An intelligent system for HIV prevention and detection

Abstract

This invention discloses an intelligent system for HIV prevention and detection, belonging to the field of medical and health data processing technology. The system includes an intelligent testing kit, a user terminal, and a disease control management terminal. The intelligent testing kit has a built-in image acquisition unit for acquiring the colorimetric image of the HIV test strip; the user terminal receives the colorimetric image and extracts colorimetric feature parameters, generates a continuously variable risk index through a risk mapping model, and simultaneously identifies high-risk scenarios through a behavior perception module, dynamically pushing intervention information based on the risk index and / or high-risk scenarios; the disease control management terminal receives the desensitized risk index and corresponding anonymous identifiers and spatiotemporal data, constructs a risk propagation model based on a graph neural network, traces the spatiotemporal companions of risk nodes, and pushes anonymous early warning information. This invention achieves a closed loop from individual detection to group early warning, which can improve the early detection rate of HIV, achieve precise intervention, and optimize resource allocation.

Description

Technical Field

[0001] This invention relates to the field of medical and health data processing technology, specifically to an intelligent system and method for HIV prevention and detection. Background Technology

[0002] This section provides only background information related to this application to enable those skilled in the art to understand this application more thoroughly and accurately, and it is not necessarily prior art.

[0003] Current HIV prevention and control efforts mainly include prevention and intervention for high-risk groups and screening and testing for the general public. In existing technologies, these two aspects are often separated. Some information management systems are primarily used for case reporting and data statistics. While some testing technologies, such as image recognition-based test strip interpretation, are applied, they still suffer from the following problems: prevention and intervention lack specificity and cannot be dynamically adjusted based on individual real-time behavioral risks; test results, especially negative results, are difficult to effectively feed back to the disease control system to form a data loop; and there is a lack of intelligent analysis models that can integrate individual behavioral data, testing data, and spatial data. Therefore, a system is needed that can combine prevention and testing to achieve precise intervention and intelligent early warning. Summary of the Invention

[0004] The purpose of this invention is to provide an intelligent system and method for HIV prevention and detection, so as to solve the problem of the separation between prevention intervention and detection management in the prior art.

[0005] To achieve the above objectives, this invention provides an intelligent system for HIV prevention and detection, comprising an intelligent testing kit, a user terminal, and a disease control management terminal. The intelligent testing kit includes a card slot and an image acquisition unit. The card slot receives HIV test strips, and the image acquisition unit acquires colorimetric images of the reaction area on the test strip. Preferably, the intelligent testing kit also includes a multispectral light source, enabling image acquisition under different wavelengths of light, facilitating subsequent analysis of the test strip's effectiveness and colorimetric characteristics.

[0006] The user terminal communicates with the intelligent detection box and includes an image processing module, a risk assessment module, a behavior perception module, and an intervention engine module. The image processing module receives colorimetric images and extracts colorimetric feature parameters, such as the color width of the detection line area, color saturation gradient, and color continuity index. The risk assessment module generates a risk index based on the colorimetric feature parameters; this risk index is continuously evaluated. The behavior perception module identifies high-risk scenarios using data from the phone's built-in sensors, such as staying in a specific area at night or abnormal movement trajectories across multiple locations. The intervention engine module pushes personalized intervention information based on the risk index and / or high-risk scenarios, such as medication reminders, self-testing device navigation, and testing suggestions.

[0007] The disease control management terminal communicates with multiple user terminals, including a data receiving module, a risk transmission analysis module, and an early warning push module. The data receiving module receives anonymized risk indices, corresponding anonymous identifiers, and spatiotemporal data. The risk transmission analysis module constructs a risk transmission model based on a graph neural network, using anonymous identifiers as nodes and spatiotemporal co-occurrence as edges. When an abnormal risk index is detected at a node, the risk transmission probability of its neighboring nodes is calculated to identify potential risk transmission targets. The early warning push module pushes anonymous early warning information to identified risk transmission targets, reminding them to undergo testing.

[0008] The present invention also provides an HIV prevention and detection method based on the above system, including steps such as image acquisition, feature extraction, risk assessment, data uploading, graph network analysis, risk tracing, and early warning push.

[0009] Compared with the prior art, the present invention has the following beneficial effects: By combining intelligent detection boxes and user terminals, quantitative analysis of detection data and fusion of behavioral perception are achieved to generate a continuous risk index. Through graph neural network analysis at the disease control management end, risk transmission prediction and anonymous early warning based on spatiotemporal co-occurrence are realized, forming a closed loop from individual detection to group early warning. Through the intervention engine module, precise intervention based on individual risk status is realized to improve prevention and control efficiency.

[0010] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0011] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0012] Figure 1 This is a general block diagram of an intelligent system for HIV prevention and detection according to the present invention.

[0013] Figure 2 This is a block diagram of the intelligent detection box of the intelligent system of the present invention.

[0014] Figure 3 This is a block diagram of the user terminal of the intelligent system of the present invention.

[0015] Figure 4 This is a block diagram of the disease control management module of the intelligent system of the present invention.

[0016] Figure 5 This is a flowchart of the HIV prevention and detection method based on an intelligent system according to the present invention. Detailed Implementation

[0017] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0018] Example 1: like Figure 1 As shown, the intelligent system for HIV prevention and detection of the present invention includes three core components: an intelligent testing box 100, a user terminal 200, and a disease control management terminal 300. The intelligent testing box 100 communicates with the user terminal 200 via Bluetooth, and the user terminal 200 communicates with the disease control management terminal 300 via a 4G / 5G network. The data flow is as follows: test data is uploaded from the intelligent testing box 100 to the user terminal 200, processed, and then uploaded to the disease control management terminal 300; the control flow is as follows: intervention information and early warning information are distributed from the disease control management terminal 300 to the user terminal 200.

[0019] The intelligent testing kit 100 is used to acquire the colorimetric image of the HIV test strip. It integrates components such as an image acquisition unit, a multispectral light source, and control circuitry. The user terminal 200 is a smartphone with a dedicated app installed, containing at least four modules: image processing, risk assessment, behavior perception, and an intervention engine. The disease control management terminal 300 is deployed on the disease control center's server and includes modules for data reception, risk transmission analysis, and early warning push notifications.

[0020] Example 2: The smart testing kit 100 has a slot for inserting HIV test strips, and the kit contains the following modules: Image acquisition unit 101: Employs a high-resolution miniature camera to capture images of the test strip's reaction area upon triggering by the control circuit. Multispectral light source 102: Includes at least two LED light sources of different wavelengths, such as a 450nm blue LED and a 630nm red LED, used to illuminate the test strip's detection and control areas respectively, for subsequent analysis of the test strip's effectiveness and colorimetric characteristics. Control circuit 103: Responsible for image acquisition and light source switching, controlling the light source illumination and camera capture according to a preset timing sequence after receiving a trigger signal. Communication module 104: May employ a low-power Bluetooth module to transmit acquired image data to the user terminal 200. Power module 105: Contains a built-in rechargeable battery to power the aforementioned modules and may include a trigger button to initiate the detection process.

[0021] In use, the user inserts the completed test strip into the slot on the test box, presses the trigger button, and the control circuit starts image acquisition: first, the blue light is turned on to capture one frame of image, and then the red light is turned on to capture one frame of image. The two frames of image are transmitted to the user terminal via Bluetooth.

[0022] Example 3: like Figure 3 As shown, the user terminal 200 includes the following software modules: Image processing module 210: Receives the colorimetric image from the intelligent detection box 100, performs preprocessing (grayscale conversion, filtering and noise reduction, contrast enhancement), and then uses a deep learning semantic segmentation model (such as U-Net) to identify the detection line region (T-line) and control line region (C-line) in the image. For the T-line region, it extracts colorimetric feature parameters, including color width (the statistical average of pixels perpendicular to the line direction), color saturation gradient (the maximum value and half-width at half-maximum obtained from the cross-sectional analysis of the S-channel in HSV space), and color continuity index (the proportion of connected region area). For the C-line region, it analyzes the reflectance characteristics of the control line using multispectral image analysis to determine whether the test strip is valid (if the reflectance is within a preset range, it is valid; otherwise, it is invalid, and the user is prompted to replace the test strip). The color saturation gradient parameter refers to the distribution curve of pixel color saturation values ​​(S-channel in HSV space) along the direction perpendicular to the detection line (T-line) as a function of spatial distance, extracted from the HIV test strip image, and feature parameters extracted from this curve, including but not limited to peak value and half-width at half-maximum.

[0023] Risk assessment module 220: Employs a pre-trained deep neural network. Inputs are feature parameters extracted by the image processing module (such as color width, color saturation gradient, color continuity index, and quality control validity indicators). Output is a continuous risk index between 0 and 100. This network is a three-layer fully connected network, trained on a large number of samples with known infection states, capable of quantitatively assessing infection risk, especially demonstrating good identification ability for early weak positive results.

[0024] Behavior perception module 230: Collects data through the phone's built-in sensors such as the gyroscope, accelerometer, and GPS. First, it performs privacy-preserving computation and anonymization: hashing the user ID to generate an anonymous identifier, converting GPS coordinates into a geogrid code with a certain precision (e.g., 150m×150m GeoHash encoding), and coarsening the timestamp to the hourly level. Then, it uses a Hidden Markov Model to identify the user's state (home, work, mobile, entertainment), thereby identifying high-risk scenarios, such as staying in a bar area at night (22:00-04:00), frequent switching between multiple locations within a day (high trajectory entropy), and abnormal nighttime stops. These high-risk scenario labels, along with the anonymized spatiotemporal data, are output.

[0025] Intervention Engine Module 240: Based on the risk index output by the risk assessment module and the high-risk scenarios identified by the behavior perception module, personalized intervention information is pushed through the rule engine. For example: If the risk index is <30 and there are no high-risk scenarios: push routine health education information. If 30 ≤ risk index ≤ 60 or a high-risk scenario occurs: push navigation to nearby self-testing devices and PrEP medication reminders. If the risk index is >60 or (30-60 with consecutive high risks): push emergency testing suggestions and provide a free testing voucher. If the user is taking PrEP medication (through user settings or smart pill bottle data), combined with missed dose records, when consecutive missed doses are detected and the behavior perception module indicates high risk, emergency testing and make-up dose reminders are triggered.

[0026] Example 4: like Figure 4 As shown, the CDC management terminal 300 is deployed on the CDC server and communicates with tens of thousands of user terminals via the Internet. Its core modules include: Data receiving module 310: Receives anonymized data packets from various user terminals via an HTTPS interface. Each data packet contains: an anonymous identifier, timestamp, geographic grid code, risk index, and high-risk scenario label (optional). The data is encrypted and stored in the database after parsing and verification.

[0027] Risk propagation analysis module 320: Constructs a risk propagation model based on a graph neural network. Anonymous identifiers are used as nodes, and a dynamic graph is constructed with spatiotemporal co-occurrence (two nodes sharing the same grid code and a time difference less than a preset threshold, such as 2 hours) as edges. Node features include the average risk index over the past 7 days, the frequency of high-risk scenarios, and the trend of risk index changes. When a node's risk index exceeds the first threshold (e.g., 80), the risk propagation probability of all its neighboring nodes is calculated using the graph neural network's message passing mechanism. Neighboring nodes exceeding the second threshold (e.g., 0.5) are marked as risk propagation targets.

[0028] Warning push module 330: Pushes anonymous warning information to marked individuals at risk of HIV transmission, such as "Based on big data analysis, you may have recently been exposed to a high-risk environment for HIV. We recommend that you get tested after the window period. Click to receive a free testing voucher." The information is pushed through the user's mobile app and does not reveal the identity information of any associated individuals.

[0029] Resource scheduling module 340: Based on the high-risk node cluster area (hotspot grid code) output by the risk propagation analysis module, generate material replenishment instructions to guide the replenishment of self-testing consumables and intervention materials to the smart vending machines in that area.

[0030] Example 5: like Figure 5 As shown, the HIV prevention and detection method of the present invention includes the following steps: S1: The smart testing box acquires the colorimetric image of the HIV test strip. The user inserts the completed test strip into the smart testing box, presses the trigger button, and the smart testing box captures two images (blue light and red light) under multispectral light source illumination, which are then transmitted to the user's terminal via Bluetooth.

[0031] S2: Process the colorimetric image and extract colorimetric feature parameters. The image processing module of the user terminal preprocesses and semantically segments the image, identifies the C-line and T-line, calculates the colorimetric width, color saturation gradient, and colorimetric continuity of the T-line, and determines the validity of the test strip based on the multispectral reflectance characteristics of the C-line.

[0032] S3: Generate a continuously ranging risk index based on colorimetric feature parameters. The feature parameters are input into a pre-trained deep neural network, which outputs a risk index between 0 and 100. A risk mapping model is pre-trained from the deep neural network, with training data including a large number of sample test strip images with known infection statuses. Supervised learning training is performed by labeling infection risk levels.

[0033] S4: Upload the anonymized risk index, anonymous identifier, and spatiotemporal data to the disease control management terminal. After anonymizing the data (ID hashing, location gridding, and time coarsening) on ​​the user terminal, it is uploaded via an encrypted network.

[0034] S5: The disease control management platform constructs a dynamic risk map based on a graph neural network. Anonymous identifiers are used as nodes, and spatiotemporal co-occurrence is used as edges. Node features include risk index-related statistics, and the map is updated regularly.

[0035] Constructing a dynamic risk map specifically includes: S51: For each user's anonymous identifier, record its spatiotemporal grid code at different time periods; S52: Spatiotemporal co-occurrence is defined as a record in which two anonymous identifiers are in the same grid code and the time difference is less than a preset threshold; S53: Using anonymous identifiers as nodes, if spatiotemporal co-occurrence records exist, an edge is established, with the edge weight being the number of co-occurrences; S54: Periodically update the graph structure and remove edges that have exceeded their expiration date.

[0036] S6: When the risk index of the first node exceeds the first threshold, backtrack to the second node that has a spatiotemporal co-occurrence edge within a preset time period. Calculate the risk propagation probability of adjacent nodes using a graph neural network, and identify nodes exceeding the second threshold as risk propagation targets.

[0037] S7: Push anonymous warning information to the user terminal corresponding to the second node. The warning information includes anonymized risk alerts and detection suggestions to guide the user to perform detection.

[0038] Through the above steps, a closed loop from individual detection to group early warning is achieved, providing an intelligent tool for HIV prevention and control.

[0039] Example 6: Taking user "Zhang San" as an example: On Saturday night, Zhang San went to a bar (identified as a high-risk scenario by the behavior perception module). The next day, he received navigation information for a self-testing device from the app. He went to a vending machine to pick up a urine test strip, tested it at home, and inserted it into the smart testing box. Image processing detected a slight color change in the T-line, and the risk assessment module calculated a risk index of 65. The app displayed "Pending retest" and suggested going to the CDC for confirmation. At the same time, the anonymized data was uploaded to the CDC management terminal. GNN analysis traced back Zhang San's spatiotemporal companions over the past three weeks and found that anonymous user "X" had co-occurred with Zhang San in the same bar, with a calculated transmission probability of 0.7. Therefore, an anonymous warning message was pushed to X. After seeing the warning, X performed a test. If the test result was abnormal, the system would further track his spatiotemporal companions, forming a chain of risk transmission. Simultaneously, the resource scheduling module detected an increase in high-risk nodes in the grid where the bar was located and automatically sent replenishment instructions to the vending machines in that area.

[0040] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

[0041] For those skilled in the art, various changes and modifications will undoubtedly be apparent after reading the above description. Therefore, the appended claims should be construed as covering all changes and modifications that encompass the true intent and scope of the invention. Any and all equivalent scope and content within the scope of the claims should be considered to remain within the intent and scope of the invention.

Claims

1. An intelligent system for HIV prevention and detection, characterized in that, include: The intelligent testing kit includes a card slot and an image acquisition unit. The card slot is used to receive HIV test strips, and the image acquisition unit is used to acquire colorimetric images of the reaction area on the test strips. The user terminal is communicatively connected to the intelligent detection box. The user terminal includes an image processing module, a risk assessment module, a behavior perception module, and an intervention engine module. The image processing module is used to receive color images and extract color feature parameters. The risk assessment module is used to generate a risk index based on the color feature parameters. The behavior perception module is used to identify high-risk scenarios through sensor data. The intervention engine module is used to push intervention information based on the risk index and / or high-risk scenarios. The disease control management terminal communicates with multiple user terminals. The disease control management terminal includes a data receiving module, a risk transmission analysis module, and an early warning push module. The data receiving module is used to receive the risk index and corresponding anonymous identifier and spatiotemporal data after desensitization. The risk transmission analysis module is used to construct a risk transmission model based on a graph neural network and identify risk transmission targets. The early warning push module is used to push anonymous early warning information to the identified risk transmission targets.

2. The intelligent system according to claim 1, characterized in that, The intelligent detection box also includes a multispectral light source, and the image acquisition unit acquires color images under the illumination of the multispectral light source; the image processing module is also used to determine the validity of the test strip based on the spectral characteristics of the quality control area under the multispectral light source.

3. The intelligent system according to claim 1, characterized in that, The color rendering feature parameters include at least one of the following: color rendering width of the detection line area, color saturation gradient, and color rendering continuity index; the risk assessment module uses a neural network to map the color rendering feature parameters into a risk index with continuously taking values.

4. The intelligent system according to claim 1, characterized in that, The behavior perception module collects data through the built-in sensors of the mobile phone, and identifies high-risk scenarios after privacy calculation and desensitization. The high-risk scenarios include at least one of staying in a specific area at night and abnormal movement trajectories in multiple locations.

5. The intelligent system according to claim 1, characterized in that, The intervention information pushed by the intervention engine module includes at least one of the following: pre-exposure prophylaxis medication reminders, self-test device navigation, and testing suggestions; the intervention engine module also triggers emergency testing reminders based on medication adherence data.

6. The intelligent system according to claim 1, characterized in that, The graph neural network constructed by the risk propagation analysis module uses anonymous identifiers as nodes and spatiotemporal co-occurrence as edges. The node features include risk index-related statistics. When the risk index of the first node exceeds the threshold, the risk propagation probability of its neighboring nodes is calculated, and the neighboring nodes that exceed the propagation probability threshold are regarded as risk propagation targets.

7. A method for HIV prevention and detection based on the intelligent system according to any one of claims 1 to 6, characterized in that, Includes the following steps: S1: Acquire the colorimetric image of the HIV test strip through the smart detection box; S2: Process the color image and extract color feature parameters; S3: A risk index with continuously varying values ​​is generated based on colorimetric feature parameters; S4: Upload the desensitized risk index, anonymized label, and spatiotemporal data to the disease control management terminal; S5: The disease control management terminal is based on a graph neural network, using anonymous identifiers as nodes and spatiotemporal co-occurrence as edges to construct a dynamic risk map; S6: When the risk index of the first node exceeds the first threshold, backtrack the second node that has a spatiotemporal co-occurrence edge within a preset time period; S7: Push anonymous warning information to the user terminal corresponding to the second node.

8. The method according to claim 7, characterized in that, The extraction of colorimetric feature parameters in step S2 specifically includes: identifying the detection line area and the control line area, calculating the colorimetric width and color saturation gradient of the detection line area, and determining the validity of the test strip based on the spectral characteristics of the control line area.

9. The method according to claim 7, characterized in that, In step S3, the risk index is generated by a pre-trained deep neural network. The input of the neural network is the color feature parameters, and the output is a continuous value between 0 and 100.

10. The method according to claim 7, characterized in that, Step S5, which involves constructing a dynamic risk graph, further includes: recording the spatiotemporal grid codes of each anonymous identifier at different time periods; defining spatiotemporal co-occurrence as two anonymous identifiers sharing the same grid code and having a time difference less than a preset threshold; establishing edges and assigning weights; and calculating the risk propagation probability of adjacent nodes through the message passing mechanism of a graph neural network.

Images