Multi-disease screening system based on multi-modal data and large model

Abstract

The invention relates to the field of medical services, is used for solving the problems that chronic diseases of old people are high in morbidity, and prevention, intervention and treatment management are difficult to carry out on a large scale by means of a traditional manual mode, and particularly relates to a multi-disease screening system based on multi-modal data and a large model. According to the invention, the data port laying scheme is utilized to predict and analyze the putting setting of the public equipment, the detection equipment can be reasonably deployed according to the number of patients or high-risk crowds in the area, the utilization rate of the equipment is improved, and through the cooperation of the multi-modal data acquisition array and the dynamic port laying scheme, the detection efficiency is improved. Automatic directional collection of crowd health data is realized, detection items can be intelligently matched according to different crowd characteristics, and multi-dimensional health indexes are generated by fusing large model analysis with historical disease control information and utilizing model analysis, so that deviation of a single data source is effectively avoided, and the detection accuracy is improved. Therefore, the screening accuracy is finally improved, and the period from data acquisition to risk assessment is remarkably shortened.

Description

Technical Field

[0001] This invention relates to the field of medical services, specifically to a multi-disease screening system based on multimodal data and large models. Background Art

[0002] At present, the problem of home-based elderly care is prominent, and the refined management of elderly patients with chronic diseases is imminent. There is an urgent need to improve the standardized management rate of patients with hypertension and diabetes, and the control rate of blood pressure and blood sugar in the managed population. Relying solely on "traditional manual" methods can no longer meet the actual needs of chronic disease management. It is necessary to alleviate the problem of insufficient manpower through information technology and provide refined health management services for elderly patients with chronic diseases. Therefore, the key objectives at present are to establish an efficient and accurate integrated medical and elderly care health management system that enables "early detection, early intervention, and early treatment" for the elderly in the region, to achieve health management for the entire population throughout their life cycle, to improve the service capacity of integrated medical and elderly care, to promote the integrated development of medical and health services with home-based and institutional elderly care, and to advance high-quality medical and elderly care services.

[0003] In response to the above technical problems, this application proposes a solution. Summary of the Invention

[0004] In this invention, the automated and targeted collection of population health data is achieved through the synergy of a multimodal data acquisition array and a dynamic port deployment scheme. It can intelligently match detection items based on the characteristics of different population groups and ensure comprehensive data coverage by comparing the completion rate in real time via the cloud. The large-scale model analysis is integrated with historical disease management information, and multidimensional health indicators are generated using model analysis, effectively avoiding the bias of a single data source. This ultimately improves screening accuracy and significantly shortens the cycle from data collection to risk assessment, addressing the problem of high incidence of chronic diseases among the elderly and the difficulty in large-scale prevention, intervention, and treatment management using traditional manual methods. Therefore, a multi-disease screening system based on multimodal data and a large-scale model is proposed.

[0005] The purpose of the present invention can be achieved through the following technical solutions: A multi-disease screening system based on multimodal data and large models includes a multimodal data acquisition array, a data port deployment module, a large model tracking and analysis module, a change risk early warning module, and a management terminal interaction module. The data port deployment module is used to acquire the population that needs to be collected and the types of data to be collected, and to make a data port deployment plan based on the population information, and at the same time send the data port deployment plan to the multimodal data acquisition array. The multimodal data acquisition array acquires data from the terminal through the data port, compares the acquired results with the set acquisition requirements, uploads the acquired terminal data to the cloud, and generates the acquisition completion rate based on the comparison results. The large-scale model tracking and analysis module acquires terminal data through the cloud, and simultaneously acquires disease control information corresponding to the identity. It then performs identity matching based on the terminal data, and uses the large-scale model to simulate and analyze the identity-matched terminal data and disease control information to obtain patient health indicators. The change risk early warning module obtains the patient's health indicators through the large model tracking and analysis module, and generates treatment follow-up information based on the patient's health indicators; The management terminal interaction module acquires treatment follow-up information and patient health indicators, regenerates the patient's disease risk level, and outputs it through the network. At the same time, it sends the data port laying module according to the disease risk level, and the data port laying module performs data port laying reconstruction.

[0006] In a preferred embodiment of the present invention, the data port deployment module obtains the patient population with chronic diseases through the cloud, and at the same time obtains the distribution of the patient population in a set area through information reconstruction technology. Based on the distribution of the patient population, weight simulation is performed to obtain the patient distribution weights in different areas. The data port deployment module categorizes the data to be collected into individual data and centralized data. Individual data is collected using personal devices, while centralized data is collected using shared devices. The data port deployment module generates the distribution density of shared devices based on the patient distribution weights and obtains the data port deployment plan.

[0007] In a preferred embodiment of the present invention, the method for the data port deployment module to obtain the distribution density of public equipment is as follows: The data port deployment module obtains the preset radiation range and radiation quantity of each public device. The data port deployment module obtains the number of patients based on the patient distribution weight, and calculates the distribution quantity by using the number of patients, the radiation quantity of each public device, and the adjustment coefficient. At the same time, the data port deployment module analyzes the radiation range of each public device using a preset algorithm to obtain the distribution location of public devices when the coverage reaches a set threshold, and records the distribution location and distribution quantity of public devices as the public device distribution density.

[0008] In a preferred embodiment of the present invention, the terminal data collected by the multimodal data acquisition array includes exercise intensity, heart rate curve, blood glucose variation and blood pressure data, wherein blood pressure data is collected through public equipment, while exercise intensity, heart rate curve and blood glucose variation are collected through personal equipment, and the collection frequency of public equipment is lower than that of personal equipment. The multimodal data acquisition array records each piece of terminal data acquired as a sample data. The multimodal data acquisition array obtains the sample threshold for each data category, and when the number of acquired sample data exceeds the set sample threshold, it records that the acquisition requirement has been completed, and packages the sample data and uploads it to the cloud. When the number of sample data does not reach the set sample threshold, the ratio of the number of samples to the set sample threshold is recorded as the collection completion rate. The multimodal data acquisition array compares the collection completion rate of each data category and records the data category with the lowest collection completion rate as the abnormal category.

[0009] In a preferred embodiment of the present invention, the large model tracking and analysis module takes exercise intensity, heart rate curve, blood glucose variation and blood pressure data as input parameters. When performing simulation analysis, the large model tracking and analysis module compares the exercise intensity with the set exercise range, the heart rate curve with the set heart rate parameter, the blood glucose variation with the set blood glucose parameter, and the blood pressure data with the set blood pressure range. When the output parameters match the set range and parameters, it is recorded as exponential coincidence. When it exceeds the set range, the proportion of the excess is recorded as exponential deviation. The large model tracking and analysis module statistically analyzes index deviation and index overlap, performs normalization processing to obtain index samples, and calculates an arithmetic mean of all health index samples to obtain the final patient health index.

[0010] In a preferred embodiment of the present invention, after the change risk warning module obtains the patient's health indicators, it will simultaneously obtain the patient's disease risk level through the cloud, and perform weight calculation based on the patient's disease risk level and the patient's health indicators. The weight calculation result will be compared with a set threshold to obtain a follow-up deterioration indicator or a follow-up normal indicator, and the follow-up deterioration indicator or the follow-up normal indicator will be recorded as treatment follow-up information.

[0011] In a preferred embodiment of the present invention, the management terminal interaction module expands the patient's health indicators through treatment follow-up information, compares the expanded patient health indicators with the set threshold levels to obtain the health change level, and regenerates the new patient disease risk level based on the patient's existing disease risk level and health change level.

[0012] In a preferred embodiment of the present invention, after obtaining the disease risk level, the management terminal interaction module sends the disease risk level to the data port deployment module. The data port deployment module adds the disease risk level to the patient distribution weight, regenerates the new patient distribution weight, and obtains a new data port deployment scheme based on the new patient distribution weight.

[0013] Compared with the prior art, the present invention has the following beneficial effects: In this invention, compared with the traditional single-point detection mode, the automated and targeted collection of population health data is achieved through the synergy of multimodal data acquisition array and dynamic port deployment scheme. It can intelligently match detection items according to the characteristics of different populations, and compare the collection completion rate in real time through the cloud to ensure the comprehensiveness of data coverage. The large model analysis is integrated with historical disease control information, and multidimensional health indicators are generated by model analysis, which effectively avoids the bias of a single data source. In the end, it achieves the effect of improving the accuracy of screening and significantly shortening the cycle from data collection to risk assessment.

[0014] In this invention, based on the health indicators output by a large model, a treatment follow-up plan is automatically generated using a risk warning method. This solves the problem of traditional screening that emphasizes detection but neglects follow-up. Furthermore, the disease risk level of different patients is regenerated in real time through interaction with the management terminal, thereby dynamically optimizing the detection strategy. This shifts disease intervention from passive response to proactive prevention, significantly reducing the conversion rate to severe illness and forming a cyclical assessment and management plan of evaluation, warning, correction, and follow-up evaluation.

[0015] In this invention, the deployment and setting of public equipment is predicted and analyzed using a data port deployment scheme. The detection equipment can be deployed reasonably according to the number of patients or high-risk groups in the area, avoiding idle or overloaded hardware resources. This can adapt to large-scale public health screening scenarios in residential areas, thereby solving the problem of insufficient efficiency of traditional manual screening. Attached Figure Description

[0016] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings:

[0017] Figure 1 This is a system block diagram of the present invention; Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0018] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. Example 1

[0019] Please see Figure 1 - Figure 2As shown, the multi-disease screening system based on multimodal data and large models includes a multimodal data acquisition array, a data port deployment module, a large model tracking and analysis module, a change risk early warning module, and a management terminal interaction module. The data port deployment module is used to obtain the population that needs to be collected and the types of data to be collected, and to make a data port deployment plan based on the population information, while sending the data port deployment plan to the multimodal data acquisition array. Regarding the population to be collected: The data port deployment module obtains individuals with chronic diseases or at high risk of chronic diseases from the cloud-based medical system database and records them as the patient population. At the same time, it obtains the distribution of the patient population and the distribution of the non-patient population within the designated area through information reconstruction technology. It also divides the jurisdiction into multiple sub-regions and calculates the ratio of the patient population to the total population in each sub-region to obtain the patient percentage. After the patient percentages of all sub-regions are calculated, the data port deployment module performs weighted simulation based on the patient percentages. It processes the data based on the number of patients and the patient percentage in each sub-region, with the number of patients as the main factor and the patient percentage as the correction factor, to obtain the representation result of the number of patients. The representation result of each sub-region is then normalized so that the representation result of each sub-region falls within the range of [0, 1], thus obtaining the patient distribution weights of different sub-regions. The data port deployment module categorizes the data to be collected into individual data and centralized data. Individual data is collected through personal devices, such as smartwatches, while centralized data is collected through public devices, such as blood pressure monitors. The data port deployment module generates the distribution density of public devices based on the patient distribution weight and obtains the data port deployment plan. The method for obtaining the distribution density of public equipment by the data port deployment module is as follows: The data port deployment module acquires the preset radiation range and number of devices for each public facility. The radiation range is typically defined by residential area and is set up using units such as communities, neighborhood committees, and pharmacies. The data port deployment module obtains the number of patients based on patient distribution weights and calculates the distribution quantity S using the number of patients, the number of devices each public facility can radiate, and an adjustment coefficient. Meanwhile, the data port deployment module analyzes the radiation range of each public device using a preset algorithm. By distributing a fixed number of public devices in different locations, it obtains the maximum coverage rate based on the radiation range of the public devices. The distribution location of the public devices is determined when the coverage rate reaches a set threshold, and the distribution location and number of public devices are recorded as the public device distribution density. The multimodal data acquisition array collects data from the terminal through the data port. The terminal data collected by the multimodal data acquisition array includes exercise level, heart rate curve, blood glucose change and blood pressure data. Blood pressure data is collected through public equipment, while exercise level, heart rate curve and blood glucose change are collected through personal equipment. The collection frequency of public equipment is lower than that of personal equipment. The multimodal data acquisition array records each piece of terminal data acquired as a sample data. The multimodal data acquisition array obtains the sample threshold for each data category and compares the acquired results with the set acquisition requirements. When the number of acquired sample data exceeds the set sample threshold, it is recorded as the acquisition requirement is completed, and the sample data is packaged and uploaded to the cloud. When the number of sample data does not reach the set sample threshold, the ratio of the number of samples to the set sample threshold is recorded as the collection completion rate. The multimodal data acquisition array compares the collection completion rate of each data category and records the data category with the lowest collection completion rate as the abnormal category. The large-scale model tracking and analysis module acquires terminal data from the cloud, along with corresponding disease management information. It then performs identity matching based on the terminal data and uses the matched data and disease management information for simulation analysis through a large-scale model. The module takes exercise intensity, heart rate curve, blood glucose fluctuations, and blood pressure data as input parameters. During simulation analysis, it compares the exercise intensity with a set exercise range to determine whether the daily activity level of the patient or high-risk chronic disease population meets prevention and treatment requirements. It also compares the heart rate curve with set heart rate parameters. The comparison points include average heart rate, maximum heart rate, minimum heart rate, and abnormal heart rate to confirm whether there are any abnormalities in heart rate. Blood glucose changes are compared with the set blood glucose parameters. During the comparison, predictions are made based on time to confirm the rise and fall of fasting blood glucose and blood glucose at multiple time points after meals to confirm whether blood glucose is in a healthy state. Blood pressure data is compared with the set blood pressure range to confirm whether systolic and diastolic blood pressure are in a healthy state. If they are not in a healthy state, they are judged to be outside the range. When the output parameters match the set range and parameters, it is recorded as exponential coincidence. When it exceeds the set range, the proportion of the excess is recorded as exponential deviation. The large model tracking and analysis module statistically analyzes the index deviation and index overlap, and performs normalization to obtain index samples. The index sample with all data categories having index overlap is the optimal value of 0. The arithmetic mean of all health index samples is then calculated to obtain the final patient health index. After acquiring patients' health indicators, the risk warning module simultaneously obtains the patients' disease risk levels via the cloud. Disease risk levels are distinguished by color, such as green, blue, yellow, orange, and red. Green indicates a low risk of chronic diseases, blue and yellow indicate high risk, with yellow indicating a higher risk than blue. Orange and red indicate patients with existing chronic diseases, with red indicating a more severe condition than yellow. Weights are calculated based on the patient's disease risk level and health indicators. Specifically, different disease risk levels are converted into expansion coefficients using a pre-set database. These expansion coefficients are multiplied by the patient's health indicators to obtain the weight calculation result. The expansion coefficients for green, blue, and yellow gradually approach 1, while those for orange and red are greater than 1. The weight calculation result is compared with a set threshold. If the result is greater than the threshold, a worsening indicator is obtained; otherwise, a normal indicator is obtained. These indicators are recorded as treatment follow-up information, enabling large-scale management of all chronic disease patients or high-risk groups within the jurisdiction through data visualization. Example 2

[0020] Please see Figure 1 - Figure 2 As shown, the management terminal interaction module obtains treatment follow-up information and patient health indicators. It expands the patient health indicators through the treatment follow-up information, compares the expanded patient health indicators with the set threshold levels to obtain the health change level, and regenerates the new patient disease risk level based on the patient's existing disease risk level and health change level, and outputs it through the network. After obtaining the disease risk level, the management terminal interaction module sends the disease risk level to the data port deployment module. The data port deployment module adds the disease risk level to the patient distribution weight, thereby assigning a corresponding weight to each patient or high-risk individual. That is, without a weight, each patient or high-risk individual is counted as 1, and the number of people is used as the basis for the weight. After the weight is assigned, the count of each patient or high-risk group with different risk levels may be less than 1 or greater than 1. The count result is used as the basis for the weight. After combining the weights, the patient distribution weight is recalculated to generate a new patient distribution weight, and a new data port deployment scheme is obtained based on the new patient distribution weight.

[0021] In the technical solutions involved in this invention patent, the use of personal privacy data strictly complies with relevant laws and regulations and is fully legal. Before using personal privacy data, explicit consent from the data subject is obtained in advance, and this consent is given voluntarily by the data subject with full knowledge. The purpose of data use is clear and reasonable, used only to achieve the specific function of the technical solution involved in the invention, directly related to the processing purpose, and adopted in a manner that minimizes the impact on personal rights. When collecting personal privacy data, it is limited to the minimum scope required to achieve the processing purpose, and excessive collection is avoided. At the same time, the entire data use process follows the principles of openness and transparency, clearly informing the data subject of the purpose, method, and scope of processing, and protecting the data subject's right to know and right to control.

[0022] Thresholds, preset values, preset ranges, etc. are set for result comparison and analysis to determine whether they are good or bad. The value of these thresholds is determined by a combination of large-scale model analysis of sample data and human experience. They can also be adjusted appropriately based on seasonal or common-sense influences. Furthermore, the settings for weighting ratios, influence factors, etc., are based on the magnitude of each parameter's influence on the results. The specific values ​​are allocated to ultimately reflect the impact on the results. The settings for input and storage are also determined by a combination of large-scale model analysis of sample data and human experience. Appropriate adjustments can also be made based on seasonal or rational influence conditions.

[0023] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to specific embodiments. Obviously, many modifications and variations are possible based on the contents of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-disease screening system based on multimodal data and large models, characterized in that, It includes a multimodal data acquisition array, a data port deployment module, a large model tracking and analysis module, a change risk early warning module, and a management terminal interaction module. The data port deployment module is used to obtain the population that needs to be collected and the types of data to be collected, and to make a data port deployment plan based on the population information, and at the same time send the data port deployment plan to the multimodal data acquisition array. The multimodal data acquisition array acquires data from the terminal through the data port, compares the acquired results with the set acquisition requirements, uploads the acquired terminal data to the cloud, and generates the acquisition completion rate based on the comparison results. The large-scale model tracking and analysis module acquires terminal data through the cloud, and simultaneously acquires disease control information corresponding to the identity. It then performs identity matching based on the terminal data, and uses the large-scale model to simulate and analyze the identity-matched terminal data and disease control information to obtain patient health indicators. The change risk early warning module obtains the patient's health indicators through the large model tracking and analysis module, and generates treatment follow-up information based on the patient's health indicators; The management terminal interaction module acquires treatment follow-up information and patient health indicators, regenerates the patient's disease risk level, and outputs it through the network. At the same time, it sends the data port laying module according to the disease risk level, and the data port laying module performs data port laying reconstruction.

2. The multi-disease screening system based on multimodal data and large models according to claim 1, characterized in that, The data port deployment module obtains the patient population with chronic diseases through the cloud, and at the same time obtains the distribution of the patient population in a set area through information reconstruction technology. It performs weight simulation based on the patient population distribution to obtain the patient distribution weight in different areas. The data port deployment module categorizes the data to be collected into individual data and centralized data. Individual data is collected using personal devices, while centralized data is collected using shared devices. The data port deployment module generates the distribution density of shared devices based on the patient distribution weights and obtains the data port deployment plan.

3. The multi-disease screening system based on multimodal data and large models according to claim 2, characterized in that, The method by which the data port deployment module obtains the distribution density of public equipment is as follows: The data port deployment module obtains the preset radiation range and radiation quantity of each public device. The data port deployment module obtains the number of patients based on the patient distribution weight, and calculates the distribution quantity by using the number of patients, the radiation quantity of each public device, and the adjustment coefficient. At the same time, the data port deployment module analyzes the radiation range of each public device using a preset algorithm to obtain the distribution location of public devices when the coverage reaches a set threshold, and records the distribution location and distribution quantity of public devices as the public device distribution density.

4. The multi-disease screening system based on multimodal data and large models according to claim 1, characterized in that, The terminal data collected by the multimodal data acquisition array includes exercise intensity, heart rate curve, blood glucose variation and blood pressure data. Blood pressure data is collected through public devices, while exercise intensity, heart rate curve and blood glucose variation are collected through personal devices. The collection frequency of public devices is lower than that of personal devices. The multimodal data acquisition array records each piece of terminal data acquired as a sample data. The multimodal data acquisition array obtains the sample threshold for each data category, and when the number of acquired sample data exceeds the set sample threshold, it records that the acquisition requirement has been completed, and packages the sample data and uploads it to the cloud. When the number of sample data does not reach the set sample threshold, the ratio of the number of samples to the set sample threshold is recorded as the collection completion rate. The multimodal data acquisition array compares the collection completion rate of each data category and records the data category with the lowest collection completion rate as the abnormal category.

5. The multi-disease screening system based on multimodal data and large models according to claim 1, characterized in that, The large-scale model tracking and analysis module takes exercise intensity, heart rate curve, blood glucose fluctuation and blood pressure data as input parameters. When performing simulation analysis, the large-scale model tracking and analysis module compares the exercise intensity with the set exercise range, the heart rate curve with the set heart rate parameter, the blood glucose fluctuation with the set blood glucose parameter, and the blood pressure data with the set blood pressure range. When the output parameters match the set range and parameters, it is recorded as exponential coincidence. When it exceeds the set range, the excess ratio is recorded as exponential deviation. The large model tracking and analysis module statistically analyzes index deviation and index overlap, performs normalization processing to obtain index samples, and calculates an arithmetic mean of all health index samples to obtain the final patient health index.

6. The multi-disease screening system based on multimodal data and large models according to claim 1, characterized in that, After obtaining the patient's health indicators, the change risk warning module will simultaneously obtain the patient's disease risk level through the cloud, and perform weight calculation based on the patient's disease risk level and the patient's health indicators. The weight calculation result will be compared with a set threshold to obtain the follow-up deterioration index or the follow-up normal index, and the follow-up deterioration index or the follow-up normal index will be recorded as treatment follow-up information.

7. The multi-disease screening system based on multimodal data and large models according to claim 1, characterized in that, The management terminal interaction module expands the patient's health indicators through treatment follow-up information, compares the expanded patient health indicators with the set threshold levels to obtain the health change level, and regenerates the new patient disease risk level based on the patient's existing disease risk level and health change level.

8. The multi-disease screening system based on multimodal data and large models according to claim 1, characterized in that, After obtaining the disease risk level, the management terminal interaction module sends the disease risk level to the data port deployment module. The data port deployment module adds the disease risk level to the patient distribution weight, regenerates the new patient distribution weight, and obtains a new data port deployment scheme based on the new patient distribution weight.

Images