Health early warning management system and method based on physical examination information

Abstract

The invention discloses a health early warning management system and method based on physical examination information, and particularly relates to the field of skeleton structure recognition, and the method comprises the steps: constructing a structural posture imaging data set based on a physical examination posture image of a student, extracting three-dimensional structure coordinates of parts, such as a scapula, a spine and a pelvis; and uniformly mapping to a skeleton reference system to calculate a body inclination index, a spine offset rate and a pelvic inclination index. Then, constructing a time sequence of posture structure parameters, and generating a trend feature vector sequence according to academic difference and gradient extraction; and further constructing a posture structure evolution trajectory diagram by taking the structure parameters as nodes and the trend vectors as edge attributes, and capturing a cross-year posture change path of the students. On the basis, all paths are aligned according to ages, the sudden change trend is extracted, the posture deviation high-incidence risk of a specific age group is recognized, and accurate student posture early warning management is achieved.

Description

Technical Field

[0001] This invention relates to the field of skeletal structure recognition technology, and more specifically, to a health early warning management system and method based on physical examination information. Background Technology

[0002] In the current health check-up system for primary and secondary school students, skeletal developmental abnormalities related to posture are receiving increasing attention, especially developmental skeletal problems such as scoliosis, scapular asymmetry, and pelvic tilt, whose incidence is rising year by year. However, the existing health check-up system mainly relies on single observations or static photographic assessments to identify these problems, lacking the ability to accumulate and analyze continuous and trend-based data, making it difficult to capture the evolution of abnormal postures in a timely manner. Especially during a student's growth cycle, skeletal development has significant stage-specific characteristics, and the trends in posture changes at different developmental stages often do not have immediately visible pathological manifestations, but their slow deviations may evolve into irreversible skeletal structural abnormalities years later. Traditional methods cannot effectively integrate historical posture indicators from students' health check-ups across academic years, resulting in blind spots in the identification of potential developmental abnormalities. In addition, posture data often lacks structured management due to inconsistent subjective assessment standards and fragmented recording content, further limiting the systematic nature and accuracy of risk identification.

[0003] Therefore, there is an urgent need to construct a method for modeling the physical evolution trajectory based on multi-period physical examination data, so as to realize the trend identification and risk warning of skeletal hidden abnormalities and improve the group development and health management capabilities of students. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, embodiments of the present invention provide a health early warning management method based on physical examination information to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A health early warning management method based on physical examination information includes the following steps:

[0007] S1: Establish a standardized imaging environment to collect posture images of students during physical examinations and construct a structural posture imaging dataset;

[0008] S2: Perform skeletal structure calculation on structural pose imaging and extract the three-dimensional structural coordinate sets of the scapula, spine, and pelvis.

[0009] S3: Map the three-dimensional structural coordinate set to a unified skeletal reference system and calculate the body structure parameters, including body tilt index, spinal offset rate and pelvic tilt index.

[0010] S4: Construct time series student body structure parameters, perform difference and gradient extraction on the changes in structure parameters in the order of academic years, and generate a trend feature vector sequence;

[0011] S5: Construct an evolutionary trajectory map using trend feature vectors as edge attributes and body structure parameters as nodes to capture cross-year body structure evolution paths;

[0012] S6: In the evolutionary trajectory diagram, all cross-year body structure evolution paths are aligned by age, and body trend mutation analysis is used to provide early warning of body deviation risk for the target age group.

[0013] In a preferred embodiment, in S1, establishing a standardized imaging environment to collect posture images of students during physical examinations and constructing a structural posture imaging dataset specifically includes:

[0014] Obtain the student ID and corresponding age from the physical examination registration form, and use them as the time sequence index for the acquisition of posture images;

[0015] Set up image acquisition positions with fixed height and shooting distance, set up background vertical rulers and ground foot distance guide frames, and construct a standardized imaging environment to acquire frontal and lateral posture images;

[0016] Image distortion correction and shooting angle offset measurement were performed on the frontal and lateral pose images respectively. Image frames that did not meet the preset composition constraints were filtered out to form a structural pose imaging dataset based on age division.

[0017] In a preferred embodiment, in S2, the skeletal structure calculation of the structural pose imaging and the extraction of the three-dimensional structural coordinate sets of the scapula, spine, and pelvis specifically include:

[0018] After classifying the image sequences in the structural pose imaging dataset according to the shooting angle, they are input into the bone point recognition network to perform full-body key point location detection. The key points refer to the set of human skeletal anatomical landmarks extracted from the pose images through image recognition.

[0019] The identified two-dimensional keypoint sequence is analyzed, the joint pair connection relationship is determined according to the preset bone connection rules, and the bone point connection topology is constructed.

[0020] By fusing the spatial coordinates of key points with the same name in frontal and side images, the skeletal structure is reconstructed through triangular projection. The scapular angle, the midline node of the spine, and the pelvic reference point are extracted from the reconstructed skeletal structure to form a three-dimensional structural coordinate set.

[0021] In a preferred embodiment, in S3, the three-dimensional structural coordinate set is mapped to a unified skeletal reference system, and postural structural parameters, including body tilt indices, spinal offset rate, and pelvic tilt indices, are calculated. Specifically, these include:

[0022] With the center point of the pelvis as the origin of the coordinate system, a three-dimensional skeleton reference system is constructed, and the reference axes in the left-right, vertical and front-back directions are defined.

[0023] The three-dimensional coordinates of the key points of the whole body are translated to the skeletal reference system, the coordinate difference of the left and right shoulder key points in the vertical axis direction is calculated, and the height difference of the two shoulders is output as the body tilt index.

[0024] Connect the sequence of key points of the spine, fit the spine center curve and calculate the maximum lateral offset rate relative to the principal axis as the spine offset rate;

[0025] Construct a pelvic plane vector, measure its angular offset relative to the left and right axes, and output the pelvic tilt angle as a pelvic tilt index.

[0026] In a preferred embodiment, the construction of a three-dimensional skeleton reference system with the center point of the pelvis as the origin is specifically defined by taking the midpoint of the line connecting the left and right anterior superior iliac spines as the origin, defining the X-axis with the direction of the line connecting from right to left, defining the Y-axis with the vector direction from the sacrum to the seventh cervical vertebra, and determining the direction of the Z-axis by the right-hand cross product of the X-axis and the Y-axis. The three axes constitute an orthogonal Cartesian coordinate system.

[0027] In a preferred embodiment, in S4, constructing time-series student postural structure parameters, and performing difference and gradient extraction on the changes in structure parameters in academic year order to generate a trend feature vector sequence specifically includes:

[0028] Extract body tilt indicators, spinal deviation rate and pelvic tilt indicators from the physical examinations of all students with the same student ID over the years, and sort them by academic year.

[0029] Perform interannual difference calculation on the sorted body structure parameter sequence, extract the interannual variation and mark the direction of change, and calculate the second gradient based on the interannual difference;

[0030] The differences and gradient values ​​of body structure parameters are combined into feature triples to form an annual trend feature vector set, and a complete sequence of individual body trend feature vectors is output.

[0031] In a preferred embodiment, in S5, an evolutionary trajectory graph is constructed using trend feature vectors as edge attributes and body structure parameters as nodes to capture the cross-year body structure evolution path, specifically including:

[0032] Use the physical structure parameters from past physical examinations corresponding to the student ID as graph nodes, and add age labels as the time attributes of the nodes.

[0033] Nodes in adjacent academic years are connected by trend feature vectors. The direction of the directed edge is set as the direction of age label increase. Based on the difference value and gradient direction of the trend vector, edge attributes representing evolution intensity and deformation trend are assigned.

[0034] A topological index is established for all nodes and their directed edge connections to form a cross-year body structure evolution path, thus creating a body structure evolution trajectory diagram representing the order of structure evolution.

[0035] In a preferred embodiment, in S6, in the evolutionary trajectory diagram, all cross-year body structure evolution paths are aligned by age, and body trend mutation analysis is used to provide a risk warning of body deviation for the target age group, specifically including:

[0036] Traverse the evolution trajectory diagram by student number, and align the nodes in all cross-year physical structure evolution paths in the evolution trajectory diagram according to the age label.

[0037] Extract the node path topology index aligned with age labels, and construct a continuous node subsequence aligned with age groups;

[0038] The percentage of edges in a subsequence of a statistical node where the gradient direction changes in all adjacent nodes is positive and the difference value exceeds a preset threshold is used as a scoring index for the abrupt change in the physical trend of the corresponding age group.

[0039] By integrating the scoring indicators of sudden changes in body posture across different age groups, a trend map of body posture deviation is obtained, enabling risk warnings of body posture deviation for target age groups.

[0040] On the other hand, the present invention provides a health early warning management system based on physical examination information, including a posture image acquisition module, a skeleton reconstruction module, a parameter calculation module, a trend extraction module, and a risk trajectory analysis module:

[0041] Attitude image acquisition module: Constructs a unified imaging environment, acquires frontal and lateral images and corrects distortion and angular offset, and constructs a structural attitude imaging dataset indexed by age;

[0042] Skeleton Reconstruction Module: Identifies key points of human skeleton in images, parses connection topology, fuses frontal and side images to reconstruct a 3D skeleton and extracts the coordinates of key bone points;

[0043] Parameter calculation module: Constructs a three-dimensional skeleton reference system with the pelvic center as the origin, unifies the coordinates of key points, and calculates the shoulder height difference, spinal offset rate and pelvic tilt angle;

[0044] Trend extraction module: Constructs a sequence of body posture parameters sorted by academic year, performs difference and gradient extraction, and generates an annual trend feature vector sequence describing changes in body posture;

[0045] Risk trajectory analysis module: Constructs a physical evolution trajectory map, extracts and aggregates path segments of the same age, calculates the mutation trend score, marks the risk age group and outputs early warning results.

[0046] The technical effects and advantages of the health early warning management system and method based on physical examination information of this invention are as follows:

[0047] This study employs multi-angle skeletal structure calculation technology to reconstruct a three-dimensional skeleton, improving the accuracy and stability of structural extraction in the scapula, spine, and pelvis. A unified skeletal reference system is used to standardize the three-dimensional structural coordinates, ensuring a basis for lateral comparison of structural parameters across different students and time points. A time series of postural parameters is constructed based on continuous academic year data, and difference and gradient features are extracted to quantitatively express postural evolution trends. A postural evolution trajectory map is constructed using structural parameters as nodes and trend vectors as edge attributes, clearly describing the path and patterns of cross-year structural evolution. Finally, age alignment and trend clustering analysis of the trajectory paths are performed to identify the tendency for sudden changes in postural characteristics in specific age groups, providing crucial evidence for early warning and targeted intervention. This study possesses strong engineering feasibility and health risk prediction value. Attached Figure Description

[0048] Figure 1 This is a schematic diagram of a health early warning management method based on physical examination information according to the present invention;

[0049] Figure 2 This is a schematic diagram of the structure of a health early warning management system based on physical examination information according to the present invention. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0051] Example 1

[0052] Figure 1 This invention presents a health early warning management method based on physical examination information, which includes the following steps:

[0053] S1: Establish a standardized imaging environment to collect posture images of students during physical examinations and construct a structural posture imaging dataset;

[0054] S2: Perform skeletal structure calculation on structural pose imaging and extract the three-dimensional structural coordinate sets of the scapula, spine, and pelvis.

[0055] S3: Map the three-dimensional structural coordinate set to a unified skeletal reference system and calculate the body structure parameters, including body tilt index, spinal offset rate and pelvic tilt index.

[0056] S4: Construct time series student body structure parameters, perform difference and gradient extraction on the changes in structure parameters in the order of academic years, and generate a trend feature vector sequence;

[0057] S5: Construct an evolutionary trajectory map using trend feature vectors as edge attributes and body structure parameters as nodes to capture cross-year body structure evolution paths;

[0058] S6: In the evolutionary trajectory diagram, all cross-year body structure evolution paths are aligned by age, and body trend mutation analysis is used to provide early warning of body deviation risk for the target age group.

[0059] In S1, a standardized imaging environment is established to collect posture images of students during physical examinations, and a structural posture imaging dataset is constructed.

[0060] Each student's unique ID and current age information are extracted from the physical examination registration form and used as the timing identifier for posture image acquisition. The student ID serves as the primary key field for image data naming and archiving, while the age information is used for subsequent postural evolution sequence sequencing. To ensure image acquisition quality and structural consistency, a standardized imaging area needs to be established in the physical space. This area is constructed by uniformly configuring the position of the shooting equipment and background reference markers. Specifically, the installation height of the shooting device and the distance between it and the subject are set. The shooting equipment height is fixed at 130 cm, measured vertically from the ground and fixed in place using a stable tripod. The distance to the subject is set at 250 cm, with the corresponding standing position marked by a ground-based distance measuring tape, ensuring a constant shooting distance between the camera and the student in different shooting rounds and eliminating near-far distortion interference. A vertical scale is set in the shooting background to clarify the head and shoulder coordinates and postural symmetry reference. The vertical scale uses a gray-white high-contrast line design, with the starting point flush with the ground, an overall height of 2.2 meters, and a horizontal reference line set every 5 cm for spatial calibration processing in subsequent image correction. Simultaneously, foot positioning guide frames were set up on the ground in the shooting area. These frames were arranged parallel to each other at shoulder width intervals, marking the landing areas of both feet and defining the standing posture direction. This ensured good consistency in the standing posture, foot positioning, and body orientation of each student during the frontal and lateral image acquisition. The image acquisition included two types of images: frontal and lateral. For frontal images, students were required to face the camera, with their heads naturally looking forward and their arms hanging at their sides. For lateral images, students needed to maintain a standing posture rotated 90 degrees, with their left or right shoulder facing the shooting device, keeping their center of gravity stable, and not leaning forward or backward. After each round of acquisition, the image data was immediately numbered and archived, recording meta-information such as student number, age, acquisition date, shooting angle, and whether it met the composition standards, forming the original dataset of structural posture images.

[0061] After standardized data acquisition, to ensure the processing quality and usability of the structural posture image data, image correction and composition quality screening were performed on each image. Distortion correction was then applied to the images. Due to the potential for slight nonlinear optical distortions caused by fixed camera equipment and lens parameters, especially in edge areas where bending and stretching are common, a known calibration template was used for distortion correction. This correction process analyzed the shape of the vertical ruler line in the image background, extracted the difference between its ideal geometric baseline and the actual line segment contour in the image, established an image distortion model, and then reverse-mapped and corrected the image pixel distribution, restoring the vertical ruler to a linear, vertical state. The corrected image maintained linear structural morphology in the shoulder and head regions, facilitating subsequent keypoint extraction. Subsequently, the image shooting angle offset was measured. The offset measurement was based on the relative geometric relationship between the foot distance positioning frame in the image and the background vertical ruler. In frontal images, if the left and right feet are not symmetrically positioned within the guide frame, or if the shoulder center line deviates from the vertical scale by more than 3 degrees, it is considered an angular deviation. In side images, preset composition constraints are applied; for example, if the foot positioning deviates from the ground guide line by more than 5 centimeters, or if the head and shoulders deviate horizontally beyond a set threshold, it is considered an angular mismatch. This angle is automatically detected by the background marker lines in the image. Finally, the images that pass distortion correction and angular deviation judgment are filtered and classified, removing image frames that do not meet the composition standards, forming a set of effective posture images with good structural consistency. These images are then divided into annual data subsets based on age information, with each subset containing frontal and lateral structural posture images of students in the corresponding age group.

[0062] In S2, the skeletal structure is solved for the structural pose imaging, and the three-dimensional structural coordinate sets of the scapula, spine and pelvis are extracted.

[0063] The acquired image dataset is explicitly classified according to the shooting angle, that is, frontal and lateral images are classified and labeled separately. After classification, each type of image is input into a bone point recognition network (such as OpenPose, HRNet, and other open-source models) for human key point detection. This recognition process is based on the learning ability of image recognition algorithms on human anatomical structures, automatically detecting and labeling a set of specific anatomical locations, i.e., key bone points, from the input images. The selection range of key points covers the top of the head, middle neck, acromion, elbow, wrist, midline of the spine, pelvis, knee joint, and ankle, among which areas closely related to posture analysis, such as the scapular angle (left and right acromions), midpoints of each segment of the spine, sacral point, and anterior superior iliac spine point, are the focus of detection. This recognition task does not rely on manual annotation or template matching, but uses a method of fusing convolutional visual features and spatial geometric features. The most likely key point pixel positions are determined through model training, and the key point positions are output in two-dimensional coordinate form. The system ensures that at least 16 standard key points are identified in each image, and occlusion markers are set for missing points to support subsequent fusion and recovery mechanisms. Finally, this step outputs a two-dimensional location set with standard skeletal anatomical key points, along with location confidence information, laying the foundation for the next step of skeletal topology construction.

[0064] After obtaining the location information of two-dimensional key points in each pose image, logical connections are established between these key points according to preset human skeletal connection rules to construct a complete human skeletal topology. These connection rules are set based on anatomical common sense and standard pose modeling procedures, clearly defining the connection order and structural hierarchy between joints. For example, the vertex of the head connects to the seventh cervical vertebra to form a cervical axis; the left and right acromions connect to the elbow and wrist respectively to form the upper limb path; the midline of the spine connects sequentially from the cervical, thoracic, and lumbar vertebrae to the sacrum to form the central axis; and the pelvic region constructs the pelvic support through the anterior superior iliac spine, the sacral point, and the greater trochanter of the femur. When establishing connections, structural continuity and physiological rationality must be ensured. The connection order is fixed, and node combinations must satisfy the logical dependencies between upper and lower hierarchical structures; for example, the scapular point must be identified before the acromion connects to the elbow. The connection process uses a graph structure construction method, defining all key points as graph nodes, with connections as directed edges, and assigning labels to each edge indicating anatomical significance (e.g., "shoulder-elbow" represents the upper arm connection). This topology not only provides a descriptive view of physiological structures, avoiding structural distortions caused by misidentification at single points, but also, to support unified processing of multi-angle images, retains the shooting angle identifier during construction and performs geometric constraint compensation based on viewpoint features. For example, in a frontal image, the midline of the spine can provide information on left-right symmetry, while in a side image, it can enhance the ability to recognize forward and backward curvature.

[0065] After keypoint identification and topology construction, the keypoint information from the frontal and side images is fused to reconstruct the 3D skeleton structure. The reconstruction process is based on triangulation, using known shooting parameters and the 2D pixel coordinates of corresponding keypoints in the frontal and side images to calculate the actual position of each keypoint in 3D space. Specifically, firstly, the camera position, optical axis direction, and imaging scale parameters are determined through calibration data. Then, the coordinates of a keypoint in the frontal image on the image plane are paired with the coordinates of the corresponding keypoint in the side image, treating them as projections of the same spatial point onto two projection planes. By establishing the intersection relationship between the two projection planes, the intersection point in space is calculated, thus determining its 3D position. This process is repeated for all keypoints to form a complete 3D skeleton point set. High-precision reconstruction is achieved using projection rules and image geometry under standardized imaging conditions.

[0066] In S3, the three-dimensional structural coordinate set is mapped to a unified skeletal reference system, and postural structural parameters, including body tilt index, spinal offset rate, and pelvic tilt index, are calculated.

[0067] Using the identified 3D coordinates of key skeletal points as the foundation, the 3D spatial coordinates of the left and right anterior superior iliac spines are extracted, and a line connecting these two points is constructed. The midpoint of this line is set as the origin of the 3D reference system, i.e., the center point of the pelvis, serving as the spatial reference for all coordinate transformations and reference axis construction. Based on this, the horizontal X-axis is defined as the direction from the right anterior superior iliac spine to the left anterior superior iliac spine, ensuring it represents the "left-right direction" in actual anatomical semantics. The sacrum and the seventh cervical vertebra are then extracted, and a spatial vector from the sacrum to the seventh cervical vertebra is constructed, defined as the Y-axis, representing the "vertical direction" from the pelvis to the head. Then, according to the cross product rule of the right-hand coordinate system, the cross product of the X-axis and Y-axis is used as the Z-axis direction, representing the "anterior-posterior direction," thus constructing a three-axis orthogonal, directionally stable 3D skeletal reference system. This ensures that all subsequent postural parameter calculations are anatomically interpretable and comparable across individuals.

[0068] After constructing the skeletal reference system, the three-dimensional coordinates of the key points of the original whole-body skeleton are uniformly transformed to this reference coordinate system. A translation operation is performed on the coordinates of all key points, making the center point of the pelvis the origin of the new coordinate system. Then, the coordinate axes of all key points are rotated to reposition them according to the defined X, Y, and Z axes. After the coordinate system transformation, the coordinate values ​​of the left and right acromion points in the Y-axis direction (i.e., the vertical direction) are extracted, and the difference between the two points in this axis direction is directly calculated; this value is the shoulder height difference. This height difference reflects the degree of lateral tilt of the body during a physical examination and is often used to identify early signs of postural deviation, scoliosis, etc. The entire calculation process is performed based on the unified reference system, ensuring reproducibility and stability.

[0069] All key points of the spine from the neck to the sacrum, such as the seventh cervical vertebra, the middle thoracic vertebrae, the upper lumbar vertebrae, and the sacrum, are extracted to form a sequence of key spinal points. This sequence is then fitted with a spatial curve, and a spinal center curve is generated using a sequential connection and smoothing method. Subsequently, in a skeletal reference frame, the perpendicular Y-axis is selected as the symmetry reference axis for the fitted curve. The offset of each point on the fitted curve in the X-axis direction (i.e., laterally) is statistically analyzed, and its maximum value is recorded. The proportion of this maximum offset value to the entire length of the spine is calculated, which is the spinal offset rate. The offset rate, as a quantitative indicator, reflects the degree of lateral curvature of the spine and, when combined with historical data, can be used for early screening of scoliosis or asymmetrical growth trends.

[0070] The left and right anterior superior iliac spines and the sacral point are extracted to form the pelvic plane. This plane can be determined by three points, constructing a pelvic reference plane vector in space. The angle between this vector and the X-axis vector in the skeletal reference frame is then calculated to determine the spatial angle between the pelvic plane and the left-right direction. This angle is the pelvic tilt angle, reflecting whether the pelvis rotates in the left-right direction, or whether postural instability causes one side of the pelvis to rise or fall. This angle, as an independent indicator, can be used in conjunction with the shoulder height difference and spinal offset rate to assess the coordination and stability of an individual's overall postural structure.

[0071] In S4, time series student body structure parameters are constructed, and differential and gradient extraction are performed on the changes in structure parameters in the order of academic years to generate a sequence of trend feature vectors.

[0072] Based on the calculated postural structural parameters, all students were iterated through by student ID. For each student, three core postural structural parameters obtained from their annual physical examinations were extracted: shoulder height difference (reflecting body tilt), spinal deviation rate (reflecting the degree of scoliosis), and pelvic tilt angle (reflecting pelvic horizontal stability). These parameters are all defined within a standard skeletal reference frame, possessing a unified spatial meaning and comparability. During the extraction process, the physical examination records were sorted according to the academic year information recorded in the examination registration, ensuring that each student's structural parameter sequence was strictly arranged in ascending order over time. After completing the construction of the structural parameter sequence, the changing trends of each parameter over time were further extracted. First-order differencing was performed on each type of structural parameter sequence for each student, i.e., calculating the difference between the indicator values ​​of two adjacent academic years. This operation directly reflects the annual change range of the indicator over time. A positive difference value indicates an upward trend in the structural indicator, a negative difference value indicates a downward trend, and a zero value indicates that the indicator remained stable during that period. In the differencing results, the direction of change between each pair of adjacent academic years was clearly marked, forming a trend label, constituting the first level of trend information. This process transforms the static values ​​of the original indicators into dynamically changing values, enabling subsequent models to identify trend changes rather than absolute value differences. Furthermore, by appropriately normalizing and quantifying the difference values, they are used as vector representations in subsequent multidimensional trend modeling, enhancing the completeness of the trend description.

[0073] Second-order differencing is introduced to identify changes in the rate of change. For each student and each structural parameter, the difference between adjacent differencing values ​​is calculated to form a trend gradient sequence. This gradient sequence can be used to determine whether structural changes are accelerating, slowing down, or reaching an inflection point. For example, if the shoulder height difference increases by 5 mm in one year and by 15 mm in the next, it indicates an acceleration of the tilt change, with a positive gradient value; conversely, a decrease indicates a slowdown. This gradient calculation provides higher-order trend features, particularly suitable for identifying abrupt changes or phases of intensified structural deformation. Performing this calculation requires ensuring the time series data is complete and the time intervals are consistent to guarantee the practical significance of the gradient changes. After completing the first-order differencing and second-order gradient calculations, a complete trend description unit is formed for each year. The trend feature of each node consists of three elements: the original value of the structural parameter for that year, the difference between that parameter and the previous year, and the gradient between that difference and the previous year. These three data points are combined into a structural feature triplet, describing the changes in the current state, direction of change, and rate of change, respectively. By combining the triples from each year in chronological order, a sequence of annual trend feature vectors for each structural parameter dimension of the student is constructed. This sequence preserves the temporal evolution of body posture changes while also possessing complete local trends. Finally, by fusing the trend feature vector sequences of the three types of structural parameters, a complete set of individual body posture trend feature sequences can be constructed.

[0074] In S5, an evolutionary trajectory map is constructed using trend feature vectors as edge attributes and body structure parameters as nodes to capture the cross-year body structure evolution path.

[0075] The postural structural parameters in the time series are represented using a graph structure. Taking each student's physical examination record as a unit, structural graph nodes are constructed based on their postural characteristics at different times. Each node represents the student's structural state in a given academic year, composed of three previously output core parameters (shoulder height difference, spinal offset rate, and pelvic tilt angle). This three-dimensional structural parameter set serves as the node's basic descriptive attribute. Furthermore, to ensure the temporal order of nodes in the evolutionary path, the student's age at the time of the physical examination can be explicitly attached to the node, forming a temporally ordered set of nodes. After defining the structural nodes, directed connections can be established between them, constructing the physical basis of the paths in the graph. For each student, two structural nodes formed in adjacent academic years should be connected by a directed edge, with the direction always increasing in age, i.e., from the node of the previous academic year to the node of the next academic year. This connection edge not only represents the temporal evolution order but also needs to carry attribute information reflecting structural changes. Therefore, a trend feature vector generated in the previous stage is embedded in the edge, which is the combination of the first-order difference and second-order gradient values ​​of the structural parameters between nodes. These edge attributes are used to quantitatively describe the intensity of change (reflected by the difference magnitude) and the trend of change (reflected by the gradient direction) in the evolution path segment of the structure.

[0076] After all nodes and edges are constructed, the topological relationships of the entire graph structure are further established. This topological structure is not merely a storage of connection information, but rather organizes the continuous sequence of physical state nodes for each student into a complete structural evolution path, using student ID as the index unit. Each path represents a student's physical structure changes across multiple age groups, consisting of a set of nodes arranged in chronological order and their connecting edges. When constructing the topological index, the system first traverses by student ID, sorts all nodes for each student by age label, then connects them sequentially to generate path segments, and registers the entire path in the topological index set. This index set allows for rapid location of any student's physical state at any age group and its subsequent changes, while also supporting group-level path clustering and trend analysis. Finally, all structural path indexes are merged to form a complete physical structure evolution trajectory graph. This graph structure fully preserves information on individual change patterns, evolutionary order across age spans, and trend directions, providing a structured framework for physical risk assessment and group trend mining.

[0077] In S6, the evolutionary trajectory diagram aligns all cross-year body structure evolution paths by age, and through body trend mutation analysis, it provides early warning of body deviation risk for target age groups.

[0078] Using student IDs as the index, the previously constructed physical structure evolution trajectory map is traversed. In each individual path, nodes are clearly labeled with age tags, representing the student's physical structure status at different age levels. During path traversal, alignment matching is performed according to a unified age sequence template. For example, setting 6 to 18 years old as the reference alignment age range, the corresponding nodes of each student in the path are assigned to their corresponding positions in the standard age sequence in ascending order of age. During this process, individuals missing certain age nodes will have their corresponding positions left blank or data from the most recent year used to fill in the missing nodes. All cross-year path structures are uniformly mapped to the same age sequence template, forming a structured set of nodes with age alignment semantics. After the alignment operation is completed, the set of nodes aligned under each specific age tag is extracted, and continuous path subsequences are reconstructed according to their preceding and following connections in the original path. This process preserves the continuity of individual physical evolution while ensuring the comparability of group evolution paths within the same age group. Each set of age-aligned nodes forms a path subsequence, which topologically represents a local segment shared by multiple student paths that traverses that age group. Nodes within the subsequence retain structural parameters and trend characteristics from the original evolutionary path. The system organizes this topological index by age label, treating each age group as an independent analysis unit and recording structural indicators such as the number of participating path segments, node connectivity, and edge attributes (trend characteristics) within that age group. The construction of path subsequences ensures that trend mutation analysis not only focuses on fluctuations in individual evolutionary trajectories but also supports the identification of common trend change characteristics within a population.

[0079] After extracting the path subsequences, a trend mutation analysis is performed on the node path segments for each age group. Specifically, the connection edges between each pair of consecutive nodes are traversed, and the difference components and gradient direction attributes contained therein are read. If the gradient direction of the edge is positive, it indicates that the structural parameters are in a state of accelerated change, and the difference amplitude exceeds the predefined mutation reference amplitude (using twice the average body shape change of the same age group in history as the evaluation benchmark), then the edge is marked as a trend mutation edge. Subsequently, the ratio between the number of edges marked as trend mutation edges in the current age group and the total number of connection edges in the age group is calculated. This ratio is the group body shape trend mutation score index for that age group. This score index quantitatively reflects the severity of the general changes in body shape structure among individuals in that age group and is used to assess whether there is a concentrated risk of change trend. In the example, if the mutation edge ratio in a certain age group exceeds 40%, it can be determined that there is a significant structural shift risk in that age group.

[0080] After completing trend change scoring for all age groups, the scores are sorted by age label to form a continuous sequence of group postural deviation scores. This sequence can be viewed as a postural trend evolution curve, where each point represents the deviation intensity for a specific age group. Further analysis of this trend curve identifies intervals with significant spikes, sustained high levels, or abrupt fluctuations, and these intervals are designated as high-risk age groups. For example, if the score jumps from 20% to 55% between the ages of 11 and 13, and remains above 50% for two consecutive years, this interval will be marked as a "high-incidence period of structural deviation." Finally, the risk labeling results are output as a visualized postural deviation trend chart, and a structural abnormality warning is issued for the target age group. This warning can be used by schools, medical examination institutions, or health management personnel as a basis for intervention, developing targeted monitoring and training strategies to achieve the health management goal of early identification and early intervention.

[0081] Example 2

[0082] The difference between Embodiment 2 and Embodiment 1 is that this embodiment introduces a health early warning management system based on physical examination information.

[0083] Figure 2 A schematic diagram of a health early warning management system based on physical examination information is provided according to the present invention. The health early warning management system based on physical examination information includes a posture image acquisition module, a skeleton reconstruction module, a parameter calculation module, a trend extraction module, and a risk trajectory analysis module.

[0084] Attitude image acquisition module: Constructs a unified imaging environment, acquires frontal and lateral images and corrects distortion and angular offset, and constructs a structural attitude imaging dataset indexed by age;

[0085] Skeleton Reconstruction Module: Identifies key points of human skeleton in images, parses connection topology, fuses frontal and side images to reconstruct a 3D skeleton and extracts the coordinates of key bone points;

[0086] Parameter calculation module: Constructs a three-dimensional skeleton reference system with the pelvic center as the origin, unifies the coordinates of key points, and calculates the shoulder height difference, spinal offset rate and pelvic tilt angle;

[0087] Trend extraction module: Constructs a sequence of body posture parameters sorted by academic year, performs difference and gradient extraction, and generates an annual trend feature vector sequence describing changes in body posture;

[0088] Risk trajectory analysis module: Constructs a physical evolution trajectory map, extracts and aggregates path segments of the same age, calculates the mutation trend score, marks the risk age group and outputs early warning results.

[0089] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0090] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0091] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0092] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0093] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0094] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0095] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0096] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0097] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0098] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A health early warning management method based on physical examination information, characterized in that, Includes the following steps: S1: Establish a standardized imaging environment to collect posture images of students during physical examinations and construct a structural posture imaging dataset; S2: Perform skeletal structure calculation on structural pose imaging and extract the three-dimensional structural coordinate sets of the scapula, spine, and pelvis. S3: Map the three-dimensional structural coordinate set to a unified skeletal reference system and calculate the body structure parameters, including body tilt index, spinal offset rate and pelvic tilt index. S4: Construct time series student body structure parameters, perform difference and gradient extraction on the changes in structure parameters in the order of academic years, and generate a trend feature vector sequence; S5: Construct an evolutionary trajectory map using trend feature vectors as edge attributes and body structure parameters as nodes to capture cross-year body structure evolution paths; S6: In the evolutionary trajectory diagram, all cross-year body structure evolution paths are aligned by age, and body trend mutation analysis is used to provide early warning of body deviation risk for the target age group.

2. The health early warning management method based on physical examination information according to claim 1, characterized in that, In S1, a standardized imaging environment is established to collect posture images of students during physical examinations, and a structural posture imaging dataset is constructed, specifically including: Obtain the student ID and corresponding age from the physical examination registration form, and use them as the time sequence index for the acquisition of posture images; Set up image acquisition positions with fixed height and shooting distance, set up background vertical rulers and ground foot distance guide frames, and construct a standardized imaging environment to acquire frontal and lateral posture images; Image distortion correction and shooting angle offset measurement were performed on the frontal and lateral pose images respectively. Image frames that did not meet the preset composition constraints were filtered out to form a structural pose imaging dataset based on age division.

3. The health early warning management method based on physical examination information according to claim 1, characterized in that, In S2, the skeletal structure is solved for structural pose imaging, and the three-dimensional structural coordinate sets of the scapula, spine, and pelvis are extracted, specifically including: After classifying the image sequences in the structural pose imaging dataset according to the shooting angle, they are input into the bone point recognition network to perform full-body key point location detection. The key points refer to the set of human skeletal anatomical landmarks extracted from the pose images through image recognition. The identified two-dimensional keypoint sequence is analyzed, the joint pair connection relationship is determined according to the preset bone connection rules, and the bone point connection topology is constructed. By fusing the spatial coordinates of key points with the same name in frontal and side images, the skeletal structure is reconstructed through triangular projection. The scapular angle, the midline node of the spine, and the pelvic reference point are extracted from the reconstructed skeletal structure to form a three-dimensional structural coordinate set.

4. The health early warning management method based on physical examination information according to claim 1, characterized in that, In S3, the three-dimensional structural coordinate set is mapped to a unified skeletal reference frame, and postural structural parameters are calculated, including body tilt indices, spinal offset rate, and pelvic tilt indices. Specifically, these include: With the center point of the pelvis as the origin of the coordinate system, a three-dimensional skeleton reference system is constructed, and the reference axes in the left-right, vertical and front-back directions are defined. The three-dimensional coordinates of the key points of the whole body are translated to the skeletal reference system, the coordinate difference of the left and right shoulder key points in the vertical axis direction is calculated, and the height difference of the two shoulders is output as the body tilt index. Connect the sequence of key points of the spine, fit the spine center curve and calculate the maximum lateral offset rate relative to the principal axis as the spine offset rate; Construct a pelvic plane vector, measure its angular offset relative to the left and right axes, and output the pelvic tilt angle as a pelvic tilt index.

5. A health early warning management method based on physical examination information according to claim 4, characterized in that, The construction of a three-dimensional skeleton reference system with the center point of the pelvis as the origin is specifically as follows: the midpoint of the line connecting the left and right anterior superior iliac spines is used as the origin; the X-axis is defined by the direction of the line connecting from right to left; the Y-axis is defined by the vector direction from the sacrum to the seventh cervical vertebra; and the Z-axis is determined by the right-hand cross product of the X-axis and the Y-axis. The three axes constitute an orthogonal Cartesian coordinate system.

6. A health early warning management method based on physical examination information according to claim 1, characterized in that, In S4, the time series student posture structure parameters are constructed, and the changes in structure parameters are extracted by difference and gradient extraction according to the academic year to generate a trend feature vector sequence. Specifically, this includes: Extract body tilt indicators, spinal deviation rate and pelvic tilt indicators from the physical examinations of all students with the same student ID over the years, and sort them by academic year. Perform interannual difference calculation on the sorted body structure parameter sequence, extract the interannual variation and mark the direction of change, and calculate the second gradient based on the interannual difference; The differences and gradient values ​​of body structure parameters are combined into feature triples to form an annual trend feature vector set, and a complete sequence of individual body trend feature vectors is output.

7. A health early warning management method based on physical examination information according to claim 1, characterized in that, In S5, an evolutionary trajectory graph is constructed using trend feature vectors as edge attributes and body structure parameters as nodes to capture cross-year body structure evolution paths, specifically including: Use the physical structure parameters from past physical examinations corresponding to the student ID as graph nodes, and add age labels as the time attributes of the nodes. Nodes in adjacent academic years are connected by trend feature vectors. The direction of the directed edge is set as the direction of age label increase. Based on the difference value and gradient direction of the trend vector, edge attributes representing evolution intensity and deformation trend are assigned. A topological index is established for all nodes and their directed edge connections to form a cross-year body structure evolution path, thus creating a body structure evolution trajectory diagram representing the order of structure evolution.

8. A health early warning management method based on physical examination information according to claim 1, characterized in that, In S6, the evolutionary trajectory diagram aligns all cross-year body structure evolution paths by age. Through body trend mutation analysis, it provides early warning of body deviation risks for target age groups, specifically including: Traverse the evolution trajectory diagram by student number, and align the nodes in all cross-year physical structure evolution paths in the evolution trajectory diagram according to the age label. Extract the node path topology index aligned with age labels, and construct a continuous node subsequence aligned with age groups; The percentage of edges in a subsequence of a statistical node where the gradient direction changes in all adjacent nodes is positive and the difference value exceeds a preset threshold is used as a scoring index for the abrupt change in the physical trend of the corresponding age group. By integrating the scoring indicators of sudden changes in body posture across different age groups, a trend map of body posture deviation is obtained, enabling risk warnings of body posture deviation for target age groups.

9. A health early warning management system based on physical examination information, used to implement the health early warning management method based on physical examination information as described in any one of claims 1-8, characterized in that, It includes a posture image acquisition module, a skeleton reconstruction module, a parameter calculation module, a trend extraction module, and a risk trajectory analysis module. Attitude image acquisition module: Constructs a unified imaging environment, acquires frontal and lateral images and corrects distortion and angular offset, and constructs a structural attitude imaging dataset indexed by age; Skeleton Reconstruction Module: Identifies key points of the human skeleton in images, parses the connection topology, fuses frontal and side images to reconstruct a 3D skeleton and extracts the coordinates of key bone points; Parameter calculation module: Constructs a three-dimensional skeleton reference system with the pelvic center as the origin, unifies the coordinates of key points, and calculates the shoulder height difference, spinal offset rate and pelvic tilt angle; Trend extraction module: Constructs a sequence of body posture parameters sorted by academic year, performs difference and gradient extraction, and generates an annual trend feature vector sequence describing changes in body posture; Risk trajectory analysis module: Constructs a physical evolution trajectory map, extracts and aggregates path segments of the same age, calculates the mutation trend score, marks the risk age group and outputs early warning results.

Images