Schistosomiasis transmission risk dynamic simulation method based on data fusion and AI

Abstract

The invention discloses a schistosomiasis transmission risk dynamic simulation method based on data fusion and AI, and relates to the technical field of intelligent medical treatment. The method comprises the following steps: firstly, collecting environment, biological media, crowd society and intervention measure multi-source heterogeneous data, and unifying the data to a standard spatio-temporal data cube; then data preprocessing is carried out, and feature variables strongly related to propagation risks are constructed; then the processed data and the characteristic variables are input into a risk simulation engine, the engine adopts a hybrid framework in which a mechanism model and an AI prediction model are deeply fused, a dynamic assimilation algorithm is introduced, and risk level maps of different spatial and temporal scales in the future are output; finally, intervention measure scene simulation is carried out based on the risk level map, the cost-benefit ratio is quantitatively evaluated, and an optimal prevention and control strategy suggestion is generated. According to the method, the hysteresis and static defects of a traditional method are overcome, and real-time sensing, dynamic prediction and intelligent decision support of the schistosomiasis transmission risk are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent medical treatment and public health information technology, and in particular to a schistosomiasis transmission risk dynamic simulation method based on data fusion and AI. BACKGROUND

[0002] Schistosomiasis is a serious parasitic disease that endangers people's health and social and economic development. Its transmission process is complex and is influenced by both natural factors (such as climate, hydrology, and Oncomelania distribution) and social factors (such as population flow and prevention and control measures). Traditional schistosomiasis risk assessment methods mainly rely on historical epidemic data statistics and limited field Oncomelania surveys, and have obvious drawbacks:

[0003] Lag: The assessment results are seriously lagging behind the actual development of the epidemic, and cannot realize forward-looking early warning.

[0004] Unilaterality: It relies on a single or small number of data sources, and it is difficult to comprehensively and accurately reflect the complex transmission dynamics process.

[0005] Static: Most of them are static assessments, and cannot dynamically simulate and predict the risk in space and time.

[0006] Weak decision support: Usually only risk level judgment is provided, and there is a lack of intuitive and quantitative intervention measure effect simulation, making it difficult to support precise prevention and control decisions.

[0007] In the prior art, although there are some models for predicting infectious diseases, most of them are based on a single statistical model or simple machine learning algorithm, and do not deeply integrate multi-source heterogeneous data, nor do they lack special mechanism models for the schistosomiasis transmission chain (infectious source- Oncomelania-susceptible population), resulting in insufficient prediction accuracy and practicality.

[0008] Therefore, there is an urgent need in the art for a method that can integrate multi-source data in real time, dynamically simulate transmission risks, and provide intelligent decision support. SUMMARY

[0009] The purpose of the embodiments of the present application is to propose a schistosomiasis transmission risk dynamic simulation method based on data fusion and AI, aiming to solve the problems of lag, unilateralism and staticness existing in the existing schistosomiasis risk assessment methods.

[0010] The embodiment of the present application is implemented in the following way: a schistosomiasis transmission risk dynamic simulation method based on data fusion and AI, the method comprising the following steps: collecting multi-source heterogeneous data, and unifying the collected multi-source heterogeneous data to a standard space-time data cube; preprocessing the data in the space-time data cube, and constructing a characteristic variable related to the schistosomiasis transmission risk; inputting the preprocessed data and the characteristic variable into a risk simulation engine, and outputting a risk level map of different space-time scales in the future through the risk simulation engine; According to the risk level map, the scenario simulation of the corresponding intervention measure combination is carried out, and the cost-benefit ratio is quantitatively evaluated to generate an optimal prevention and control strategy suggestion.

[0011] Further, the method further comprises the step of: dynamically displaying the spatial distribution of the risk, the space-time evolution animation and the decision support result through a visual human-computer interaction interface.

[0012] In the embodiment of the present application, the multi-source heterogeneous data sources comprise: environmental data including remote sensing satellite data, ground meteorological data and hydrological monitoring data; biological vector data including electronic snail survey data and wild animal distribution data; population and social data including historical epidemic report data, population census and flow data, resident production and life behavior data and land use and cover change data; intervention measure data including drug chemotherapy coverage and snail eradication measure implementation data.

[0013] In the embodiment of the present application, the step of unifying the collected multi-source heterogeneous data to a standard space-time data cube comprises: adopting an ontology-based semantic mapping technology and a space-time alignment algorithm to unify and integrate the collected multi-source data into a standard space-time data cube; the ontology-based semantic mapping method comprises: constructing a special ontology in the field of schistosomiasis, the ontology uniformly defines the core concepts, attributes and their mutual relationships involved in the multi-source heterogeneous data in a formalized way; the fields and coding rules of the multi-source heterogeneous data are mapped to the unified standard concepts according to the ontology; the space-time alignment algorithm comprises: unifying the data with different spatial resolutions and time frequencies to a preset space-time reference through spatial interpolation, resampling or data aggregation method, to form the space-time data cube.

[0014] In the embodiment of the present application, the characteristic variable comprises effective accumulated temperature in a preset historical period, precipitation anomaly in a preset recent period and snail breeding environment suitability index.

[0015] In the embodiment of the present application, the risk simulation engine adopts a hybrid framework of deep fusion of mechanism model and AI prediction model, and introduces a dynamic assimilation algorithm.

[0016] In the embodiment of the present application, the mechanism model is a WAQFWV model based on the transmission dynamics of schistosomiasis, which is used to simulate the transmission chain of pathogens in snail populations and between human and animal hosts, and calculate the basic reproductive number of key epidemiological indicators.

[0017] In the embodiment of the present application, the algorithm adopted by the AI prediction model is LightGBM, XGBoost, spatio-temporal graph convolution network or long short-term memory network.

[0018] In the embodiment of the present application, the dynamic assimilation algorithm is Kalman filtering or variational assimilation method. In the embodiment of the present application, the scenario simulation of the corresponding intervention measure combination is performed according to the risk level map, and the cost-benefit ratio is quantitatively evaluated to generate optimal prevention and control strategy suggestions, which include: According to the risk level map, the corresponding intervention measure combination is set, the risk simulation engine is called to deduce and visualize the dynamic change process of the future risk level map under the intervention measure combination; Based on the preset economic evaluation model, the cost-benefit ratio between the execution cost and the expected reduced disease burden of each simulated intervention measure combination is quantitatively calculated; By comparing the output results of different scenario simulations and their cost-benefit ratios, a report containing optimal prevention and control strategy suggestions is automatically identified and generated.

[0019] Advantages of the present application The present application provides a schistosomiasis transmission risk dynamic simulation method based on data fusion and AI, which relates to the field of intelligent medical treatment. The method first collects multi-source heterogeneous data of environment, biological vectors, human society and intervention measures, and unifies them to a standard spatio-temporal data cube; then performs data preprocessing and constructs characteristic variables strongly related to transmission risk; then inputs the processed data and characteristic variables into a risk simulation engine, which adopts a hybrid framework of deep fusion of mechanism model and AI prediction model, and introduces a dynamic assimilation algorithm, and outputs risk level maps at different spatio-temporal scales in the future; finally, intervention measure scenario simulation is performed based on the risk level map, the cost-benefit ratio is quantitatively evaluated, and optimal prevention and control strategy suggestions are generated. The present application overcomes the lagging and static defects of traditional methods, realizes real-time perception, dynamic prediction and intelligent decision support of schistosomiasis transmission risk. BRIEF DESCRIPTION OF DRAWINGS

[0020] Figure 1 is a flowchart of the schistosomiasis transmission risk dynamic simulation method based on data fusion and AI of the preferred embodiment of the present application; Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. For ease of explanation, only the parts related to the embodiments of this invention are shown. It should be understood that the specific embodiments described herein are merely for explaining this invention and are not intended to limit this invention.

[0022] This invention proposes a dynamic simulation method for schistosomiasis transmission risk based on data fusion and AI, belonging to the field of smart healthcare technology. The method first collects multi-source heterogeneous data on the environment, biological vectors, social populations, and intervention measures, unifying them into a standard spatiotemporal data cube. Then, data preprocessing is performed to construct feature variables strongly correlated with transmission risk. Next, the processed data and feature variables are input into a risk simulation engine. This engine employs a hybrid framework that deeply integrates mechanistic models and AI prediction models, and incorporates a dynamic assimilation algorithm to output risk level maps at different future spatiotemporal scales. Finally, based on the risk level maps, intervention scenario simulations are conducted to quantitatively evaluate the cost-benefit ratio and generate optimal prevention and control strategy recommendations. This invention overcomes the lag and static nature of traditional methods, achieving real-time perception, dynamic prediction, and intelligent decision support for schistosomiasis transmission risk.

[0023] Figure 1 This is a flowchart of a preferred embodiment of the present invention for a dynamic simulation method of schistosomiasis transmission risk based on data fusion and AI; the method includes the following steps: S1, collect multi-source heterogeneous data and unify the collected multi-source heterogeneous data into a standard spatiotemporal data cube; The multi-source heterogeneous data includes: Environmental data includes remote sensing satellite data (such as vegetation index NDVI, land surface temperature LST, and water body distribution), ground meteorological observation data (such as temperature, precipitation, and humidity), and hydrological monitoring data (such as water level and flow velocity).

[0024] Biological vector data: including electronic Oncomelania snail survey data (including snail density and infection rate) and distribution data of wild animals (such as insect hosts).

[0025] Population and social data: including historical epidemic report data, population census and mobility data, residents' production and living behavior data, and land use and cover change data.

[0026] Intervention data: including coverage of drug chemotherapy and data on the implementation of molluscicide measures (including the geographical scope and time of implementation of molluscicide measures, the status of health education activities and the population covered).

[0027] The resident production and life behavior data includes various data for directly or indirectly evaluating the resident epidemic water contact risk, and specifically includes: Resident epidemic water contact history, inferred through questionnaire survey or mobile device signaling data, including the contact geographical position, contact frequency, duration and activity type (such as farming, washing, swimming, etc.); Population spatiotemporal distribution and flow trajectory, based on mobile device signaling data inversion, used for macro-inference of the exposure probability of the population in the epidemic area periphery; High-risk occupation information, obtained from population census or administrative records, such as occupation as a farmer or fisherman; Land use data, land use data (such as the distribution of farmland and fishponds) related to the production and life activities of residents.

[0028] In the embodiment of the application, the collected multi-source data is uniformly integrated into a standard spatiotemporal data cube by using an ontology-based semantic mapping technology and a spatiotemporal alignment algorithm.

[0029] The ontology-based semantic mapping method is to construct a special ontology in the field of schistosomiasis, and the ontology uniformly defines the core concepts (such as air temperature, snail density and chemotherapy coverage) involved in the multi-source heterogeneous data, attributes and their mutual relationships in a formalized manner. Subsequently, the fields and coding rules of the multi-source heterogeneous data are mapped to the unified standard concept according to the ontology;

[0030] The spatiotemporal alignment algorithm is to unify the data with different spatial resolutions (such as kilometer grid, administrative boundary) and time frequencies (such as daily, monthly, irregular) to a preset spatiotemporal reference (such as 1km×1km grid, daily scale) through spatial interpolation, resampling and data aggregation, etc., to form the spatiotemporal data cube. S2, the data in the spatiotemporal data cube is preprocessed, and the feature variables related to the schistosomiasis transmission risk are constructed;

[0031] The preprocessing includes cleaning, denoising and missing value interpolation; The cleaning method is to identify and eliminate obviously abnormal values (such as temperature values exceeding the reasonable range) that do not conform to the physical meaning; The denoising method is to smooth random fluctuations and eliminate data noise by using a filtering algorithm (such as moving average); The missing value interpolation method is to interpolate the missing values in the data by using a spatial interpolation method (such as spatiotemporal Kriging interpolation, inverse distance weighting method) or a time series interpolation method (such as prediction filling based on a time series model (such as ARIMA, SARIMA), linear interpolation), to ensure the continuity and integrity of the data set; The characteristic variables include: effective accumulated temperature in a preset historical period, precipitation anomaly in a preset recent period, and snail breeding environment suitability index. The effective accumulated temperature in a preset historical period is the accumulated temperature higher than the snail and schistosome development threshold in a preset historical period (such as the previous 30 days), which is used to measure the cumulative effect on the development stage of the pathogen.

[0032] The precipitation anomaly in a preset recent period is the deviation of precipitation in a preset recent period (such as the last two weeks) from the historical average value, which is used to reflect the influence of precipitation anomaly on the formation of snail breeding sites.

[0033] The snail breeding environment suitability index is an index calculated by a specific model (such as analytic hierarchy process, logistic regression) based on environmental variables such as water body distribution, vegetation condition, and soil humidity, which is used to evaluate the potential suitability of a specific area for snail breeding.

[0034] S3, input the preprocessed data and the characteristic variables into the risk simulation engine, and output the risk level map of different spatio-temporal scales in the future through the risk simulation engine; The risk simulation engine adopts a hybrid framework of deep fusion of mechanism model and AI prediction model, and introduces a dynamic assimilation algorithm; In the embodiment of the application, the risk simulation engine processing process includes: Model input: input the preprocessed data and the characteristic variables into the risk simulation engine; Mechanism model unit: based on the schistosomiasis transmission dynamics model (such as WAQFWV model), simulate the transmission chain of the pathogen in the snail population and between human and animal hosts, and calculate the key epidemiological indicators (basic reproduction number (R0)); AI prediction model unit: adopt machine learning models (such as LightGBM, XGBoost) or deep learning models (such as spatio-temporal graph convolution network ST-GCN, long short-term memory network LSTM), take the characteristic variables and the intermediate output or final output of the mechanism model as input, and perform risk prediction. Learn complex nonlinear relationships and hidden patterns from data, and correct and enhance the mechanism model;

[0035] Model fusion and dynamic assimilation: set a fusion and optimization layer to weight or integrate the outputs of the mechanism model and the AI model to generate the final prediction; at the same time, adopt a dynamic assimilation algorithm (such as Kalman filter, ensemble Kalman filter, variational assimilation algorithm) to fuse the real-time incoming new observation data (such as newly discovered infectious snail points, latest confirmed epidemic cases) with the model prediction results, and update and correct the state variables of the model in real time, to ensure the timeliness and accuracy of the prediction results; Model output: The risk simulation engine outputs risk level maps for different time periods (e.g., future 1 week, 1 month) and different spatial grids (e.g., 1 km grid), which are characterized by quantitative risk indices or basic reproduction number R0 values.

[0036] S4, according to the risk level map, scenario simulation of corresponding intervention measure combination is performed, and cost-benefit ratio is quantitatively evaluated, and optimal prevention and control strategy suggestion is generated.

[0037] Comprise: According to the risk level map, a corresponding intervention measure combination (for example, 70% drug chemotherapy coverage is implemented in region A, and mechanical snail eradication is carried out in region B for two weeks) is set, a risk simulation engine is called to deduce and visualize the dynamic change process of the future risk level map under the intervention measure combination; Based on a preset economic evaluation model, the cost-benefit ratio between the execution cost and the expected reduced disease burden (such as the number of infection years that can be avoided) of each simulated intervention measure combination is quantitatively calculated; By comparing the output results of different scenario simulations and their cost-benefit ratios, a report containing optimal prevention and control strategy suggestions is automatically identified and generated; In the embodiment of the application, the intervention measure combination is set by the user; Further, in some embodiments, after step S4, the method further comprises: S5, the spatial distribution of risk, the time-space evolution animation and the decision support result are dynamically displayed through a visual human-computer interaction interface.

[0038] The embodiment of the application dynamically displays the spatial distribution and time-space evolution animation of historical risk, real-time risk and future predicted risk through a visual human-computer interaction interface based on WebGIS, taking an electronic map as a carrier. At the same time, interactive controls are provided for the user to set simulation parameters, start scenario simulation and intuitively view decision support results.

[0039] The dynamic risk spatial distribution display is to dynamically display the risk level map on the electronic map in the form of a heat map, a hierarchical color setting map or a grid rendering. The user can view the risk level (such as low, medium and high) or specific risk index / R0 value corresponding to different colors through a legend;

[0040] The time-space evolution animation presents a time axis control, allowing the user to customize the time range (such as the past month and the future two weeks). By continuously playing the risk level maps on different time slices, a time-space evolution animation is generated to intuitively display the dynamic process of the spatial and temporal diffusion, aggregation and dissipation of the schistosomiasis transmission risk;

[0041] The decision support result comprises: Display the risk comparison under different intervention scenarios in the form of a thematic map side by side; Display the change of the key area risk index over time in the form of a chart (e.g. bar chart, trend line); Present the quantitative evaluation results such as cost-benefit ratio in the form of a table or dashboard; The implemented interactive controls include: Map interaction: support basic operations such as zooming, panning, ranging, and area calculation, and support clicking on map elements (e.g. risk grid) to display detailed attribute information of the location in the form of a pop-up window.

[0042] Parameter setting and simulation triggering: provide interactive controls (e.g. drop-down menu, slider, checkbox) to allow users to set simulation parameters (e.g. prediction length, risk threshold) or build intervention measure combinations (e.g. check specific areas, set chemotherapy coverage) on the front-end interface. Users can trigger the scenario simulation function of the system by clicking the "start simulation" button.

[0043] Data layer management: provide a layer control panel, users can switch to display or hide background data layers (e.g. snail distribution points, water body network, administrative boundary, population density, etc.), realize the superposition analysis and comparison of risk information and multi-source background data.

[0044] The embodiment of the present application adopts a front-end and back-end separation architecture. The back-end provides processed risk data, map services and simulation results through RESTful API; the front-end uses a JavaScript map library (e.g. Leaflet, OpenLayers or Mapbox GL JS) for rendering and interaction.

[0045] The spatio-temporal evolution animation is realized by calling time-series risk data, and performing smooth interpolation and rendering on the front-end using a graphics library.

[0046] Those skilled in the art can understand that all or part of the steps in the above-mentioned embodiment methods can be completed by program instruction related hardware, and the program can be stored in a computer readable storage medium, such as ROM, RAM, magnetic disk, optical disc, etc.

[0047] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. Any modification, equivalent replacement and improvement within the spirit and principle of the present application shall be included in the protection scope of the present application.

Claims

1. A schistosomiasis transmission risk dynamic simulation method based on data fusion and AI, characterized in that, The method comprises the following steps: collecting multi-source heterogeneous data, and unifying the collected multi-source heterogeneous data into a standard space-time data cube; preprocessing data in the space-time data cube and constructing feature variables related to schistosomiasis transmission risk; inputting the preprocessed data and the feature variables into a risk simulation engine, and outputting a risk level map of different space-time scales in the future through the risk simulation engine; According to the risk level map, scenario simulation of corresponding intervention measure combinations is carried out, and the cost-benefit ratio is quantitatively evaluated to generate an optimal prevention and control strategy suggestion.

2. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 1, characterized in that, The multi-source heterogeneous data sources include: environmental data including remote sensing satellite data, ground meteorological data and hydrological monitoring data; biological vector data including electronic snail survey data and wild animal distribution data; population and social data including historical epidemic report data, population census and flow data, resident production and life behavior data and land use and cover change data; intervention measure data including drug chemotherapy coverage and snail eradication measure implementation data.

3. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 1, characterized in that, The unification of the collected multi-source heterogeneous data into a standard space-time data cube comprises: adopting an ontology-based semantic mapping technology and a space-time alignment algorithm to unify and integrate the collected multi-source data into a standard space-time data cube; the ontology-based semantic mapping method comprises: constructing a special ontology in the field of schistosomiasis, the ontology uniformly defines the core concepts, attributes and their mutual relationships involved in the multi-source heterogeneous data in a formalized manner; and mapping the fields and coding rules of the multi-source heterogeneous data to uniform standard concepts according to the ontology; the space-time alignment algorithm comprises: unifying data with different spatial resolutions and time frequencies to a preset space-time reference through spatial interpolation, resampling or data aggregation methods to form the space-time data cube.

4. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 1, characterized in that, The feature variables include effective accumulated temperature in a preset historical period, precipitation anomaly in a preset recent period and snail breeding environment suitability index.

5. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 1, characterized in that, The risk simulation engine adopts a hybrid framework of deep fusion of mechanism model and AI prediction model, and introduces a dynamic assimilation algorithm.

6. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 5, characterized in that, The mechanism model is a WAQFWV model based on the dynamics of schistosomiasis transmission, which is used to simulate the transmission chain of pathogens in snail populations and between human and animal hosts, and calculate the basic reproduction number of key epidemiological indicators.

7. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 5, characterized in that, The AI prediction model adopts algorithms such as LightGBM, XGBoost, space-time graph convolution network or long short-term memory network.

8. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 5, characterized in that, The dynamic assimilation algorithm is a Kalman filter or a variational assimilation method.

9. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 1, wherein according to the risk level map, scenario simulation of corresponding intervention measure combinations is carried out, and the cost-benefit ratio is quantitatively evaluated to generate an optimal prevention and control strategy suggestion, which comprises: According to the risk level map, set the corresponding intervention measure combination, call the risk simulation engine to deduce and visualize the dynamic change process of the future risk level map under the intervention measure combination; Based on the preset economic evaluation model, the cost-effectiveness ratio between the execution cost and the expected disease burden reduction of each simulated intervention combination is quantitatively calculated; By comparing the output results and cost-effectiveness ratios of different scenario simulations, a report containing optimal prevention and control strategy suggestions is automatically identified and generated.

10. The schistosomiasis transmission risk dynamic simulation method based on data fusion and AI according to claim 1, characterized in that, The method further comprises the step of dynamically displaying the spatial distribution of the risk, the spatiotemporal evolution animation and the decision support result through a visual human-computer interaction interface.

Images