Medical detection device based on big data analysis and control method thereof

Abstract

The invention relates to the technical field of medical detection, in particular to a medical detection device based on big data analysis and a control method thereof, and the medical detection device comprises a distributed sensor network, an edge computing node and a cloud server. The device adjusts the contribution degree of a multi-modal data source in real time through a dynamic weight distribution algorithm, optimizes health assessment in combination with a deep reinforcement learning model, and generates a personalized intervention strategy. The multi-modal health data can be fused in real time, the abnormal health state can be accurately recognized, accurate medical resource matching is provided through the cross-platform interaction system, and the efficiency and accuracy of health monitoring and intervention are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of medical detection technology, and in particular to a medical detection device based on big data analysis and a control method thereof. Background Art

[0002] Currently, in existing technologies, medical detection devices have certain limitations in data collection and integration. For example, traditional equipment mostly focuses on single-function monitoring (such as blood pressure, heart rate, etc.), resulting in insufficient correlation analysis capabilities for cross-modal health information. In addition, most devices rely on manual operation and offline data analysis, and there is a certain delay in real-time dynamic monitoring and immediate feedback. For example, routine oral health monitoring usually requires patients to visit the hospital multiple times for examinations, which is difficult to meet high-frequency needs.

[0003] Some solutions are disclosed in the prior art, such as:

[0004] CN117637158A discloses a big data-based oral health monitoring device that implements telemedicine through cloud-based diagnosis. However, the data interaction efficiency of this solution is affected by network latency, and it does not incorporate a real-time learning algorithm to optimize model performance.

[0005] CN114557713A discloses a medical detection device and control method based on big data analysis, which can integrate multiple physiological indicator detection functions. However, this solution does not solve the problems of multi-source data fusion and dynamic health assessment, limiting its application in complex scenarios.

[0006] The above existing technologies have certain limitations in data fusion and dynamic analysis. For example, the traditional method uses a fixed weight distribution strategy to process multimodal data, which can be expressed as:

[0007]

[0008] Among them, W i represents the weight of the i-th category of data, and n is the total number of data sources. This fixed weighting method cannot be dynamically adjusted according to the patient's health status, resulting in limited accuracy of the analysis results.

[0009] In addition, existing health assessment models mostly rely on static rules or traditional machine learning algorithms and lack dynamic optimization mechanisms. For example, traditional risk prediction models use the following formula:

[0010]

[0011] Among them, RR represents risk value, ω i and x iare the weight and measurement value of the i-th indicator respectively. Due to the fixed parameters of the model, its ability to identify abnormal patterns is weak, especially when facing complex health data.

[0012] When it comes to personalized health interventions, existing technologies mostly provide standardized recommendations, lacking customized services that integrate with individual health records. For example, while hemodialysis equipment can detect leaks, it doesn't integrate with patients' long-term health data to optimize treatment plans. In existing methods, health intervention strategies are typically based on preset thresholds, using the following formula:

[0013]

[0014] Here, S represents the intervention signal, V is the detection value, and T is the preset threshold. This simple threshold determination method cannot meet the needs of precision medicine.

[0015] Therefore, there is an urgent need for an innovative solution that can integrate multimodal data in real time, dynamically optimize analysis models, and provide personalized health intervention strategies to improve the intelligence level and service capabilities of medical detection devices. Summary of the Invention

[0016] This invention relates to a medical detection device and control method based on big data analysis, designed to integrate multimodal health data in real time and dynamically optimize health assessment models. Specifically, the device achieves millisecond-level data acquisition through a collaborative architecture of edge computing and the cloud. It then incorporates a deep reinforcement learning algorithm to dynamically model a patient's health status, ultimately generating personalized health intervention strategies.

[0017] The main purpose of the present invention is to provide a medical detection device and a control method thereof, which can integrate multimodal health data in real time, dynamically optimize health assessment models, and generate personalized health intervention strategies.

[0018] The purpose of the present invention can be achieved by the following methods:

[0019] First, a multimodal data real-time fusion system is constructed using a collaborative architecture of edge computing and the cloud. This system collects physiological parameters (such as blood pressure and heart rate), imaging data (such as intraoral images), and environmental information (such as temperature and humidity) through a distributed sensor network. Distributed sensors transmit data to edge computing nodes via wireless communication modules. The edge computing nodes perform preliminary processing on the data before uploading it to the cloud server.

[0020] Secondly, a dynamic weight allocation algorithm is designed to adjust the contribution of different data sources in real time according to the patient's health status. Specifically, the algorithm calculates the dynamic weight of each data source using the following formula:

[0021]

[0022] Among them, W i (t) represents the weight of the i-th type of data at time t, d i (t) represents the deviation between the data in category II and the current health status, and α is the adjustment coefficient. Through this algorithm, the system can automatically increase the weight of key data sources. For example, for diabetic patients, the weight of blood sugar data can be automatically increased to 70%, and other indicators can be dynamically adapted.

[0023] Next, a dynamic health assessment model based on deep reinforcement learning is constructed, using a "detection-feedback-optimization" closed-loop system to model health status in real time. The model updates parameters using the following formula:

[0024] Q(s t ,a t )←Q(s t ,a t )+η·[r t +γ·max a Q(s t+1 ,a)-Q(s t , a t )]

[0025] Among them, S t Indicates the current health status, a t Indicates the action taken, r t Denotes the reward value, η and γ are the learning rate and discount factor, respectively. Using this model, the system can generate risk level predictions within 10 seconds, with an accuracy improvement of 23% compared to traditional models.

[0026] Finally, a cross-platform interactive health intervention system was designed, integrating medical testing devices with mobile terminals and hospital information systems (HIS) to provide personalized recommendations across multiple terminals. The system uses natural language processing (NLP) to generate patient-readable "health prescriptions" and integrates with Guangzhou's "Smart Equipment Cloud" platform to automatically match nearby medical resources. Specifically, the system generates intervention priorities using the following formula:

[0027]

[0028] Where P represents the intervention priority, ω i and f i are the weight and score of the i-th health indicator, g j represents the availability score of the jth type of resource. Through this formula, the system can prioritize and recommend the most appropriate medical resources.

[0029] A method for controlling a medical detection device employs a distributed sensor network, edge computing nodes, and a cloud server. The method is characterized in that the distributed sensor network includes physiological parameter sensors, image data collectors, and environmental information monitors, and scans a target area at a predetermined frequency to collect multimodal data from the scanned area, including physiological parameters, image data, and environmental information. When the device enters an operating state, the medical detection device is confirmed to be in an on state. The control method for the medical detection device is as follows:

[0030] Step 1: First, obtain multimodal data values of all areas of the first environment through the distributed sensor network, that is, obtain data values of all areas of the environment state at the first time, calculate the average value of the data values of all scanning points as the environmental baseline value, and use this average value as the mathematical expectation value of the environmental baseline value:

[0031]

[0032] Step 2: Obtain the numerical points (x1, x2…xn) of the key data sources through query, that is, all numerical points greater than the environmental benchmark value, and calculate the average value of the sum of the squares of the differences between the numerical points of the key data sources and the average value. That is, to obtain all key data sources

[0033] Step 3: Obtain multimodal data values of all areas of the second environment through distributed sensor network scanning, and calculate the mean value M1 and variance as in steps 1 and 2.

[0034] Step 4: When the second environment is obtained Above the first environment When a certain threshold is reached, the health status is judged to be abnormal, and the system automatically adjusts relevant parameters, such as intervention priority and resource matching.

[0035] Furthermore, a data correction value △D is added in step 2, specifically:

[0036] Step 2: Obtain the numerical points (x1, x2…xn) of the key data sources through query, that is, all numerical points greater than the environmental benchmark value, and calculate the average value of the sum of the squares of the differences between the numerical points of the key data sources and the average value. That is, to find the variance of all key data sources

[0037]

[0038] Among them, △D is the data correction value.

[0039] Furthermore, the distributed sensor network scans and outputs 1024 pixels, and calculates an average value M of the data values of the 1024 pixels as a mathematical expectation value of the environmental reference value.

[0040] A medical detection device includes a distributed sensor network, an edge computing node, and a cloud server. The distributed sensor network collects multimodal health data and transmits the collected data to the edge computing node. The edge computing node performs preliminary data processing and uploads the processed data to the cloud server. The cloud server performs in-depth analysis and generates corresponding control commands based on the analysis results. The control commands are transmitted to the terminal device to change the terminal device's working state.

[0041] The present invention has the following advantages:

[0042] The medical detection device control method provided by this invention uses a dynamic weight allocation algorithm to adjust the contribution of different data sources in real time, more accurately reflecting changes in a patient's health status. By comparing the variance of two measurements, the system can accurately identify health abnormalities and generate personalized health intervention strategies.

[0043] In order to make the above and other objects, features and advantages of the present invention more obvious and easy to understand, preferred embodiments are given below and described in detail with reference to the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 This is a schematic diagram of the overall architecture of the medical detection device of the present invention, showing the connection relationship between the distributed sensor network, edge computing nodes and cloud servers.

[0046] Figure 2 This is a structural diagram of the distributed sensor network in the present invention, showing the distribution of physiological parameter sensors, image data collectors and environmental information monitors and their data flow.

[0047] Figure 3 This is a flow chart of the dynamic weight allocation algorithm of the present invention, which describes in detail how to adjust the contribution of different data sources in real time according to the patient's health status.

[0048] Figure 4This is a working principle diagram of the dynamic health assessment model based on deep reinforcement learning in the present invention, showing the operating mechanism of the "detection-feedback-optimization" closed-loop system.

[0049] Figure 5 This is a logical framework diagram of the cross-platform interactive health intervention system of the present invention, which embodies the collaborative working mode of the medical detection device, mobile terminal and hospital information system (HIS).

[0050] Figure 6 This is a flow chart of multimodal data processing in the control method of the present invention, including complete steps from data acquisition to abnormality judgment.

[0051] Figure 7 Schematic diagram of variance comparison for abnormal health status judgment in an embodiment of the present invention, marking the changing trends of the baseline variance and the real-time variance. DETAILED DESCRIPTION

[0052] The present invention provides a medical detection device and control method based on big data analysis. The core of the device is to achieve real-time collection, processing, and analysis of multimodal health data through the collaborative work of a distributed sensor network, edge computing nodes, and cloud servers. The following describes specific embodiments of the present invention in detail with reference to the accompanying drawings.

[0053] like Figure 1 As shown in Figure 1, the overall architecture of a medical detection device consists of a distributed sensor network, edge computing nodes, and cloud servers. The distributed sensor network includes physiological parameter sensors, image data collectors, and environmental information monitors. These sensors are distributed at different locations within the target area to collect multimodal data. Physiological parameter sensors primarily acquire physiological indicators such as blood pressure and heart rate; image data collectors capture intraoral images or other medical images; and environmental information monitors record environmental parameters such as temperature, humidity, and light. These sensors transmit data to edge computing nodes via wireless communication modules. As the first layer of data processing, edge computing nodes perform preliminary cleaning and compression on the raw data before uploading it to the cloud server. The cloud server further performs in-depth data analysis, generates corresponding control commands, and ultimately feeds the results back to the terminal device to change its operating state.

[0054] In a distributed sensor network, the distribution of sensors and data flow are as follows: Figure 2As shown. Physiological parameter sensors are usually installed on key parts of the patient's body, such as the wrist or chest, to ensure that physiological signals can be accurately captured. Image data collectors are deployed on specific medical equipment, such as endoscopes or dental scanners, to obtain high-resolution images in real time. Environmental information monitors are generally placed in the main areas where patients move, such as wards or home environments. These sensors are connected to the data aggregation node via wired or wireless means. The node is responsible for integrating and transmitting the data from all sensors to the edge computing node. In this process, the flow of data is unidirectional, that is, from the sensor to the edge computing node, ensuring efficient data transmission.

[0055] The dynamic weight allocation algorithm is one of the core technologies of the present invention, and its implementation process is as follows: Figure 3 As shown in the figure, the system first calculates the deviation value d_i(t) for each data source based on the patient's current health status. The deviation value is calculated as d_i(t) = |x_i(t) - x_ref|, where x_i(t) represents the actual measured value of the i-th data type at time t, and x_ref is the reference value. The system then calculates the dynamic weight W_i(t) for each data source using the formula W_i(t) = exp(-α·d_i(t)) / Σ_j = 1^n exp(-α·d_j(t)), where α is the adjustment coefficient used to control the sensitivity of the weight adjustment. For example, when a diabetic patient's blood sugar level deviates from the normal range, the weight of the blood sugar data is automatically increased to 70%, while the weights of other indicators are correspondingly reduced. This dynamic weight allocation mechanism can significantly improve the contribution of key data sources, thereby more accurately reflecting changes in the patient's health status.

[0056] The working principle of the dynamic health assessment model based on deep reinforcement learning is as follows Figure 4 As shown. The model uses a "detection-feedback-optimization" closed-loop system to model the patient's health status in real time. First, the system collects multimodal data through a distributed sensor network and inputs it into the health assessment model. The model selects an action a_t based on the current health status s_t and calculates the reward value r_t. The reward value is calculated as r_t=f(s_t,a_t), where f is the reward function used to evaluate the effectiveness of the action. Subsequently, the model uses the Q-learning algorithm to update the parameters, with the formula Q(s_t,a_t)←Q(s_t,a_t)+η·[r_t+γ·max_aQ(s_t+1,a)-Q(s_t,a_t)], where η is the learning rate and γ is the discount factor. Through continuous iteration, the model can generate a risk level prediction within 10 seconds, and the prediction accuracy is 23% higher than that of the traditional model.

[0057] The logical framework of the cross-platform interactive health intervention system is as follows Figure 5As shown. The system integrates medical detection devices, mobile terminals and hospital information systems (HIS), and generates patient-readable "health prescriptions" through natural language processing (NLP). The system first calculates the intervention priority P based on the results of the health assessment model 7. The formula is P = Σ_i = 1^mw_i·f_i / √(Σ_j = 1^k g_j^2), where w_i and f_i are the weights and scores of the i-th health indicators, respectively, and g_j represents the availability score of the j-th resource. Subsequently, the system automatically matches the nearest medical resources through the Guangzhou "Smart Machinery Cloud" platform and sends intervention recommendations to the patient's mobile terminal. For example, when the system detects that a patient has a risk of hypertension, it will give priority to recommending nearby community hospitals and generate personalized health prescriptions containing dietary recommendations, exercise plans, and other content.

[0058] Multimodal data processing process Figure 6 As shown in the figure, a distributed sensor network scans the target area at a constant frequency, collecting multimodal data values from the scanned area. This data includes physiological parameters, imaging data, and environmental information. The system then calculates the average value M_0 of the data from all scanned points as the mathematical expectation of the environmental baseline value. The formula is M_0 = Σ_i = 1^n x_i / n, where x_i is the data value of the i-th scan point and n is the total number of scan points. Next, the system queries the key data sources for numerical points (x_1, x_2, …, x_n) and calculates the average value S_baseline^2 of the sum of the squares of the differences between these numerical points and the average value, using the formula S_baseline^2 = Σ_i = 1^n(M-x_i)^2 / n. In the second environment, the system repeats the above steps to calculate the new average value M_1 and variance S_moment^2. When S_moment^2 exceeds a certain threshold of S_baseline^2, the system determines that the health status is abnormal and performs relevant parameter adjustments.

[0059] In some embodiments, a data correction value ΔD is introduced to improve the accuracy of data processing. The specific correction process is as follows: First, the system still calculates the average value S_base^2 of the sum of the squares of the differences between the numerical points of the key data source and the average value according to the above steps, but multiplies this value by the data correction value ΔD. The formula is: The data correction value ΔD ranges from 0.8 to 1.2, and the specific value is adjusted according to the actual application scenario. For example, in scenarios with high environmental noise, ΔD can be set to 0.9 to reduce the impact of noise on the data; while in medical scenarios requiring high precision, ΔD can be set to 1.1 to enhance data sensitivity.

[0060] In practical applications, the output of a distributed sensor network is usually the data value of 1024 pixels. The system calculates the average value M of the data values of these 1024 pixels as the mathematical expectation value of the environmental reference value. Subsequently, the system calculates the variance S_reference^2 and S_moment^2 according to the aforementioned steps, and determines whether the health status is abnormal by comparing the difference between the two. For example, when a patient uses the device at home, the system collects his heart rate, blood pressure, and indoor temperature and humidity data through a distributed sensor network, and calculates the reference variance S_reference^2. If the patient suddenly has an abnormal heart rate, the real-time variance S_moment^2 collected by the system will be significantly higher than the reference variance, such as Figure 7 At this point, the system automatically triggers the health intervention mechanism, generates a personalized intervention strategy, and notifies the patient and their family.

[0061] The embodiments of the present invention also include optimization solutions for different scenarios. For example, in the oral health monitoring scenario, the image data collector can be configured as a high-resolution camera to capture images of the oral cavity in real time. The system determines whether there are diseases such as caries or periodontitis by analyzing features such as tooth color and gum condition in the image. At the same time, the physiological parameter sensor collects the patient's saliva pH value and oral temperature, and the environmental information monitor records the indoor air quality. Through the dynamic weight allocation algorithm, the system can adjust the weight of each data source according to the patient's specific situation, thereby generating more accurate health assessment results.

[0062] In another embodiment, the present invention is applied to a chronic disease management scenario. The system builds a personalized health profile by monitoring the patient's blood sugar, blood pressure, weight and other indicators over a long period of time. When the blood sugar level of a diabetic patient is continuously high, the system will not only increase the weight of the blood sugar data, but also combine historical data to predict future trends and generate targeted intervention recommendations. For example, the system may recommend that patients reduce their carbohydrate intake and increase their daily exercise time. At the same time, the system sends recommendations to the patient's mobile terminal through a cross-platform interactive health intervention system, and interacts with the hospital information system (HIS) to remind doctors to pay attention to changes in the patient's condition.

[0063] The present invention's embodiments have also demonstrated its applicability in complex scenarios. For example, in scenarios where multiple users are using the system simultaneously, the system distinguishes the health data of different users through a distributed sensor network. Each user's data is accompanied by a unique identifier, ensuring data accuracy and privacy. Furthermore, the system supports multi-terminal collaboration, allowing patients to view health reports on their mobile phones while doctors access detailed data on their computers. This multi-terminal collaboration significantly improves the system's flexibility and practicality.

[0064] The specific embodiments of the present invention fully disclose the technical details of the medical detection device and its control method, including the composition, connection, and operating principles of its components. As can be seen from the above examples, the present invention can effectively address the problems of insufficient data fusion and weak dynamic analysis capabilities in the existing technology, providing strong technical support for precision medicine.

[0065] In order to better enable relevant personnel in this technical field to fully understand and implement the present invention, the specific implementation principle of the present invention is further supplemented below with reference to a specific application scenario.

[0066] In home health monitoring scenarios, patients use medical testing devices to collect and analyze daily health data. First, physiological parameter sensors in a distributed sensor network are worn on the patient's wrist to collect real-time physiological indicators such as heart rate and blood pressure. Simultaneously, environmental information monitors are deployed in the patient's bedroom and living room, continuously recording environmental parameters such as temperature, humidity, and light intensity. An image data collector is installed on a dental scanner to regularly capture intraoral image data. These sensors transmit the collected data to an edge computing node via wireless communication modules. The edge computing node, deployed in the patient's home, receives data from the distributed sensor network and performs preliminary cleaning and compression processing to reduce data redundancy and noise interference. The processed data is then uploaded to a cloud server for further in-depth analysis.

[0067] During data processing, the system adjusts the contribution of each data source according to the dynamic weight allocation algorithm. For example, when the blood pressure value measured by a patient in the morning is significantly higher than the normal range, the system will calculate the deviation value of the blood pressure data according to the deviation value formula di(t) = |xi(t)-xref| and use the formula

[0068]

[0069] Dynamically increase the weight of blood pressure data. In this scenario, the weight of blood pressure data might increase from an initial 20% to 70%, while the weight of other data sources decreases accordingly. This dynamic adjustment mechanism ensures that key data dominates health assessments, thereby more accurately reflecting changes in a patient's health status.

[0070] The dynamic health assessment model based on deep reinforcement learning runs in the cloud server and models multimodal data in real time. The system first collects the patient's heart rate, blood pressure and environmental temperature and humidity data through a distributed sensor network and inputs them into the health assessment model 7. The model calculates the patient's health status based on the current health status S. t Select action a t , and calculate the reward value r tThe reward value is calculated as rt = f(st,at), where f is the reward function used to evaluate the effectiveness of the action. Subsequently, the model uses the Q-learning algorithm to update the parameters, as follows:

[0071] Q(s t , a t )←Q(s t ,a t )+η·[r t +γ·max a Q(s t+1 ,a)-Q(s t ,a t )]

[0072] Through continuous iteration, the model can generate a risk level prediction within 10 seconds. For example, if a patient's blood pressure and heart rate both deviate from the normal range, the system will determine that they are at risk of cardiovascular disease and generate a corresponding risk level prediction result.

[0073] The cross-platform interactive health intervention system provides patients with personalized health intervention strategies through the Guangzhou "Smart Equipment Cloud" platform and the hospital information system (HIS). The system first calculates the intervention priority P based on the results of the health assessment model 7, and the formula is

[0074] Among them, ω i and f i are the weight and score of the i-th health indicator, g j represents the availability score of the jth resource. The system then automatically matches nearby medical resources and generates a personalized health prescription that includes dietary recommendations, exercise plans, and more. For example, if the system detects a patient at risk for hypertension, it will prioritize a nearby community hospital and send health intervention recommendations to the patient via mobile device.

[0075] In actual operation, the system uses a multimodal data processing process to determine abnormal health status. First, the distributed sensor network scans the target area at a certain frequency and collects multimodal data values of the scanned area, including physiological parameters, imaging data, and environmental information. Then, the system calculates the average value M0 of the data of all scanning points as the mathematical expectation value of the environmental baseline value, which is calculated as follows:

[0076]

[0077] Next, the system queries the numerical points (x1, x2, ..., xn) of the key data sources and calculates the average value of the sum of the squares of the differences between these numerical points and the average value. To calculate, the formula is

[0078] In the second environment, the system repeats the above steps to calculate the new mean value M1 and variance when Higher than When a certain threshold is reached, the system determines that the health status is abnormal and triggers the health intervention mechanism. For example, when the patient's heart rate suddenly increases at home, the real-time variance collected by the system Significantly higher than the benchmark variance At this time, the system automatically generates personalized intervention strategies and notifies patients and their families through mobile terminals.

[0079] In addition, in some high-precision medical scenarios, the system introduces a data correction value ΔD to improve the accuracy of data processing. The specific correction process is as follows: the system still calculates the average value of the sum of squares of the difference between the numerical points of the key data sources and the average value according to the above steps. But on this basis, multiply the data correction value ΔD, the formula is

[0080]

[0081] The data correction value ΔD ranges from 0.8 to 1.2, and the specific value is adjusted according to the actual application scenario. For example, in scenarios with high environmental noise, ΔD can be set to 0.9 to reduce the impact of noise on the data; while in medical scenarios requiring high precision, ΔD can be set to 1.1 to enhance data sensitivity.

[0082] Through the above steps, the present invention effectively addresses existing issues such as insufficient data fusion and weak dynamic analysis capabilities, providing strong technical support for precision medicine. Furthermore, the system's multi-terminal collaborative model significantly enhances flexibility and practicality, ensuring that patients and doctors can easily access health data and intervention recommendations.

[0083] The above are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention description and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A medical detection device control method based on big data analysis, which utilizes a distributed sensor network, edge computing nodes, and cloud servers, and is characterized by: The distributed sensor network includes physiological parameter sensors, image data collectors, and environmental information monitors. It scans the target area at a certain frequency and collects multimodal data of the scanned area, including physiological parameters, image data, and environmental information. When the device enters the working state, it is confirmed that the medical detection device is in the turned-on state. The control method of the medical detection device is as follows: Step 1: Obtain multimodal data values of all areas of the first environment through a distributed sensor network, that is, obtain data values of all areas of the environment state at the first time, calculate the average value of the data values of all scanning points as the environmental baseline value, and use this average value as the mathematical expectation value of the environmental baseline value: Step 2: Obtain the numerical points (x1, x2…xn) of the key data sources through query, that is, all numerical points greater than the environmental benchmark value, and calculate the average value of the sum of the squares of the differences between the numerical points of the key data sources and the average value. That is, to obtain all key data sources Step 3: Obtain multimodal data values of all areas of the second environment through distributed sensor network scanning, and calculate the mean value M1 and variance as in steps 1 and 2. Step 4: When the second environment is obtained Above the first environment When a certain threshold is reached, it is judged that the health status is abnormal and the system automatically adjusts the relevant parameters.

2. The medical detection device control method according to claim 1, characterized in that: In step 2, the data correction value △D is added, specifically: Step 2: Obtain the numerical points (x1, x2…xn) of the key data sources through query, that is, all numerical points greater than the environmental benchmark value, and calculate the average value of the sum of the squares of the differences between the numerical points of the key data sources and the average value. That is, to find the variance of all key data sources Among them, △D is the data correction value.

3. The medical detection device control method according to claim 1 or 2, characterized in that: The distributed sensor network scans and outputs 1024 pixels, and calculates an average value M of the data values of the 1024 pixels as a mathematical expectation value of the environmental reference value.

4. A medical detection device for implementing the medical detection device control method according to any one of claims 1 to 3, characterized in that: It includes a distributed sensor network, edge computing nodes and cloud servers. The distributed sensor network collects multimodal health data and transmits the collected data to the edge computing nodes. The edge computing nodes perform preliminary processing of the data and upload the processed data to the cloud server. The cloud server performs in-depth analysis and generates corresponding control commands based on the analysis results. The control commands are transmitted to the terminal device to change the working state of the terminal device.

5. The medical detection device according to claim 4, characterized in that: The distributed sensor network includes physiological parameter sensors, image data collectors and environmental information monitors. The physiological parameter sensors are used to collect blood pressure and heart rate data of patients, the image data collectors are used to collect medical image data, and the environmental information monitors are used to record temperature, humidity and light data.

6. The medical detection device according to claim 4, characterized in that: The edge computing node performs preliminary cleaning and compression on the original data and then uploads it to the cloud server.

7. The medical detection device according to claim 4, characterized in that: The cloud server uses a dynamic weight allocation algorithm to adjust the contribution of different data sources in real time. The dynamic weight allocation algorithm calculates the dynamic weight of each data source using the following formula: Among them, W i (t) represents the weight of the i-th type of data at time t, d i (t) represents the deviation between the i-th type of data and the current health status, and α is the adjustment coefficient.

8. The medical detection device according to claim 4, characterized in that: The cloud server builds a dynamic health assessment model based on deep reinforcement learning, and uses the "detection-feedback-optimization" closed-loop system to model the health status in real time. The model updates the parameters through the following formula: Q(s t , a t )←Q(s t ,a t )+η·[r t +γ·max a Q(s t+1 ,a)-Q(s t ,a t )] Among them, S t Indicates the current health status, a t Indicates the action taken, r t represents the reward value, η and γ are the learning rate and discount factor respectively.

9. The medical detection device according to claim 4, characterized in that: The cloud server designs a cross-platform interactive health intervention system that integrates medical detection devices with mobile terminals and hospital information systems, and generates intervention priorities using the following formula: Where P represents the intervention priority, ω i and f i are the weight and score of the i-th health indicator, g j Represents the availability score of the j-th resource.

Images