Infant neural development intelligent screening system and method based on AI closed-loop control

Abstract

The invention provides an intelligent screening system and method for infant neural development, and belongs to the field of medical information technology and intelligent hardware. The system comprises a data acquisition device, a user terminal and an evaluation processing platform. Intelligent evaluation software in the evaluation processing platform can analyze video data acquired by a high-definition camera of the data acquisition equipment in real time, generate a control instruction according to an analysis result, and send the instruction to a driving mechanism to drive the high-definition camera to move to a target shooting position determined based on the analysis result, so that the target shooting position is shot. And closed-loop intelligent control of the acquisition process is realized. Meanwhile, the software also analyzes the collected video to evaluate the neural development condition of the infant and generate an evaluation report. Through AI closed-loop control hardware, the quality and standardization level of video acquisition are guaranteed from the source, intelligence and automation of data acquisition are realized, the problem of inaccurate evaluation results caused by non-standard video acquisition is solved, and the screening efficiency is improved.

Description

Technical Field

[0001] This invention relates to the field of medical information technology and smart hardware, and in particular to an intelligent screening system and method for infant neurodevelopment based on artificial intelligence closed-loop control. Background Technology

[0002] Early screening for infant neurodevelopment, particularly universal motor assessment for high-risk infants, is crucial for promoting healthy child growth. The accuracy of universal motor assessment is highly dependent on the quality and standardization of the video data collected. While various video acquisition methods exist in the current technology, all have limitations.

[0003] A common approach involves using parents' mobile devices (such as smartphones) to capture videos of infants' activities and upload them to an assessment platform. For example, patent document CN113903468A discloses a method for ultra-early intelligent screening and monitoring of high-risk infants based on a mobile terminal. This method involves collecting videos of infants' natural body movements while lying flat; uploading the videos to a cloud server for analysis; extracting the movement trajectories of the infants' limbs and torso through computer image processing; quantifying the movement trajectory parameters; and scoring the infants' movement status based on the quantification results. This optimizes the assessment of the infants' overall movement, making it simpler and easier to implement. Parents can collect images at home using their smartphones, quickly upload them to the cloud, and have the system automatically analyze the images for quality and make a preliminary analysis before a report is generated and verified manually. However, while this method is convenient, the quality of the videos varies greatly due to differences in shooting equipment, ambient lighting, shooting angles, and operating techniques. These videos often fail to meet professional and universal movement assessment standards, leading to inaccurate or even invalid assessment results and delaying the optimal time for early intervention.

[0004] Another approach involves using specialized medical imaging equipment, such as a camera system integrated into a medical cart. For example, patent document CN113642525A discloses a method and system for assessing infant neurodevelopment based on skeletal points. S1: Capture and collect videos of the infant's movements, and annotate the videos using a whole-body motion assessment method; S2: After data collection, extract skeletal point information based on posture estimation, and perform data cleaning and segmentation; S3: Train and test the skeletal point-based motion recognition model using few-shot learning from the field of transfer learning; S4: Determine whether further diagnosis and treatment are needed based on the model's prediction results. However, such equipment is typically used only as a simple hardware acquisition tool, lacking deep integration with backend assessment software and intelligent control. The video acquisition process still heavily relies on the operator's experience and skill, making it difficult to guarantee the consistency of data collected at different times and by different operators, affecting the reliability of longitudinal comparisons and cross-sectional analyses of the assessment.

[0005] Therefore, the market currently lacks a solution that can seamlessly integrate standardized hardware data collection, AI-powered intelligent analysis, remote expert evaluation, and convenient family participation. In particular, there is a lack of a closed-loop intelligent screening system that can proactively analyze video content through AI and control the hardware in reverse to ensure data collection quality. This means that the overall efficiency and accuracy of infant neurodevelopment screening need to be improved. Summary of the Invention

[0006] To address the shortcomings of existing technologies, the present invention aims to provide an intelligent screening system and method for infant neurodevelopment based on artificial intelligence closed-loop control. This system addresses the technical problems in existing technologies, such as inconsistent video acquisition equipment, non-standardized operation leading to unreliable video quality, and the lack of proactive and intelligent quality control from the data acquisition source. The goal is to improve the standardization, efficiency, and accuracy of infant neurodevelopment screening.

[0007] According to the present invention, an intelligent screening system for infant neurodevelopment includes: a data acquisition device comprising a movable arm, a high-definition camera mounted on the arm, and a drive mechanism for moving the arm and / or the high-definition camera; a user terminal for allowing a user to input infant information and upload a first video; an assessment processing platform communicatively connected to the data acquisition device and the user terminal, the assessment processing platform comprising: intelligent assessment software for infant neurodevelopment behavior; and a PACS system for storing the infant information, the first video, and a second video acquired by the high-definition camera, wherein the PACS system is configured to store data in a format required for general motor assessment; wherein the intelligent assessment software is configured to: analyze video data acquired in real time by the high-definition camera and generate control commands based on the analysis results; send the control commands to the drive mechanism to drive the high-definition camera to move to a target shooting position determined based on the analysis results; and analyze the second video to assess the infant's neurodevelopment status and generate an assessment report.

[0008] Preferably, the intelligent evaluation software is further configured to perform a quality pre-inspection on the first video uploaded through the user terminal.

[0009] Preferably, the intelligent evaluation software is further configured to: during the acquisition of the second video, if, based on real-time analysis of the second video, it is determined that the second video lacks key motion frames of preset body parts, then automatically control the drive mechanism to perform supplementary shooting.

[0010] Preferably, the data acquisition device further includes an all-in-one computer, the intelligent evaluation software runs on the all-in-one computer, and is configured to: provide real-time feedback on the height, angle, or distance parameters of the high-definition camera on the display interface of the all-in-one computer, and display the deviation information between the high-definition camera and the standard acquisition position on the display interface.

[0011] Preferably, the evaluation processing platform includes an evaluation mode switching module for switching between AI-assisted evaluation mode and expert remote evaluation mode; and the evaluation processing platform is configured to provide the AI ​​evaluation results as reference information to remote experts for comprehensive evaluation.

[0012] Preferably, the drive mechanism includes a rotary motor and an electric push rod; and the data acquisition device also includes a column, a chassis, medical silent wheels, a battery enclosure, and an all-in-one computer.

[0013] Preferably, the evaluation processing platform further includes a verification mechanism to ensure that the infant information uploaded through the user terminal is accurately bound to the second video collected through the data acquisition device.

[0014] This invention also provides an intelligent screening method for infant neurodevelopment based on AI closed-loop control, comprising the following steps: receiving infant information and a first video uploaded by a user via a user terminal; acquiring a second video of the infant using a data acquisition device, the data acquisition device including a movable support arm, a high-definition camera mounted on the support arm, and a drive mechanism for driving the support arm and / or the high-definition camera to move; generating control commands based on real-time analysis of the second video; sending the control commands to the drive mechanism of the data acquisition device to drive the high-definition camera to move to a target shooting position determined based on the real-time analysis results; storing the infant information, the first video, and the second video in a PACS system, wherein the PACS system is configured to store data in a format required by general motion assessment; and evaluating the second video and generating a report.

[0015] Preferably, the method further includes: during the acquisition of the second video, if, based on real-time analysis of the second video, it is determined that the second video lacks key motion frames of preset body parts, then the drive mechanism is automatically controlled to perform reshooting.

[0016] Preferably, the method further includes: performing a quality pre-inspection on the first video uploaded through the user terminal.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention employs standardized data acquisition equipment and utilizes artificial intelligence algorithms to actively control the camera for positioning, ensuring the consistency and professionalism of the acquired video in terms of angle, distance, and clarity, thus laying a solid foundation for subsequent accurate evaluation.

[0018] 2. This invention uses a closed-loop mechanism that reverse-controls hardware based on artificial intelligence analysis results, freeing operators from complex positioning tasks, reducing human error, and achieving automated and standardized video acquisition.

[0019] 3. This invention significantly improves the overall efficiency of screening and the accuracy of diagnosis by integrating pre-hospital screening, standardized in-hospital data collection, and a hybrid assessment model that combines AI-assisted assessment with remote expert assessment.

[0020] 4. This invention allows users to conveniently submit information, receive reports, and receive follow-up appointment reminders via a user terminal, making the screening process more convenient and transparent, and improving the satisfaction of medical services. Attached Figure Description

[0021] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a schematic diagram of the structure of a data acquisition device provided in an embodiment of the present invention; Figure 2 A business process diagram for infant neurodevelopment screening provided in an embodiment of the present invention; Figure 3 A schematic diagram of the layered system software architecture provided in this embodiment of the invention; Figure 4 A flowchart illustrating the AI ​​closed-loop control method provided in an embodiment of the present invention; Figure 5 This is a timing diagram illustrating the signaling interaction between various modules of the system provided in an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several changes and improvements without departing from the concept of the present invention. These all fall within the scope of protection of the present invention.

[0023] Example 1 This invention provides an intelligent screening system and method for infant neurodevelopment based on artificial intelligence closed-loop control. The system organically combines standardized data acquisition hardware, a multi-layered intelligent software platform, and a user-facing mobile terminal, thereby constructing a closed-loop management system covering the entire process from pre-hospital preparation, standardized in-hospital data collection, multi-mode intelligent assessment to report distribution and follow-up. This invention aims to solve the problems of insufficient screening efficiency and accuracy caused by inconsistent video acquisition quality, cumbersome processes, and limited assessment modes in existing technologies.

[0024] Reference Figures 1 to 3 .in, Figure 1 This is a schematic diagram of the data acquisition device used in this embodiment; Figure 2 This is a flowchart illustrating the business process for using this system for screening. Figure 3 This is a schematic diagram of the software architecture that supports the operation of the system.

[0025] Specifically, the intelligent screening system for infant neurodevelopment provided in this embodiment includes data acquisition equipment, user terminals, and an evaluation and processing platform.

[0026] The data acquisition device, such as Figure 1As shown, this can be a fully integrated medical cart designed to provide stable, flexible, and intelligent video acquisition capabilities. The device includes a sturdy chassis and multiple silent medical wheels to ensure smooth and convenient movement in quiet environments such as hospitals. A column is mounted on the chassis, integrating a battery compartment, a work surface, and an all-in-one computer. The battery compartment can house a high-capacity lithium battery system (e.g., 39AH), enabling continuous operation for over 5 hours without external power, thus enhancing mobility and versatility. The work surface provides space for medical personnel to place relevant items. The all-in-one computer serves as the local computing and interaction core, running on a doctor's end of intelligent assessment software for infant neurodevelopment and behavior.

[0027] As the core component of this data acquisition device, the image acquisition assembly is software-controlled. This assembly includes a movable support arm, a high-definition camera mounted at the end of the support arm, and a drive mechanism for its movement. In this embodiment, the drive mechanism specifically consists of a rotary motor and an electric push rod. The rotary motor drives the support arm to rotate 360 ​​degrees horizontally and can precisely hover at any angle, while the electric push rod controls the raising and lowering of the support arm, thereby changing the height of the high-definition camera. Both the rotary motor and the electric push rod support forward and reverse rotation control and self-locking functions to ensure the stability of the posture after positioning. As a preferred embodiment, the high-definition camera can be a high-performance camera with a wide-angle (e.g., 70 degrees), autofocus, and ultra-low-light shooting capabilities to ensure clear and wide field of view under various lighting conditions.

[0028] In this embodiment, the user terminal can be a convenient mobile application, such as a WeChat mini-program (e.g., ...). Figure 5 The parent app shown here could be a web application or another form of application. Users (usually parents of infants and toddlers) can use this user terminal to complete the preparations before screening.

[0029] The assessment and processing platform is the core processing unit of the entire system. Logically, the platform includes intelligent assessment software for infant neurodevelopment and behavior and a PACS system. In terms of physical deployment, some functions of the platform (such as the doctor's interactive interface and device control logic) run on the all-in-one computer of the data acquisition device, while its core data processing, storage, and artificial intelligence computing functions can be deployed on cloud servers or the hospital's local server cluster.

[0030] Combination Figure 3The software architecture of the assessment processing platform is described below. This architecture adopts a layered design, consisting of a data layer, service layer, application layer, and presentation layer from bottom to top. The data layer is used to achieve persistent storage of all data. For example, a cache database cluster can be used, divided into a non-data-isolated database for storing shared information such as organizational structure and user roles, and an isolated database for storing sensitive information such as medical videos, patient information, and medical records, to ensure data security. The service layer provides a series of reusable microservices, which may include, but are not limited to, organization management services, user role management services, security services, data services, patient services, and core AI-assisted assessment services and remote expert assessment services. The application layer is built on top of the service layer to implement specific business logic. It includes a business management platform responsible for managing institutions, doctors, patients, medical records, and assessment tasks, and an AI detection platform embedded with various artificial intelligence models (such as analysis models for the twisting and restless movement stages in infant neurodevelopment assessment). The presentation layer serves as the entry point for user interaction with the system. It includes a web platform for doctors and institutions, a WeChat mini-program for parents, and device software running on the all-in-one computer of the data acquisition device.

[0031] The PACS system serves as part of the data and service layers, its function being to uniformly store and manage all data generated through the system, including infant information and the first video uploaded by user terminals, and the second video captured by data acquisition devices. It should be noted that the PACS system is configured to store and manage data according to the data formats and standards required by General Motion Assessments (GMs). For example, the naming rules for video files, metadata tagging, and storage structure all follow GMs assessment specifications, thus laying the foundation for subsequent standardized assessments and data analysis.

[0032] The following is combined Figure 2 and Figure 5 The process of the screening method in this embodiment is described in detail below: First, pre-hospital preparation and information pre-entry steps are performed. Parents can register and log in via a user terminal (e.g., a WeChat mini-program) and enter the child's basic information (such as name, date of birth, gestational age, etc.) according to the interface instructions. The application then provides video tutorials to guide parents in recording a short video of the child's activities at home that meets the basic requirements; this video serves as the first video. After recording, parents upload the infant's information and the first video via the mini-program. At this point, the intelligent assessment software in the assessment processing platform (specifically, the AI ​​detection platform in the application layer) automatically performs a quality pre-check on the uploaded first video. This pre-check algorithm analyzes multiple dimensions, including video duration, lighting, stability, and the proportion of the infant in the frame, to determine if it meets the basic requirements of the initial assessment. If it is deemed unacceptable, the system sends a notification to the parents via the mini-program explaining the reason for the failure and requesting a re-upload; if the quality pre-check passes, the system stores the infant's information and the first video in the PACS system, automatically generates a unique patient number, and then sends feedback to the parents via the mini-program. This step enables preliminary screening of pre-hospital data, which can avoid invalid medical visits due to unqualified documents submitted by family members, thereby optimizing the medical treatment process.

[0033] Next, the standardized in-hospital data collection procedure is executed. Parents bring their child to the hospital and present the patient's medical record number to the attending physician. The physician enters this number into the doctor's software running on the all-in-one computer of the data collection device. The system can then retrieve all pre-recorded infant information and the first video from the PACS system, eliminating the tedious process of repeated on-site data entry. Subsequently, the physician places the infant on an assessment bed and uses the data collection device described in this embodiment to perform professional video capture; the captured video serves as the second video. The collection process itself is also intelligent, and its specific implementation will be detailed in Embodiment 2. After collection, the second video is automatically uploaded and stored in the PACS system. To ensure the uniqueness and accuracy of the data, the system also includes a verification mechanism. For example, through the unique medical record number, the infant information uploaded before hospitalization, the first video, and the second video collected in-hospital are precisely bound together to ensure that all data accurately belongs to the same patient file, which is crucial for ensuring the rigor of medical data.

[0034] Then, the mixed-mode assessment and diagnosis steps are performed. After the second video is acquired, the doctor can flexibly select the assessment mode on the doctor's software interface according to the child's specific condition and treatment needs. The system provides an assessment mode switching module, supporting doctors to select or combine different modes. As an optional implementation, the doctor can activate the AI-assisted assessment mode. When the doctor clicks the corresponding button, the assessment task is sent to the AI ​​detection platform. Accordingly, the AI ​​platform obtains the second video data from the PACS system and calls the corresponding algorithm model for analysis, generating a structured preliminary assessment report in a short time (e.g., a few minutes). This report may include quantitative analysis and abnormal prompts for key indicators such as the amplitude of restless movements and the frequency of twisting movements. Alternatively, in another embodiment, the doctor can also activate the expert remote assessment mode, sending the assessment task, including the second video and complete medical records, to the contracted remote expert with one click.

[0035] Finally, the report generation and distribution process begins. The doctor, combining the preliminary AI report, expert assessments, or their own clinical experience, generates a final clinical diagnostic report. This report can be printed locally using a connected printer, and the system also pushes an electronic version to the parent's user terminal (WeChat mini-program) via the assessment processing platform for convenient access. Furthermore, the system can automatically set follow-up appointment dates based on the diagnostic results and send follow-up reminders to parents via their user terminals.

[0036] In summary, through the above-described embodiments, the technical solution of the present invention integrates the originally scattered and non-standardized screening process into an efficient, closed-loop intelligent process, which not only ensures data quality from the source, but also significantly improves the experience of both doctors and patients and the overall screening efficiency.

[0037] Example 2 This embodiment, based on Embodiment 1, elaborates on one of the core innovations of the present invention: how the intelligent evaluation software, through closed-loop control, actively guides or automatically controls the hardware of the data acquisition device to achieve precise adjustment of the high-definition camera's shooting position, thereby ensuring that the acquired second video always meets the stringent standards of General Motion (GMs) evaluation. It is understood that this function significantly reduces reliance on operator experience and achieves standardization and automation of the acquisition process.

[0038] Reference Figure 1 and Figure 4 ,in Figure 4This is a flowchart illustrating the AI ​​closed-loop control method in this embodiment. This function can be collaboratively performed by the real-time posture analysis module and the device control module within the intelligent assessment software for infant neurodevelopment and behavior running on an all-in-one computer. The real-time posture analysis module utilizes computer vision technology, such as a deep learning-based skeletal keypoint recognition algorithm, to accurately detect and track the infant's body contour, overall position, orientation, and key nodes of each limb from real-time video streams captured by a high-definition camera. The device control module receives the output from the posture analysis module and converts it into specific control commands for the drive mechanisms (i.e., rotary motors and electric actuators) on the data acquisition device.

[0039] In one embodiment of the present invention, the AI ​​closed-loop control function can provide at least two operating modes to adapt to different clinical scenarios and operating habits: intelligent guidance mode and fully automatic positioning mode.

[0040] Mode 1: Intelligent Guidance Mode. When a doctor selects this mode for video capture, the operation process includes: After the doctor starts the capture program, the all-in-one computer's display not only shows the image captured by the high-definition camera in real time, but also overlays auxiliary visual elements generated by the software. These elements include a virtual frame representing the standard shooting area, and a set of real-time updated parameter prompts. For example, the interface may display text or graphical deviation information such as "Current height: 80cm, Target height: 100cm" or "Please rotate approximately 15 degrees to the left." It should be noted that this deviation information is calculated by the intelligent assessment software based on the analysis results of the infant's position by the real-time posture analysis module, comparing it with the preset Universal Movements (GMs) assessment standard shooting posture in the system (e.g., the camera should be directly above the infant's body, 1 meter away from the bed surface, with the lens pointing vertically downwards). Based on this intuitive, real-time feedback on the screen, the doctor can manually control the rotary motor and electric push rod by operating the control handle or virtual buttons on the software interface to fine-tune the position of the support arm and high-definition camera. During the adjustment process, parameter prompts will be continuously refreshed until all parameters (such as height, angle, and distance) reach the standard range. At this point, the prompt message will turn green or display confirmation information such as "Position Standard". Afterward, the doctor can begin formal video recording. In this mode, the system acts as an intelligent navigator, delegating the complex task of spatial location determination to AI. The operator only needs to follow simple instructions, thus significantly improving the accuracy and consistency of manual positioning.

[0041] Mode Two: Fully Automatic Positioning Mode. The operation process includes: after the doctor places the infant in the approximate center of the assessment bed, they simply click the "Automatic Positioning" button on the software interface to trigger the automatic positioning process. Figure 4The AI ​​closed-loop control process is shown below. Specifically, the system first enters the step of starting the AI ​​control mode (S401). Then, the high-definition camera begins to acquire real-time video frames (S402) and transmits them to the real-time posture analysis module. This module analyzes the video frames to identify the infant's main position and posture (S403). Subsequently, the system enters a judgment loop, first determining whether the current camera position meets the preset standard (S404). Initially, the position usually does not meet the standard. Accordingly, the process enters the "No" branch, and the device control module will accurately calculate the amount of motion required by the drive mechanism based on the posture analysis results and the preset general motion (GMs) standard overhead pose, such as the angle that the rotary motor needs to rotate and the length that the electric push rod needs to extend (S405). After the calculation is completed, the device control module generates the corresponding control command (S406) and sends it to the rotary motor and electric push rod via internal bus or wireless communication. After receiving the command, the drive mechanism immediately drives the arm and the high-definition camera to move (S407). During the movement, the system continuously repeats the cycle from S402 to S407, that is, continuously acquiring video frames at new locations, analyzing posture, determining position, calculating new adjustment amounts, and driving movement, thus forming a real-time closed-loop feedback control. This process continues until the judgment result of step S404 is "yes," indicating that the position, angle, distance, and other parameters of the high-definition camera have fully met the preset Universal Motion (GMs) standard shooting position. At this point, the system automatically locks the drive mechanism and can automatically start formal video recording (S408). The entire positioning process is completely driven by AI without human intervention, thus ensuring that each acquisition is performed under optimal and standardized conditions, thereby fundamentally guaranteeing the data quality of the second video.

[0042] Through the two modes described above, the solution provided in this embodiment transforms traditional passive video recording into an active, AI-enabled quality control process, thereby effectively solving the problem of data inconsistency caused by differences in operator experience.

[0043] Example 3 This embodiment, building upon Embodiments 1 and 2, further elaborates on the in-depth application of the present invention in AI closed-loop control, namely, the AI-driven dynamic supplementary shooting function. This function aims to address the problem of incomplete data collection of certain key movements during a single, fixed-position shooting session, which may be caused by infants' independent activities (such as rolling over or limb obstruction). By monitoring data integrity in real time and dynamically adjusting the shooting angle for supplementary shooting during the recording process, this solution can significantly improve the success rate and data validity of a single screening session.

[0044] Refer again Figure 4This flowchart is also applicable to describing the core logic of this embodiment, especially the parts involving steps S408, S409 and S410.

[0045] In this embodiment, the AI ​​detection platform within the evaluation processing platform, in addition to the real-time posture analysis module described in Embodiment 2, also integrates a "data integrity assessment" sub-module, which is crucial for realizing the dynamic reshoot function. This sub-module predefines the data acquisition requirements for the movement performance of various parts of the infant's body in the General Motion (GMs) assessment. It is understood that these requirements are quantifiable; for example, the evaluation model needs to observe and analyze sufficient and unobstructed autonomous movement footage of all preset body parts of the infant within a specific time window (e.g., a 3-minute recording), including the head, torso, left upper limb, right upper limb, left lower limb, and right lower limb. "Sufficient" can be defined as the cumulative effective movement duration of each part reaching a certain threshold, or capturing key movements of a specific pattern (e.g., twisting movements).

[0046] The dynamic supplementary shooting process may include the following steps: After completing the initial positioning through any of the modes described in Example 2 and starting the formal second video recording (corresponding to...) Figure 4 In step S408, the data integrity assessment submodule and the real-time posture analysis module begin to work together. The real-time posture analysis module continuously tracks the key points of the infant's entire skeleton and the movement trajectories of various body parts in the video footage, while the data integrity assessment submodule analyzes and statistically processes the collected data in real time in the background (corresponding to...). Figure 4 Step S409 in the process, namely "monitoring data integrity".

[0047] This submodule maintains a "data collection completeness" metric for each body part that needs to be evaluated. As a specific scenario, suppose that after 2 minutes of recording, the submodule analyzes and finds that because the infant has been in a side-lying position, his left arm is covered by his torso for most of the time, resulting in a serious lack of effective motion data for the left arm, and its "data collection completeness" metric is far below the preset threshold.

[0048] At this point, if the judgment result of step S409 is "No / Data incomplete", the system will automatically trigger and execute the reshoot logic (step S410). The specific execution process is as follows: First, the system can pause the timer of the main recording process to ensure that the total effective evaluation time is not affected by the reshoot process. Then, the system enters the reshoot decision stage, at which point the process can jump back to step S405 (calculating adjustment amount). The data integrity assessment submodule will instruct the device control module to obtain a clear view of the left arm. Accordingly, the device control module, combined with the infant's current posture information provided by the real-time posture analysis module, calculates an optimal reshoot position. For example, it calculates that the outrigger needs to rotate horizontally to the right by 20 degrees to avoid obstruction by the torso, thereby clearly capturing the movement of the left arm. Next, the device control module generates and sends corresponding control commands (step S406) to the rotary motor, which drives the outrigger to move to the calculated reshoot position (step S407). After reaching the reshoot position, the system begins supplementary recording, while the data integrity assessment submodule focuses on monitoring the movement data of the left arm. Once the accumulated effective movement data of the left arm collected during the reshoot reaches the required level, the reshoot ends. After the reshoot is completed, the system executes steps S405, S406, and S407 again, controlling the camera to automatically return to the original standard overhead shooting position. After the camera returns to its original position, the system resumes the timer for the main recording process and continues regular video recording (returning to step S408).

[0049] Understandably, the entire supplementary imaging process is completed automatically and quickly by AI, with minimal interference to the doctor's operation, and can even be performed seamlessly. Through this intelligent dynamic data supplementation mechanism, the solution in this embodiment can proactively respond to various unexpected situations that may arise during the acquisition process, ensuring that the final obtained second video data is complete and valid. This avoids assessment failures or the need to arrange for a second examination for the child due to incomplete data, thereby greatly improving the efficiency and success rate of clinical screening.

[0050] Example 4 This embodiment focuses on illustrating the high flexibility of the present invention's evaluation mode, and how the system organically combines the efficient analytical capabilities of artificial intelligence with the authoritative diagnostic experience of human experts to form a human-machine collaborative, efficient, and accurate hybrid evaluation mode. It is understood that this mode fully utilizes the system's multi-terminal and multi-module data fusion and signaling interaction capabilities.

[0051] Reference Figure 5 The diagram is a sequence diagram that clearly shows the order in which the various participating entities in the system—including the parent mini-program, the doctor's end, the AI ​​assessment platform, the expert end, and the PACS system—interact and call information to complete a complex collaborative assessment task.

[0052] In this embodiment, the assessment processing platform integrates an assessment mode switching module, which allows doctors to freely select or combine different assessment paths on the doctor's software interface. Furthermore, in the system's application layer, the AI ​​detection platform and the business management platform are designed to be tightly coupled, enabling the AI ​​assessment service to be executed as an independent, invoked task, and its structured assessment results to be seamlessly used as input attachments for expert remote assessment tasks.

[0053] A typical human-machine collaborative comprehensive evaluation process may include the following steps: Step 1: After the doctor uses the data acquisition device to complete the second video capture of the infant and uploads it to the PACS system, they can view the child's medical records on the doctor's interface. Considering the complexity of general motor function (GM) assessment, the doctor may decide to first use AI for a rapid preliminary screening. To do this, the doctor selects the "AI-assisted assessment" mode through the assessment mode switching module.

[0054] Step 2: Accordingly, the doctor sends a signaling request to the AI ​​assessment platform to request AI assessment. This request contains the unique identifier of the video to be assessed in the PACS system.

[0055] Step 3: After receiving the request, the AI ​​evaluation platform sends a request to the PACS system to obtain video data based on the video identifier.

[0056] Step 4: After acquiring the video data, the AI ​​models within the AI ​​assessment platform (such as the restlessness movement analysis model and the twisting movement analysis model) begin in-depth analysis of the video, completing the analysis within minutes and generating a structured preliminary assessment report. It should be noted that this report not only includes quantitative data on various movement indicators (such as movement frequency, amplitude, and symmetry), but also highlights suspicious indicators or abnormal patterns identified by the AI ​​model, such as "restlessness movement amplitude is significantly lower than the normal range" or "twisting movement pattern shows spastic synchronicity." Subsequently, the AI ​​assessment platform returns this preliminary report, containing quantitative data and suspicious point markings, to the doctor.

[0057] Step 5: The doctor reviews the preliminary AI report on their end. This report helps the doctor quickly locate the most noteworthy segments from a video lasting several minutes and provides objective quantitative references. Based on the AI's suggestions, the doctor can determine that the child may have a high risk of neurodevelopmental abnormalities, requiring final confirmation from a higher-level expert. The doctor can then select "Request Remote Expert Assessment" on their end. The system will then create a remote assessment task. When creating this task, the system will automatically link the child's electronic medical record and the original second video, and will also automatically package the preliminary assessment report generated by the AI ​​platform as an attachment. This information-rich assessment task package is then sent to the designated remote expert.

[0058] Step 6: The remote expert receives and opens the assessment task on their expert-side software interface. At this point, the expert sees an integrated view that simultaneously presents the child's complete medical record, the original second video available for playback and review, and the AI ​​preliminary assessment report. While reviewing the video, the expert can refer to the abnormal time points and indicators marked in the AI ​​report, directly jump to key segments of the video for focused observation, or use the AI's quantitative data to verify their qualitative judgment. Understandably, this method greatly improves the expert's review efficiency and diagnostic accuracy, allowing them to focus more on interpreting complex and subtle patterns and making comprehensive diagnostic decisions, rather than watching every video from beginning to end. After completing the assessment, the expert writes their diagnostic opinion on their expert-side interface and sends it back to the doctor's interface.

[0059] By providing AI assessment results as reference information to remote experts for comprehensive evaluation, this embodiment achieves a complementary advantage between artificial intelligence and human experts. AI handles the tedious and repetitive preliminary quantitative analysis, acting as an efficient "screening assistant"; while experts, with AI assistance, focus on higher-level diagnostic decisions requiring rich experience and comprehensive judgment. This human-machine collaborative hybrid assessment model is an effective way to improve the quality and efficiency of infant neurodevelopment screening.

[0060] Those skilled in the art will understand that, besides implementing the system and its various devices, modules, and units provided by this invention in the form of purely computer-readable program code, the same functions can be achieved entirely through logical programming of the method steps, making the system and its various devices, modules, and units of this invention function in the form of logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded microcontrollers. Therefore, the system and its various devices, modules, and units provided by this invention can be considered as a hardware component, and the devices, modules, and units included therein for implementing various functions can also be considered as structures within the hardware component; alternatively, the devices, modules, and units for implementing various functions can be considered as both software modules implementing the method and structures within the hardware component.

[0061] Specific embodiments of the present invention have been described above. It should be understood that the present invention is not limited to the specific embodiments described above, and those skilled in the art can make various changes or modifications within the scope of the claims, which do not affect the essence of the present invention. Unless otherwise specified, the embodiments and features described in this application can be arbitrarily combined with each other.

Claims

1. An intelligent screening system for infant neurodevelopment based on AI closed-loop control, characterized in that, include: Data acquisition equipment, user terminals, and evaluation and processing platform; The data acquisition device includes a movable support arm, a high-definition camera mounted on the support arm, and a drive mechanism for driving the support arm and / or the high-definition camera to move. The user terminal is used for users to input infant information and upload the first video; The assessment and processing platform is communicatively connected to the data acquisition device and the user terminal. The assessment and processing platform includes: intelligent assessment software for infant neurodevelopment and behavior and a PACS system, used to store the infant information, the first video and the second video captured by the high-definition camera. The PACS system is configured to store data in the format required for general motion assessment. The intelligent assessment software is configured to: analyze video data collected in real time by the high-definition camera and generate control commands based on the analysis results; send the control commands to the drive mechanism to drive the high-definition camera to move to the target shooting position determined based on the analysis results; and analyze the second video to assess the neurodevelopmental status of the infant and generate an assessment report.

2. The intelligent screening system for infant neurodevelopment based on AI closed-loop control according to claim 1, characterized in that, The intelligent assessment software is also configured to: A quality pre-inspection is performed on the first video uploaded through the user terminal.

3. The intelligent screening system for infant neurodevelopment based on AI closed-loop control according to claim 1, characterized in that, The intelligent assessment software is also configured to: During the acquisition of the second video, if, based on real-time analysis of the second video, it is determined that the second video lacks key motion shots of preset body parts, the drive mechanism is automatically controlled to perform supplementary shooting.

4. The intelligent screening system for infant neurodevelopment based on AI closed-loop control according to claim 1, characterized in that, The data acquisition device also includes an all-in-one computer, on which the intelligent evaluation software runs and is configured to: The height, angle, or distance parameters of the high-definition camera are fed back in real time on the display interface of the all-in-one computer, and the deviation information between the high-definition camera and the standard acquisition position is displayed on the display interface.

5. The intelligent screening system for infant neurodevelopment based on AI closed-loop control according to claim 1, characterized in that, The evaluation processing platform includes an evaluation mode switching module for switching between AI-assisted evaluation mode and expert remote evaluation mode; and the evaluation processing platform is configured to provide AI evaluation results as reference information to remote experts for comprehensive evaluation.

6. The intelligent screening system for infant neurodevelopment based on AI closed-loop control according to claim 1, characterized in that, The drive mechanism includes a rotary motor and an electric push rod; and the data acquisition device also includes a column, a chassis, medical silent wheels, a battery enclosure, and an all-in-one computer.

7. The intelligent screening system for infant neurodevelopment based on AI closed-loop control according to claim 1, characterized in that, The evaluation and processing platform also includes a verification mechanism to ensure that the infant information uploaded through the user terminal is accurately linked to the second video collected through the data acquisition device.

8. A method for intelligent screening of infant neurodevelopment based on AI closed-loop control, characterized in that, Includes the following steps: Receive infant information entered by the user and the first video uploaded through the user terminal; A second video of an infant is captured using a data acquisition device, which includes a movable support arm, a high-definition camera mounted on the support arm, and a drive mechanism for moving the support arm and / or the high-definition camera. Based on real-time analysis of the second video, control commands are generated; The control command is sent to the drive mechanism of the data acquisition device to drive the high-definition camera to move to the target shooting position determined based on the real-time analysis results; The infant information, the first video, and the second video are stored in a PACS system, wherein the PACS system is configured to store data in the format required for general motor assessment. as well as The second video is evaluated and a report is generated.

9. The intelligent screening method for infant neurodevelopment based on AI closed-loop control according to claim 8, characterized in that, Also includes: During the acquisition of the second video, if, based on real-time analysis of the second video, it is determined that the second video lacks key motion shots of preset body parts, the drive mechanism is automatically controlled to perform supplementary shooting.

10. The intelligent screening method for infant neurodevelopment based on AI closed-loop control according to claim 8, characterized in that, Also includes: A quality pre-inspection is performed on the first video uploaded through the user terminal.

Images