.

[0021] The aforementioned multi-parameter fusion intelligent medical electronic application training platform has the following functional modules set up in the cloud server cluster: Heart rate variability analysis module: extracts HRV indicators through time domain and frequency domain analysis of heart rate signals to evaluate heart health status; Blood oxygen trend analysis module: monitors the changing trend of blood oxygen saturation and identifies potential hypoxia risks; oxygen reduction event judgment conditions are set as: blood oxygen decline rate threshold , the lowest blood oxygen saturation threshold ; Emotion recognition module: Based on the characteristic changes of skin electrical response, the machine learning model is used to identify the user's emotional state, including: Select a machine learning algorithm model; The characteristic data set of skin electrical response is divided into a training set and a test set to train the machine learning algorithm model; Evaluate the performance of machine learning algorithm models through cross-validation and adjust hyperparameters to optimize the models; The trained model is used to classify the skin electrical response data collected in real time and output the user's current emotional state, including anxiety, relaxation or neutrality; Health risk assessment module: Combines data from multiple physiological parameters and uses a multivariate regression analysis model to assess the user's health risk. The health risk assessment process includes: Normalize multi-parameter physiological data; Integrate multiple physiological parameter features, including HRV, SpO2, and GSR, through a hierarchical attention mechanism; The risk index was calculated using a pre-trained multivariate regression model; Risks are graded according to the risk index, including low / medium / high risks. If medium / high risks occur, a risk warning will be triggered.

[0022] Personalized health advice module: Generates personalized health management advice and intervention measures based on health risk assessment results.

[0023] In the aforementioned multi-parameter fusion intelligent medical electronic application training platform, a mobile terminal App is provided on the mobile terminal, and the mobile terminal App includes the following functional modules: Teaching experiment modules include: Database, used to store basic knowledge of embedded systems, including basic concepts and development environment. Basic concepts include real-time operating system, task scheduling, semaphore, message queue, and common communication protocols; development environment includes common programming languages ​​and development tools; store measurement data information received from different sensors; Development environment setting module: used to select different development environments, including programming languages ​​and development tools; Embedded Programming Module: used to provide a complete software development environment, including: Code editor: used to complete basic code numbering and editing, check syntax errors in real time, and complete code folding and unfolding; Project file management module: create project files, import and export project files, and manage source code files, configuration files, and library files; Online compiler: Use the ARM-GCC tool chain to compile and generate firmware files, and provide real-time prompts when compilation errors occur; Multi-task demonstration module: Create and manage multiple tasks in RTOS through example demonstration, showing the principle and implementation method of multi-task collaboration; Experimental module: used to configure the communication protocols of WiFi and Bluetooth, realize data transmission between different simulated hardware, and show the steps and principles of wireless data interaction; The data algorithm module sets up a variety of algorithm units, including signal filtering, feature extraction and machine learning algorithms, and completes and displays the corresponding calculation results by calling the algorithm units, such as calling the Kalman filter unit for signal smoothing, and calling the decision tree or support vector machine unit for data classification.

[0024] The aforementioned multi-parameter fusion intelligent medical electronic application training platform, the mobile terminal App includes the following functional modules: Data query module: used to view the values ​​and dynamic changes of various physiological parameters in real time and playback historical data; Heart rate variability analysis module: extracts HRV indicators through time domain and frequency domain analysis of heart rate signals to evaluate heart health status; Blood oxygen trend analysis module: monitors the changing trend of blood oxygen saturation and identifies potential hypoxia risks; Emotion recognition module: Based on the characteristic changes of skin electrical response, the machine learning model is used to identify the user's emotional state, including: Select support vector machine or neural network as the machine learning algorithm model; The characteristic data set of skin electrical response is divided into a training set and a test set to train the machine learning algorithm model; Evaluate the performance of machine learning algorithm models through cross-validation and adjust hyperparameters to optimize the models; The trained model is used to classify the skin electrical response data collected in real time and output the user's current emotional state, including anxiety, relaxation or neutrality; Health risk assessment module: Combines data from multiple physiological parameters and uses a multivariate analysis model to assess the user's health risk; Personalized health advice module: Generates personalized health management recommendations and intervention measures based on health risk assessment results, and provides health guidance.

[0025] The beneficial effects achieved by the present invention are as follows: the multi-parameter fusion intelligent medical electronic application training platform of the present invention uses a variety of sensors to collect a variety of physiological parameters to achieve comprehensive monitoring of human health status, uses a variety of data processing algorithms and models to detect abnormal data and assess health risks, and uses wireless communication methods such as WiFi and Bluetooth to achieve remote transmission, storage and visual display of data. It adopts a modular design to achieve multiple invention configurations and functional expansions to meet the needs of teaching and scientific research at different levels. It is equipped with a high-definition display screen and a variety of interactive interfaces to display measurement results and data charts in real time for easy operation and analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a system framework diagram of the multi-parameter fusion intelligent medical electronic application training platform in Example 1 of the present invention; Figure 2 This is a schematic diagram of the structure of a multi-sensor module in Example 1 of the present invention; Figure 3 This is an oxygen reduction event detection diagram in Example 1 of the present invention; Figure 4 This is a flow chart of a dynamic frequency band division method in Embodiment 1 of the present invention; Figure 5 This is a schematic diagram of the interface data summary and analysis of the human-computer interaction module in Example 1 of the present invention. DETAILED DESCRIPTION

[0027] In order to better illustrate the technical features of the present invention, the following drawings are provided. The drawings are only schematic illustrations and do not constitute a limitation to the present invention.

[0028] Example 1 like Figure 1 As shown, this embodiment provides a multi-parameter fusion intelligent medical electronic application training platform, including: A medical terminal device, wherein the medical terminal device is connected to a data acquisition and processing unit, and the acquisition and processing unit is connected to a wireless communication module; a plurality of medical terminal devices may be provided; The medical terminal device includes a main controller, the multi-sensor module is connected to the main controller through a signal conditioning circuit, and the main controller is connected to the wireless communication module; The main controller receives the data collected by the multi-sensor module and processes the collected data. The main controller acts as an edge computing node, and its processing tasks include real-time signal filtering, simple feature extraction, and data encryption. The load balancer receives data from the wireless communication module and distributes it to the local monitoring server or cloud server cluster to ensure the reliability and real-time performance of data transmission; The Web service platform and mobile terminal APP communicate with the cloud server cluster respectively for real-time data monitoring and data analysis.

[0029] like Figure 2 As shown, the multi-sensor module includes a body temperature sensor, a blood pressure sensor, a blood oxygen sensor, a galvanic skin sensor (GSR), a motion posture sensor, etc., and is connected to the main controller through a standard interface; The body temperature sensor adopts a contact or non-contact temperature sensor for real-time body temperature monitoring; The blood pressure sensor is used for accurate blood pressure measurement; The blood oxygen sensor uses a photoplethysmography (PPG) method to measure blood oxygen saturation (SpO2) and pulse rate in real time; The galvanic skin sensor (GSR) reflects emotional fluctuations and stress status by measuring changes in skin conductance; The motion posture sensor is used to monitor the posture and motion trajectory of the human body.

[0030] The multi-sensor module and the wireless communication module are connected through a standardized communication interface, and an expansion interface is reserved to support function upgrades. The communication interface includes SPI, I²C, ADC, UART, etc.; The wireless communication module includes WiFi / 4G / 5G and Bluetooth modules. The main controller interacts with the local monitoring server and mobile terminal data through WiFi or Bluetooth modules; the main controller directly interacts with the mobile terminal (such as mobile phone, tablet) through WiFi or Bluetooth modules, and the local monitoring server communicates with the Web service platform and mobile terminal APP wirelessly through WiFi / 4G / 5G.

[0031] In the wireless communication module, standardized communication protocols are used for data transmission, including MQTT, HTTP, TCP / IP, etc., to ensure the reliability and real-time performance of data communication. The lightweight MQTT protocol is used to achieve efficient publish / subscribe data transmission, which is suitable for real-time monitoring and large-scale device management. The HTTP RESTful API is used to achieve two-way data interaction with the cloud platform or mobile terminal.

[0032] Optimize the parameter settings of the TCP / IP protocol stack, such as adjusting the maximum transmission unit (MTU) and the receiving window size to adapt to the data transmission requirements in different network environments; use the priority management mechanism to enable high-priority data packets to be sent first, thereby improving real-time performance. The above optimization methods for the TCP / IP protocol stack on the embedded platform can effectively improve the data transmission stability and efficiency of the embedded platform and ensure the reliable operation of the system in a complex network environment.

[0033] Medical terminal equipment, local monitoring servers, and Web service platforms are all equipped with human-computer interaction modules.

[0034] In the human-computer interaction module of the medical terminal equipment, the high-definition display screen uses a variety of charts to display the values ​​and dynamic curves of various physiological parameters in real time, including line charts, bar charts, pie charts, etc. Set operation buttons or touch interfaces for parameter configuration, mode switching and system control.

[0035] The human-computer interaction module of the local monitoring server includes a Web console for medical staff to view patients' physiological data and health reports, and to display teaching experiment data and analysis results; like Figure 5 As shown, in the human-computer interaction module of the Web service platform, various display contents of the remotely accessed local monitoring server are displayed.

[0036] Through the human-computer interaction modules at three levels, a complete user interaction interface of the multi-terminal collaborative system is realized, allowing users to easily obtain and manage health information in medical institutions, homes or educational environments The medical terminal device is provided with a power supply module, including a voltage stabilization module, a power protection module and an energy management module, to supply power to the medical terminal device.

[0037] The main controller includes an STM32H743 chip, which integrates a DC / DC step-down circuit to stabilize the power supply, has an SD card holder for data storage, a built-in RTC clock module to maintain accurate time, and a USB OTG interface for flexible data transmission and device control. The onboard ST-LINK debugger is convenient for programming download and online debugging with STM32CubeIDE. At the same time, an LCD driver interface is provided to support display connection, and multiple IO expansion interfaces are provided to access more peripheral devices.

[0038] The STM32H743 chip includes an STM32H7 main control board, and the basic adapter board is connected to the STM32H7 main control board for experimental teaching and project development. The basic adapter board includes DAC output, independent buttons, 4x4 button matrix, SHT30 temperature and humidity sensor, light sensor, WS2812 driver, digital tube driver, servo control, ultrasonic module, MPU module, and DC motor drive unit. The DAC output can be used for analog signal output; the independent button provides a separate function trigger mode; the 4x4 button matrix can realize a variety of button combination operations; the SHT30 temperature and humidity sensor can monitor environmental parameters in real time; the light sensor can detect the intensity of surrounding light; the WS2812 driver is used to control the color change of the LED light ring; the digital tube driver is responsible for displaying digital information; the servo control module can accurately adjust the angle of the robotic arm; the ultrasonic module is used to measure the distance; the MPU module realizes motion posture perception; the DC motor drive unit can control the direction and speed of the DC motor. The 4.3-inch display module is connected to the STM32H7 main control board. STM32H743 can be used for hardware design and software programming, thereby improving the user's practical and innovative abilities.

[0039] The main controller includes the following modules: The signal preprocessing module is used to perform the following preprocessing on the data collected by multiple sensors: Noise filtering: Use low-pass filtering, high-pass filtering or band-pass filtering to remove high-frequency noise and low-frequency drift in the signal; Baseline correction: perform baseline correction on physiological signal data to eliminate sensor drift and environmental interference; Signal standardization: Standardize the data collected by different sensors to facilitate subsequent feature extraction and analysis.

[0040] Simple feature extraction module: used to extract simple features of different physiological parameters, including heart rate variability (HRV), blood oxygen saturation (SpO2) trend, galvanic skin response (GSR), etc. Simple characterization of physiological parameters include: Simple characteristics of heart rate variability: maximum, minimum, mean, median in the time domain; heart rate mean and heart rate standard deviation.

[0041] Simple characteristics of blood oxygen saturation: mean value, standard deviation; area under the curve; mean frequency, peak frequency, center frequency.

[0042] Simple characteristics of galvanic skin response: time domain peak, mean, variance, standard deviation; waveform factor, peak factor, pulse factor; skewness factor, margin factor.

[0043] The following functional modules are set in the local monitoring server: Physiological parameter abnormality detection module: Set safety thresholds for various physiological parameters, identify and alarm abnormal physiological signals based on threshold judgment or machine learning models. The machine learning models include support vector machines, decision trees or neural networks, etc., and use trained support vector machines (SVM), neural networks and other models to identify abnormal physiological signals.

[0044] Complex feature extraction module: used to extract complex features of different physiological parameters, including heart rate variability, blood oxygen saturation trend, skin electrical response, etc. Complex features include: Heart rate variability complex features: features obtained using Lomb-Scargle periodogram analysis.

[0045] Complex features of blood oxygen saturation: high-order features extracted by convolutional neural networks; fusion features combining time domain feature vectors and wavelet packet energy entropy.

[0046] Complex characteristics of galvanic skin response: frequency domain bandwidth, maximum bandwidth energy ratio; multi-scale nonlinear characteristics.

[0047] like Figure 4 As shown, a dynamic frequency band division method is used to extract heart rate variability features, and an adaptive power spectrum density threshold method is used to achieve dynamic reorganization of frequency bands, including: First, the RR interval sequence was resampled to 4 Hz and divided into frames, with 256 points per frame and 50% overlap, and then FFT transformation was performed after adding Hanning window; the RR interval sequence is the time limit between two R waves on the electrocardiogram; Adaptively merge adjacent frequency bands with power spectrum density ratio greater than 0.15 to achieve dynamic optimization of frequency bands; The multi-scale sample entropy algorithm is used to enhance feature expression, with the scale factor range of 3-7, the pattern dimension of 2, and the similarity tolerance of 0.15 times the standard deviation, including: (1) Coarse-graining the original time series, the scale factor From 3 to 7, each scale factor The coarse-grained sequence calculation formula is as follows:

[0048] Represents the first Data points are values ​​collected sequentially from the signal to be analyzed; is the index variable in the summation process, and its value range is arrive , used to traverse all data points in the current coarse-grained window; represents the first data points, The value range is ,in is the length of the coarse-grained sequence, usually the total number of original data points Divide by the scale factor Then take the integer part; Represents the total number of data points in the original time series; is the new sequence obtained after coarse-graining processing.

[0049] (2) For each scale, the coarse-grained sequence obtained after the coarse-grained processing in step (1) is

[0050] According to the following structure Dimensional vector:

[0051] in, is the first data points are used to capture the dynamic change characteristics of data at different scales and provide vectorized data support for subsequent distance calculation and sample entropy solution.

[0052] (3) Calculate each scale Sample entropy under :

[0053] is the value constructed in step (2) In the vector set, for each vector, all vectors whose distance to this vector is less than the similarity tolerance The number of vectors; Using the above statistical results, the calculation formula of sample entropy is:

[0054] is the model dimension, the value is 2, is the similarity tolerance, 0.15σ is taken, and σ is the standard deviation. Peacekeeping The complexity of the time series is measured by the similarity change in the dimensional vector.

[0055] The formula for power spectrum density ratio in dynamic frequency band division is:

[0056] Frequency point The power spectral density at , in ms² / Hz, is the maximum value of the power spectral density, in ms² / Hz. is the frequency point, .

[0057] The process of extracting blood oxygen saturation trend features includes: (1) Perform time domain feature extraction to obtain a time domain feature vector. Time domain feature extraction methods include the descending slope method and the area under the curve method. (2) Extract wavelet packet energy entropy features to obtain wavelet packet energy entropy; (3) The time domain feature vector and wavelet packet energy entropy are weighted feature fused to obtain the blood oxygen saturation trend feature. By fusing the time domain feature vector and wavelet packet energy entropy, the time domain change characteristics and frequency domain energy distribution information of the signal can be obtained simultaneously, which improves the feature expression ability and classification performance and realizes the accurate quantification of blood oxygen trend. The weight coefficients can be set to 0.6 and 0.4 respectively. The feature fusion formula is expressed as:

[0058] is the time domain feature vector, is the wavelet packet energy entropy, is the weight coefficient, ranging from 0 to 1, To integrate the blood oxygen saturation trend characteristics.

[0059] In the process of extracting skin electrical response features, a method combining taboo search and simulated annealing is used, including: The optimal feature subset is extracted from the 28-dimensional feature pool, including time domain features, frequency domain features, and nonlinear features; the time domain features include SCR slope and SCL mean, the frequency domain features include 0.02-0.2 Hz energy ratio, and the nonlinear features include Katz fractal dimension and multi-scale entropy.

[0060] The exponential cooling method was used to optimize the optimal feature subset, with an initial temperature of 100 °C, a cooling coefficient of 0.95, and a taboo table length of 15; The feature classification accuracy and feature dimension are then evaluated through the fitness function.

[0061] The 28-dimensional features are composed of multiple statistical features extracted from the skin electrical signal, including: Time domain characteristics: peak value, mean value, variance, standard deviation, kurtosis factor, root mean square, waveform factor, peak factor, skewness factor, pulse factor, margin factor; Frequency domain characteristics: bandwidth, maximum bandwidth energy ratio, average amplitude value, average short-time energy; Nonlinear features: wavelet packet energy spectrum, sample entropy, approximate entropy, Katz fractal dimension, multiscale entropy.

[0062] The determination of the above-mentioned feature dimensions is based on a comprehensive analysis of the skin electrical signals in the time domain, frequency domain and nonlinear characteristics. Experiments have verified that these features have good discriminability for emotion recognition.

[0063] The following functional modules are set up in the cloud server cluster: Heart rate variability analysis module: extracts HRV indicators through time domain and frequency domain analysis of heart rate signals to evaluate heart health status; Blood oxygen trend analysis module: monitors the changing trend of blood oxygen saturation and identifies potential hypoxia risks; The oxygen reduction event judgment condition is set as: blood oxygen drop rate threshold , the lowest blood oxygen saturation threshold , to ensure reliable identification of hypoxic events.

[0064] like Figure 3 As shown in Figure 3, blood oxygen trend analysis adopts a three-level judgment mechanism, including initial decline detection (ΔSpO2 ≥ 2% / s), continuous verification (≥ 10s) and morphological verification (conforming to the exponential decay model).

[0065] Emotion recognition module: Based on the characteristic changes of skin electrical response, the machine learning model is used to identify the user's emotional state, including: Choose a machine learning algorithm model, including support vector machine (SVM) or neural network; The characteristic data set of skin electrical response is divided into a training set and a test set to train the machine learning algorithm model; Evaluate the performance of machine learning algorithm models through cross-validation and adjust hyperparameters to optimize the models; The trained model is used to classify the skin electrical response data collected in real time and output the user's current emotional state, including anxiety, relaxation or neutrality; By adopting the model fusion method, the prediction results of multiple models are weighted averaged or voted to obtain the final emotion recognition result, which can further improve the recognition accuracy.

[0066] By training the above-mentioned machine learning model, accurate identification of the user's emotional state can be achieved, providing data support for subsequent health management and intervention.

[0067] Health risk assessment module: Combines data from multiple physiological parameters and uses a multivariate regression analysis model to assess the user's health risk. The health risk assessment process includes: Normalize multi-parameter physiological data; Integrate multiple physiological parameter features, including HRV, SpO2, GSR, etc., through a hierarchical attention mechanism; The risk index was calculated using a pre-trained multivariate regression model; Risks are graded according to the risk index, including low / medium / high risks. If medium / high risks occur, a risk warning will be triggered.

[0068] Personalized health advice module: Generates personalized health management recommendations and intervention measures based on health risk assessment results, and provides health guidance.

[0069] The intelligent medical electronic application training platform of the present invention realizes a complete data flow from terminal collection to cloud storage, meets the needs of various application scenarios such as medical monitoring and teaching experiments, and stores, analyzes and visualizes the collected data.

[0070] Data is uploaded from the sensors and main controllers of medical terminal devices to cloud servers or local monitoring servers in real time. Data cleaning and preprocessing are performed in the main controller to remove noise and outliers. Statistical analysis methods are used to analyze data to identify potential health trends and patterns. Predictive analysis is performed through machine learning models to assess users' health risks and generate corresponding health reports. Chart tools are used to present the analysis results in graphical form, including line charts, bar charts, and heat maps, so that users can intuitively understand data changes and health status.

[0071] The present invention improves data processing efficiency and system response speed by distributing computing tasks to multiple nodes. In order to achieve multi-terminal collaboration, a RESTful API interface is designed so that each human-computer interaction module and mobile terminal application can access and control data in real time. Users can view real-time monitoring data, historical records and health reports through the human-computer interaction interface of the mobile terminal App or Web service platform. At the same time, it supports multi-user collaboration and information sharing. Whether in a medical institution, home or educational environment, users can easily obtain and manage health information to achieve more efficient health management and decision support.

[0072] It also includes a mobile terminal, on which a mobile terminal App is arranged. The mobile terminal App supports Android, iOS and Hongmeng systems to realize data access and control of multiple terminals.

[0073] The mobile terminal App includes the following functional modules: Teaching experiment modules include: Database, used to store basic knowledge of embedded systems, including basic concepts and development environment. Basic concepts include real-time operating system, task scheduling, semaphore, message queue, common communication protocols such as UART, SPI, I2C, etc., sensor measurement principle; development environment includes common programming languages ​​and development tools such as Keil, CubeIDE, etc.; store the measurement data information received from different sensors for students to learn; Development environment setting module: used to select different development environments, including programming languages ​​and development tools; Embedded Programming Module: used to provide a complete software development environment, including: Code editor: used to complete basic code numbering and editing, check syntax errors in real time, and complete code folding and unfolding; Project file management module: create project files, import and export project files, and manage source code files, configuration files, and library files; Online compiler: Use the ARM-GCC tool chain to compile and generate firmware files, and provide real-time prompts when compilation errors occur; Program download module: use JTAG / SWD interface serial port ISP download.

[0074] The teaching experiment module helps students master the complete process of embedded software development and improve their practical ability by providing a complete development tool chain.

[0075] Multi-task demonstration module: Create and manage multiple tasks in RTOS through example demonstration, showing the principle and implementation method of multi-task collaboration; Experimental module: used to configure the communication protocols of WiFi and Bluetooth, realize data transmission between different simulated hardware, and show the steps and principles of wireless data interaction; For example, in setting up an experiment, the user implements data interaction between the main controller and the sensor through the UART protocol; The data algorithm module sets up a variety of algorithm units, including signal filtering, feature extraction and machine learning algorithms, and completes and displays the corresponding calculation results by calling the algorithm units, such as calling the Kalman filter unit for signal smoothing, and calling the decision tree or support vector machine unit for data classification.

[0076] The algorithm addition module is used to add machine learning and deep learning algorithms, and to run and verify the algorithms completed by students, which can improve students' data analysis and model development capabilities.

[0077] Through the above modules, students can apply the knowledge they have learned when completing specific projects. For example, they can design a health monitoring system and let students be responsible for collecting data from sensors, implementing data processing algorithms, and finally visualizing the data, so as to comprehensively improve their comprehensive capabilities. Students can deeply understand all aspects of embedded software development, master multi-task management, communication protocol implementation and data processing algorithms.

[0078] Example 2 The mobile terminal App includes the following functional modules: Data query module: used to view the values ​​and dynamic changes of various physiological parameters in real time and playback historical data; Heart rate variability analysis module: extracts HRV indicators through time domain and frequency domain analysis of heart rate signals to evaluate heart health status; Blood oxygen trend analysis module: monitors the changing trend of blood oxygen saturation and identifies potential hypoxia risks; The oxygen reduction event judgment condition is set as: blood oxygen drop rate threshold , minimum blood oxygen saturation threshold , to ensure reliable identification of hypoxic events.

[0079] like Figure 3 As shown in Figure 3, blood oxygen trend analysis adopts a three-level judgment mechanism, including initial decline detection (ΔSpO2 ≥ 2% / s), continuous verification (≥ 10s) and morphological verification (conforming to the exponential decay model).

[0080] Emotion recognition module: Based on the characteristic changes of skin electrical response, the machine learning model is used to identify the user's emotional state, including: Choose machine learning algorithm models such as support vector machines or neural networks; The characteristic data set of skin electrical response is divided into a training set and a test set to train the machine learning algorithm model; Evaluate the performance of machine learning algorithm models through cross-validation and adjust hyperparameters to optimize the models; The trained model is used to classify the skin electrical response data collected in real time and output the user's current emotional state, including anxiety, relaxation or neutrality; The recognition accuracy can be further improved by adopting the model fusion method to obtain the final result by weighted averaging or voting the prediction results of multiple models.

[0081] By training the above-mentioned machine learning model, accurate identification of the user's emotional state can be achieved, providing data support for subsequent health management and intervention.

[0082] Health risk assessment module: Combines data from multiple physiological parameters and uses a multivariate analysis model to assess the user's health risk; Personalized health advice module: Generates personalized health management recommendations and intervention measures based on health risk assessment results, and provides health guidance.

[0083] Other technical features are the same as those of Example 1.

[0084] The above description is only a preferred embodiment of the present invention and does not constitute a limitation on the protection scope of the present invention. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the claims of the present invention.

Claims

1. A multi-parameter fusion intelligent medical electronic application training platform, characterized in that: include: A medical terminal device, wherein the medical terminal device is connected to a data acquisition processing unit, and the acquisition processing unit is connected to a wireless communication module; The medical terminal device includes a main controller, the multi-sensor module is connected to the main controller through a signal conditioning circuit, and the main controller is connected to the wireless communication module; The main controller receives the data collected by the multi-sensor module and processes the collected data. The processing tasks include real-time signal filtering, simple feature extraction, and data encryption. The load balancer receives data from the wireless communication module and distributes it to the local monitoring server or cloud server cluster; The Web service platform and mobile terminal APP communicate with the cloud server cluster respectively for real-time data monitoring and data analysis.

2. According to claim 1, a multi-parameter fusion intelligent medical electronic application training platform is characterized in that: The multi-sensor module includes a body temperature sensor, a blood pressure sensor, a blood oxygen sensor, a skin electrical sensor, and a motion posture sensor, and is connected to the main controller through a standard interface.

3. According to claim 1, a multi-parameter fusion intelligent medical electronic application training platform is characterized in that: The wireless communication module includes WiFi / 4G / 5G and Bluetooth modules. The main controller interacts with the local monitoring server and the mobile terminal through the WiFi or Bluetooth module; the main controller directly interacts with the mobile terminal through the WiFi or Bluetooth module, and the local monitoring server communicates with the Web service platform and the mobile terminal APP through WiFi / 4G / 5G.

4. According to claim 1, a multi-parameter fusion intelligent medical electronic application training platform is characterized in that: Medical terminal equipment, local monitoring servers, and Web service platforms are all equipped with human-computer interaction modules; In the human-computer interaction module of the medical terminal equipment, the numerical values ​​and dynamic curves of various physiological parameters are displayed in real time through a high-definition display screen using a variety of charts, and operation buttons or touch interfaces are set for parameter configuration, mode switching and system control; The human-computer interaction module of the local monitoring server includes a Web console for users to view patients' physiological data and health reports, and to display teaching experiment data and analysis results; In the human-computer interaction module of the Web service platform, various display contents of the remotely accessed local monitoring server are displayed.

5. According to claim 1, a multi-parameter fusion intelligent medical electronic application training platform is characterized in that: The main controller includes the following modules: Signal preprocessing module, used to perform noise filtering, baseline correction, and signal standardization on data collected by multiple sensors; Simple feature extraction module: used to extract simple features of different physiological parameters, including heart rate variability, blood oxygen saturation trend, and skin electrical response.

6. The multi-parameter fusion intelligent medical electronic application training platform according to claim 1 is characterized in that: The following functional modules are set in the local monitoring server: Physiological parameter abnormality detection module: Set safety thresholds for various physiological parameters, identify and alarm abnormal physiological signals based on threshold judgment or machine learning models, and use trained machine learning models to identify abnormal physiological signals; Complex feature extraction module: used to extract complex features of different physiological parameters, including heart rate variability, blood oxygen saturation trend, and skin electrical response; Complex features include: Complex features of heart rate variability: features obtained using Lomb-Scargle periodogram analysis; Complex features of blood oxygen saturation: high-order features extracted by convolutional neural networks; fusion features combining time domain feature vectors and wavelet packet energy entropy; Complex characteristics of galvanic skin response: frequency domain bandwidth, maximum bandwidth energy ratio, and multi-scale nonlinear characteristics.

7. The multi-parameter fusion intelligent medical electronic application training platform according to claim 6 is characterized in that: The dynamic frequency band division method is used to extract the heart rate variability features, and the adaptive power spectrum density threshold method is used to achieve dynamic reorganization of the frequency band, including: First, the RR interval sequence was resampled to 4 Hz and divided into frames, with 256 points per frame and 50% overlap, and then FFT transformation was performed after adding Hanning window; the RR interval sequence is the time limit between two R waves on the electrocardiogram; Adaptively merge adjacent frequency bands with power spectrum density ratio greater than 0.15 to achieve dynamic optimization of frequency bands; Use multi-scale sample entropy algorithm to enhance feature expression, including: (1) Coarse-graining the original time series, the scale factor From 3 to 7, each scale factor The coarse-grained sequence calculation formula is as follows: ; Represents the first data points; is the index variable in the summation process, and its value range is arrive , used to traverse all data points in the current coarse-grained window; represents the first data points, The value range is ,in is the length of the coarse-grained sequence; Represents the total number of data points in the original time series; is the new sequence obtained after coarse-graining processing; (2) For each scale, the coarse-grained sequence obtained after the coarse-grained processing in step (1) is ; According to the following structure Dimensional vector: ; in, is the first data points; (3) Calculate each scale Sample entropy under : ; is the value constructed in step (2) In the vector set, for each vector, all vectors whose distance to this vector is less than the similarity tolerance The number of vectors; The calculation formula of sample entropy is: ; is the pattern dimension, is a similar tolerance.

8. The multi-parameter fusion intelligent medical electronic application training platform according to claim 7 is characterized in that: In the dynamic frequency band division method, the formula for the power spectrum density ratio in dynamic frequency band division is: ; Frequency point The power spectral density at is the maximum value of the power spectral density, is the frequency point.

9. The multi-parameter fusion intelligent medical electronic application training platform according to claim 6 is characterized in that: The process of extracting blood oxygen saturation trend features includes: (1) Perform time domain feature extraction to obtain a time domain feature vector. Time domain feature extraction methods include the descending slope method and the area under the curve method. (2) Extract wavelet packet energy entropy features to obtain wavelet packet energy entropy; (3) The time domain feature vector and wavelet packet energy entropy are weighted feature fused to obtain the blood oxygen saturation trend feature. The feature fusion formula is expressed as: ; is the time domain feature vector, is the wavelet packet energy entropy, is the weight coefficient, ranging from 0 to 1, It is the fusion blood oxygen saturation trend feature.

10. The multi-parameter fusion intelligent medical electronic application training platform according to claim 6 is characterized in that: In the process of extracting skin electrical response features, it includes: Extracting an optimal feature subset from the 28-dimensional feature pool, including time domain features, frequency domain features, and nonlinear features; the time domain features include SCR slope and SCL mean, the frequency domain features include 0.02-0.2 Hz energy ratio, and the nonlinear features include Katz fractal dimension and multi-scale entropy; The exponential cooling method was used to optimize the optimal feature subset, with an initial temperature of 100 °C, a cooling coefficient of 0.95, and a taboo table length of 15; The feature classification accuracy and feature dimension are then evaluated through the fitness function.

11. The multi-parameter fusion intelligent medical electronic application training platform according to claim 1 is characterized in that: The following functional modules are set up in the cloud server cluster: Heart rate variability analysis module: extracts HRV indicators through time domain and frequency domain analysis of heart rate signals to evaluate heart health status; Blood oxygen trend analysis module: monitors the changing trend of blood oxygen saturation and identifies potential hypoxia risks; oxygen reduction event judgment conditions are set as: blood oxygen decline rate threshold , the lowest blood oxygen saturation threshold ; Emotion recognition module: Based on the characteristic changes of skin electrical response, the machine learning model is used to identify the user's emotional state, including: Select a machine learning algorithm model; The characteristic data set of skin electrical response is divided into a training set and a test set to train the machine learning algorithm model; Evaluate the performance of machine learning algorithm models through cross-validation and adjust hyperparameters to optimize the models; The trained model is used to classify the skin electrical response data collected in real time and output the user's current emotional state, including anxiety, relaxation or neutrality; Health risk assessment module: Combines data from multiple physiological parameters and uses a multivariate regression analysis model to assess the user's health risk. The health risk assessment process includes: Normalize multi-parameter physiological data; Integrate multiple physiological parameter features, including HRV, SpO2, and GSR, through a hierarchical attention mechanism; The risk index was calculated using a pre-trained multivariate regression model; Risks are graded according to the risk index, including low / medium / high risk. If medium / high risk occurs, a risk warning is triggered; Personalized health advice module: Generates personalized health management advice and intervention measures based on health risk assessment results.

12. The multi-parameter fusion intelligent medical electronic application training platform according to claim 1 is characterized in that: The mobile terminal is provided with a mobile terminal App, which includes the following functional modules: Teaching experiment modules include: Database, used to store basic knowledge of embedded systems, including basic concepts and development environment. Basic concepts include real-time operating system, task scheduling, semaphore, message queue, and common communication protocols; development environment includes common programming languages ​​and development tools; store measurement data information received from different sensors; Development environment setting module: used to select different development environments, including programming languages ​​and development tools; Embedded Programming Module: used to provide a complete software development environment, including: Code editor: used to complete basic code numbering and editing, check syntax errors in real time, and complete code folding and unfolding; Project file management module: create project files, import and export project files, and manage source code files, configuration files, and library files; Online compiler: Use the ARM-GCC tool chain to compile and generate firmware files, and provide real-time prompts when compilation errors occur; Multi-task demonstration module: Create and manage multiple tasks in RTOS through example demonstration, showing the principle and implementation method of multi-task collaboration; Experimental module: used to configure the communication protocols of WiFi and Bluetooth, realize data transmission between different simulated hardware, and show the steps and principles of wireless data interaction; The data algorithm module sets up a variety of algorithm units, including signal filtering, feature extraction and machine learning algorithms, and completes and displays the corresponding calculation results by calling the algorithm units, such as calling the Kalman filter unit for signal smoothing, and calling the decision tree or support vector machine unit for data classification.

13. According to the multi-parameter fusion intelligent medical electronic application training platform of claim 12, the mobile terminal App includes the following functional modules: Data query module: used to view the values ​​and dynamic changes of various physiological parameters in real time and playback historical data; Heart rate variability analysis module: extracts HRV indicators through time domain and frequency domain analysis of heart rate signals to evaluate heart health status; Blood oxygen trend analysis module: monitors the changing trend of blood oxygen saturation and identifies potential hypoxia risks; Emotion recognition module: Based on the characteristic changes of skin electrical response, the machine learning model is used to identify the user's emotional state, including: Select support vector machine or neural network as the machine learning algorithm model; The characteristic data set of skin electrical response is divided into a training set and a test set to train the machine learning algorithm model; Evaluate the performance of machine learning algorithm models through cross-validation and adjust hyperparameters to optimize the models; The trained model is used to classify the skin electrical response data collected in real time and output the user's current emotional state, including anxiety, relaxation or neutrality; Health risk assessment module: Combines data from multiple physiological parameters and uses a multivariate analysis model to assess the user's health risk; Personalized health advice module: Generates personalized health management recommendations and intervention measures based on health risk assessment results, and provides health guidance.

Images