Autism early intervention management integrated platform based on telemedicine

Abstract

The invention relates to the technical field of integrated platforms, in particular to an autism early intervention management integrated platform based on telemedicine, which comprises a cloud service layer, an application layer, a network adaptability module and a front-end display layer, the cloud service layer is deployed on at least one server, and the cloud service layer further comprises a child information database used for storing child personal information, an evaluation scale and behavior video multi-source heterogeneous data; the AI model service unit comprises a deep learning sub-model for behavior recognition and a statistical sub-model for evaluation scale analysis; the recommendation engine unit is used for outputting a personalized IEP target and an intervention scheme by adopting a collaborative filtering and rule matching algorithm based on the child characteristics and historical intervention data loaded in the database; according to the invention, a client network adaptability module and a local caching strategy are introduced, so that the continuity and integrity of video evaluation and interaction data under network fluctuation are ensured.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of integrated platforms, and in particular to an autism early intervention management integrated platform based on telemedicine. BACKGROUND

[0002] With the development of the Internet and artificial intelligence technology, telemedicine platforms have achieved certain results in the fields of chronic disease management and rehabilitation training; in particular, in the early intervention of autistic children, through video assessment, online questionnaires and expert guidance, a number of scattered evaluation systems and training platforms have emerged.

[0003] However, the prior art usually has the following problems:

[0004] First, the platform architecture is bloated, lacks a clear hierarchical design, and is difficult to meet the growing data storage and computing needs.

[0005] Second, the front-end display function is single, lacks process navigation and visual interaction means, resulting in high use threshold and poor experience for parents and institutions.

[0006] Third, the intervention scheme relies on manual matching, and it is difficult to achieve personalized recommendation based on large-scale historical data.

[0007] Fourth, video assessment and data synchronization are prone to interruption when the network environment changes; fifth, the development of each terminal is scattered, the maintenance cost is high, and the cross-platform consistency is poor, which restricts the management efficiency and intervention effect of the early intervention of autism. SUMMARY

[0008] The purpose of the present application is to provide an autism early intervention management integrated platform based on telemedicine to solve the problems of the prior art that:

[0009] First, the platform architecture is bloated, lacks a clear hierarchical design, and is difficult to meet the growing data storage and computing needs.

[0010] Second, the front-end display function is single, lacks process navigation and visual interaction means, resulting in high use threshold and poor experience for parents and institutions.

[0011] Third, the intervention scheme relies on manual matching, and it is difficult to achieve personalized recommendation based on large-scale historical data.

[0012] Fourth, video assessment and data synchronization are prone to interruption when the network environment changes; fifth, the development of each terminal is scattered, the maintenance cost is high, and the cross-platform consistency is poor, which restricts the management efficiency and intervention effect of the early intervention of autism.

[0013] To achieve the above object, the present application provides the following technical solutions:

[0014] The autism early intervention management integrated platform based on telemedicine comprises a cloud service layer, an application layer, a network adaptability module and a front-end display layer.

[0015] The cloud service layer is deployed on at least one server, and the cloud service layer further comprises:

[0016] A children information database is used to store children personal information, evaluation scales and behavior videos multi-source heterogeneous data.

[0017] An AI model service unit comprises a deep learning sub-model for behavior recognition and a statistical sub-model for evaluation scale analysis.

[0018] A recommendation engine unit is used to output personalized IEP targets and intervention schemes by using collaborative filtering and rule matching algorithms based on the children characteristics and historical intervention data loaded in the database.

[0019] A security management module is responsible for data encryption, access control and log auditing, and communicates with the AI model service unit and the recommendation engine through TLS.

[0020] The application layer comprises an institutional end Web application and a parent end mobile application, and data exchange is performed between the cloud service layer and the application layer through RESTful API.

[0021] The network adaptability module is deployed on the client side and is used to monitor the current network bandwidth in real time and automatically switch the code stream and local cache strategy under different bandwidth environments.

[0022] The front-end display layer comprises a step-by-step navigation bar component, a drag-and-drop card component and a chart visualization component, and realizes a five-step process operation of "evaluation, planning, training, monitoring and review".

[0023] As a further optimization of the present application, the institutional end Web application is developed based on the front-end framework React or Vue and integrates a UI component library, and the parent end mobile application is developed based on Flutter or React Native.

[0024] As a further optimization of the present application, the step-by-step navigation bar component comprises current stage highlighting, novice prompt bubble, progress indication bar and customizable stage name, the user can click the navigation to complete any jump between stages, and the unfinished stages can be displayed in gray.

[0025] As a further optimization of the present application, the recommendation engine unit adopts a prediction model based on collaborative filtering, and the prediction score formula is:

[0026]

[0027] wherein, represents the predicted fit score of user u to intervention program i, represents the average score of user u in historical projects, N(u) represents a set of neighborhood users similar to user u, w u,v represents the similarity weight between user u and user v, r v,i and respectively represent the actual score of user v to program i and its average score, the recommendation engine selects the top 3-5 personalized IEP target recommendations from high to low.

[0028] As a further optimization of the application, wherein: the drag card component supports card grouping, card search and label classification, and each card is bound with metadata, and the front-end drag, addition and deletion operations are notified through the event bus to update the cloud synchronization.

[0029] As a further optimization of the application, wherein: the chart visualization component is implemented by calling ECharts or Chart.js visualization library, including:

[0030] Gantt chart component for displaying IEP target time scheduling;

[0031] Line / Radar chart component for displaying key behavior indicator trends;

[0032] Suggestion panel for presenting next stage key targets and intervention content.

[0033] As a further optimization of the application, wherein: the AI model service unit includes a convolutional neural network sub-model for video and behavior data analysis, and a linear conversion algorithm for evaluating scale score standardization and dynamic scoring, the linear conversion algorithm is represented as:

[0034]

[0035] wherein, R represents the original evaluation scale score, μ and σ represent the mean and standard deviation of the scale in the corresponding age group, respectively, and S represents the standardized dynamic score for subsequent trend analysis and recommendation engine input.

[0036] As a further optimization of the application, wherein: the network adaptability module implements adaptive code stream switching and local data caching strategy on the client, and the switching judgment formula is:

[0037] ​

[0038] In the formula, B represents the network bandwidth measured in real time at present, B th represents a preset bandwidth threshold, and the client limits the video resolution and frame rate to R low and F low respectively in the low-code flow mode, and temporarily stores user interaction data in a local cache, and when the network recovers to B >= B th , automatically uploads the cache data and synchronizes with the cloud incrementally.

[0039] Compared with the prior art, the beneficial effects of the present application are:

[0040] 1. In the present application, the cloud service layer, the application layer, the network adaptability module and the front-end display layer are organically combined to realize high cohesion and low coupling of data storage, AI behavior recognition and evaluation, personalized recommendation and multi-end display; a collaborative filtering algorithm and a statistical standardization model are used to automatically generate accurate personalized IEP targets and intervention schemes, thereby improving intervention matching degree; a client network adaptability module and a local cache strategy are introduced to ensure the continuity and integrity of video evaluation and interaction data under network fluctuations;

[0041] 2. In the present application, the front end is constructed based on React / Vue, Flutter / React Native, cooperates with step-by-step navigation, drag-and-drop cards and diversified chart components, realizes a closed-loop process management of "evaluation -> planning -> training -> monitoring -> review", greatly reduces the use threshold and improves the operation efficiency; the overall system has good scalability, maintainability and cross-platform consistency, and effectively improves the effect and user experience of early intervention management of autism. BRIEF DESCRIPTION OF DRAWINGS

[0042] Fig. 1 is a schematic diagram of the overall structure of the present application;

[0043] Fig. 2 is a network adaptability module workflow diagram of the present application;

[0044] Fig. 3 is an AI model service unit and recommendation engine workflow diagram of the present application. DETAILED DESCRIPTION

[0045] Referring to Figs. 1-3 , the present application provides a technical solution:

[0046] The autism early intervention management integrated platform based on telemedicine includes a cloud service layer, an application layer, a network adaptability module and a front-end display layer.

[0047] The cloud service layer is deployed on at least one server, and the cloud service layer further includes:

[0048] A children information database is configured to store children personal information, assessment scales, and multi-source heterogeneous behavior video data.

[0049] An AI model service unit includes a deep learning sub-model for behavior recognition and a statistical sub-model for assessment scale analysis.

[0050] A recommendation engine unit is configured to output personalized IEP targets and intervention programs based on the loaded children characteristics and historical intervention data in the database by using collaborative filtering and rule matching algorithms.

[0051] A security management module is responsible for data encryption, access control, and log auditing, and communicates with the AI model service unit and the recommendation engine through TLS.

[0052] An application layer includes an institutional end Web application and a parent end mobile application, and exchanges data with the cloud service layer through RESTful API.

[0053] A network adaptability module is deployed on the client side to monitor the current network bandwidth in real time and automatically switch the code stream and local caching strategy under different bandwidth environments.

[0054] A front-end display layer includes a step-by-step navigation bar component, a drag-and-drop card component, and a chart visualization component, realizes a five-step process operation of "assessment -> planning -> training -> monitoring -> review", and separates data storage, AI calculation, application display, and network adaptation through a hierarchical architecture, so that the system has high scalability and high maintainability. At the same time, the network adaptability module ensures smooth operation when the network fluctuates, improving the user experience.

[0055] As a further implementation of the present scheme, the institutional end Web application is developed based on the front-end framework React or Vue and integrates a UI component library, and the parent end mobile application is developed based on Flutter or React Native, which can shorten the development cycle and ensure interface consistency. The React / Vue and Flutter / React Native communities are mature and easy to iterate and maintain subsequent functions.

[0056] As a further implementation of the present scheme, the step-by-step navigation bar component includes current stage highlighting, novice prompt bubble, progress indicator bar, and customizable stage name. Users can click on the navigation to jump between stages, and gray-scale display of unfinished stages is supported. The step-by-step navigation clearly guides users to follow the "assessment -> planning -> training -> monitoring -> review" process, reducing the use threshold. Gray-scale display helps users quickly identify the steps to be completed, improving operational efficiency.

[0057] As a further implementation of the present scheme, the recommendation engine unit adopts a prediction model based on collaborative filtering, and the prediction score formula is:

[0058]

[0059] In the formula, represents the predicted adaptation score of user u to intervention scheme i, represents the average score of user u in historical projects, N(u) represents the set of neighborhood users similar to user u, w u,v represents the similarity weight between user u and user v, r v,i and respectively represent the actual score of user v to scheme i and its average score, and the recommendation engine selects the top 3-5 as the personalized IEP target recommendation in descending order. Collaborative filtering model can accurately recommend intervention schemes according to the historical effects of similar users, enhance personalization, improve intervention matching degree and efficiency, and is easy to calculate and expand;

[0060] As a further implementation of the present scheme, the drag-and-drop card component supports card grouping, card search and label classification, and each card is bound to metadata. The front-end drag-and-drop, addition and deletion operations are notified by event bus to update the cloud end synchronously. The drag-and-drop interaction is intuitive and flexible, and the card grouping and label classification facilitate content organization. The event bus ensures real-time synchronization of front-end and back-end data, reducing manual intervention.

[0061] As a further implementation of the present scheme, the chart visualization component is implemented by calling ECharts or Chart.js visualization library, including:

[0062] Gantt chart component for displaying IEP target time scheduling;

[0063] Line / Radar chart component for displaying key behavior indicator trends;

[0064] Suggestion panel for presenting next-stage key targets and intervention content. Diversified charts intuitively present intervention plans and effect trends, helping experts and parents quickly understand key indicator changes. The suggestion panel provides decision support to shorten the feedback cycle.

[0065] As a further implementation of the present scheme, the AI model service unit includes a convolutional neural network sub-model for video and behavior data analysis, and a linear conversion algorithm for evaluating scale score standardization and dynamic scoring, which is represented as:

[0066]

[0067] In the formula, R represents the original evaluation scale score, μ and σ represent the mean and standard deviation of the scale in the corresponding age group, respectively, S represents the standardized dynamic score, which is used for subsequent trend analysis and recommendation engine input, combined with deep learning and statistical standardization algorithm, which can accurately identify behavior patterns and quantify evaluation results; the standardized score can be compared across age groups, facilitating trend analysis and continuous monitoring;

[0068] As a further implementation of the technical scheme, the network adaptability module implements adaptive code stream switching and local data caching strategy on the client side, and the switching judgment formula is:

[0069]

[0070] In the formula, B represents the current real-time measured network bandwidth, B th represents the preset bandwidth threshold, and the client limits the video resolution and frame rate to R low and F low respectively in the low code stream mode, and temporarily stores user interaction data in the local cache, and when the network recovers to B≥B th , automatically uploads the cached data and synchronizes with the cloud incrementally, so that adaptive code stream switching and caching strategy ensure video quality and smooth interaction, and even when the network is unstable, the operation can be uninterrupted; automatic incremental synchronization eliminates the risk of data loss.

[0071] Working principle: first, the parent end and the agency end respectively initiate evaluation or training requests through the front-end display layer ("evaluation" stage), and the front-end display layer combines step-by-step navigation bar, drag-and-drop cards and chart visualization components to guide users to complete related operations; the network adaptability module monitors the bandwidth in real time on the client side, automatically switches the video code stream and local caching strategy, and ensures the continuity of evaluation video and interaction data. After all data are transmitted to the cloud service layer through RESTful API, the children information database stores and preprocesses multi-source heterogeneous data such as personal archives, evaluation scales and behavior videos;

[0072] Then, the AI model service unit calls the convolutional neural network sub-model to recognize actions and emotions from behavior videos, and uses statistical sub-models and linear conversion algorithms to standardize the evaluation scale data and generate dynamic trend information; the recommendation engine unit filters out the optimal 3-5 personalized IEP targets and intervention schemes based on the collaborative filtering prediction model (in which the prediction adaptation degree of the intervention scheme is output by calculating the user similarity and historical score) and the rule matching algorithm, combined with the children's characteristics and the past intervention effect in the database;

[0073] Subsequently, the cloud will recommend results and model analysis data feedback to the front-end display layer through TLS encryption, chart visualization components call ECharts / Chart.js render Gantt chart, line / radar chart and suggestion panel, realize "plan → training → monitoring" three-step visualization operation, in the training process, the front-end drag-and-drop card component and event bus mechanism, support flexible adjustment and cloud synchronization of intervention module; after network recovery, the data cached locally will be automatically uploaded incrementally, ensuring that the cloud and the client data are consistent;

[0074] Finally, the system in the "review" stage, through the behavior trend analysis report generated by the AI model and the adjustment suggestion of the recommendation engine, guides the experts and parents to review and optimize the intervention effect, completes the closed-loop early intervention management from "evaluation → plan → training → monitoring → review".

[0075] The principles and implementation modes of the present application are described in this paper by applying specific examples. The above examples are only used to help understand the method of the present application and its core idea. The above description is only the preferred embodiment of the present application. It should be pointed out that due to the limitation of language expression, there are infinite specific structures. For ordinary skilled persons in the technical field, without departing from the principles of the present application, some improvements, refinements or changes can be made, or the above technical features can be combined in a proper way. These improvements, refinements, changes or combinations, or the direct application of the inventive concept and technical solution to other occasions without improvement, shall be regarded as the protection scope of the present application.

Claims

1. An integrated platform for autism early intervention management based on telemedicine, characterized in that, The platform comprises a cloud service layer, an application layer, a network adaptability module and a front-end display layer; The cloud service layer is deployed on at least one server, and further comprises: a children information database for storing children personal information, assessment scales and multi-source heterogeneous behavior video data; an AI model service unit comprising a deep learning sub-model for behavior recognition and a statistical sub-model for assessment scale analysis; a recommendation engine unit for outputting personalized IEP targets and intervention programs based on the children characteristics and historical intervention data loaded in the database by using collaborative filtering and rule matching algorithms; a security management module responsible for data encryption, access control and log auditing, and communicating with the AI model service unit and the recommendation engine through TLS; The application layer comprises an institutional end Web application and a parent end mobile application, and exchanges data with the cloud service layer through RESTful API; The network adaptability module is deployed on the client side for real-time monitoring of the current network bandwidth and automatic switching of code streams and local caching strategies under different bandwidth environments; The front-end display layer comprises a step-by-step navigation bar component, a drag-and-drop card component and a chart visualization component, realizing a five-step process operation of "assessment → planning → training → monitoring → review". 2.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The institutional end Web application is developed based on the front-end frameworks React or Vue and integrates UI component libraries, and the parent end mobile application is developed based on Flutter or React Native. 3.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The step-by-step navigation bar component comprises current stage highlighting, novice prompt bubble, progress indicator bar and customizable stage name, and the user can click on the navigation to complete any jump between stages, and the unfinished stages can be displayed in gray. 4.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The recommendation engine unit adopts a prediction model based on collaborative filtering, and the prediction score formula is: wherein, represents the predicted fit score of user u to intervention i, represents the average score of user u in historical projects, N (u) represents a set of neighborhood users similar to user u, w u,v represents the similarity weight between user u and user v, r v,i and respectively represent the actual score of user v to intervention i and its average score, the recommendation engine selects the top 3-5 as the personalized IEP target recommendation. Sort from high to low, select the top 3-5 as the personalized IEP target recommendation. 5.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The drag-and-drop card component supports card grouping, card searching and label classification, and each card is bound to metadata, and the front-end drag-and-drop, addition and deletion operations are notified through an event bus to update the cloud side synchronously. 6.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The chart visualization component is realized by calling the ECharts or Chart.js visualization library, and comprises: a Gantt chart component for displaying IEP target time scheduling; a line chart / radar chart component for displaying key behavior indicator trends; a suggestion panel for presenting next-stage key targets and intervention content. 7.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The AI model service unit comprises a convolutional neural network sub-model for video and behavior data analysis, and a linear conversion algorithm for assessment scale score standardization and dynamic scoring, and the linear conversion algorithm is represented as: wherein R represents the original assessment scale score, μ and σ represent the mean and standard deviation of the scale in the corresponding age group, and S represents the standardized dynamic score for subsequent trend analysis and recommendation engine input. 8.The telemedicine-based integrated platform for early intervention management of autism according to claim 1, wherein: The network adaptability module realizes adaptive code stream switching and local data caching strategies on the client side, and the switching judgment formula is: In the formula, B represents the current real-time measured network bandwidth, B th represents the preset bandwidth threshold, and the client limits the video resolution and frame rate to R low and F low respectively in the low code stream mode, and temporarily stores the user interaction data in the local cache. When the network recovers to B≥B th , the cache data is automatically uploaded and incrementally synchronized with the cloud.

Images