Children and teenager foot spine deformity screening method based on artificial intelligence and computer vision technology

Abstract

The invention relates to a method for screening foot spine deformity of children and adolescents based on artificial intelligence and computer vision technologies. The method comprises the following steps: firstly, acquiring foot three-dimensional data of children and teenagers, and constructing a three-dimensional foot spine model based on the data; thirdly, processing the data of the three-dimensional foot spine model to obtain key geometric features; and calculating a malformation index by using a specific formula according to the characteristics, comparing the malformation index with a normal parameter threshold range, and marking the characteristic value as a potential malformation characteristic value if the threshold value is exceeded. And then, inputting the potential malformation characteristic value into a three-dimensional convolutional neural network to obtain a malformation type classification label. And generating a deformity degree score value by means of a clustering algorithm according to the label. And finally, inputting the score value into a trained three-dimensional foot ridge model to obtain a foot suggestion report comprising targeted suggestions. According to the method, the potential features of the foot spine deformity can be identified more accurately, the screening accuracy is greatly improved, the screening efficiency is effectively improved, and a personalized correction scheme can be formulated.

Description

Technical Field

[0001] The present invention belongs to the field of artificial intelligence applications, and particularly relates to a screening method for children and adolescents' foot and spine deformities based on artificial intelligence and computer vision technologies. Background Art

[0002] With the development of artificial intelligence application technologies, screening technologies for children and adolescents' foot and spine deformities based on artificial intelligence and computer vision technologies have emerged. As an important issue affecting their physical development and quality of life, early and accurate screening is crucial. Traditional screening methods for children and adolescents' foot and spine deformities mainly rely on manual palpation and visual observation. This method is not only inefficient and difficult to quickly carry out screening work in large-scale groups, but also greatly affected by subjective factors, making it difficult to guarantee accuracy and unable to achieve a standardized diagnosis process. Summary of the Invention

[0003] Based on this, it is necessary to provide a screening method for children and adolescents' foot and spine deformities based on artificial intelligence and computer vision technologies that can achieve more efficient, accurate, and objective screening, providing strong support for early detection and intervention of foot and spine deformities.

[0004] In a first aspect, the present application provides a screening method for children and adolescents' foot and spine deformities based on artificial intelligence and computer vision technologies, including:

[0005] Obtain three-dimensional data of the feet of children and adolescents; obtain the plantar pressure distribution and spinal curvature according to the three-dimensional data of the feet, and construct a three-dimensional foot and spine model; the three-dimensional data of the feet includes heel images and footprint images.

[0006] Based on the three-dimensional foot and spine model, use a voxel segmentation algorithm to partition the data to obtain key geometric features; calculate the deformity index according to the key geometric features using a formula, and compare the deformity index with the normal parameter threshold range. If it exceeds the threshold range, it is marked as a potential deformity feature value.

[0007] Input the potential deformity feature value into a three-dimensional convolutional neural network for spatial feature analysis to obtain a deformity type classification label; analyze the matching relationship between the feature parameters and growth parameters according to the deformity type classification label using a clustering algorithm to generate a deformity degree score value.

[0008] Input the deformity degree score value into the trained three-dimensional foot and spine model to obtain a foot advice report.

[0009] In one embodiment, based on the three-dimensional foot and spine model, use a voxel segmentation algorithm to partition the data to obtain key geometric features, including:

[0010] Obtain the voxel grid data of the three-dimensional foot and arch model; the voxel grid data includes the bone structure point cloud and the pressure distribution information.

[0011] Perform region growing segmentation on the voxel grid data according to a preset density threshold to obtain the plantar pressure gradient partition.

[0012] Perform morphological closing operation on the plantar pressure gradient partition to generate the boundary of the plantar contact area.

[0013] Extract the pressure peak coordinates in the bone structure point cloud based on the boundary of the plantar contact area and combine its curvature distribution to obtain the coordinates of the arch vertex.

[0014] Perform spatial vector matching between the coordinates of the arch vertex and the midline point cloud of the spine, and calculate the foot and arch biomechanical coupling parameters.

[0015] Use the random forest algorithm to perform feature fusion on the foot and arch biomechanical coupling parameters to obtain the key geometric features reflecting the morphological features of the foot and spine; the key geometric features include the arch height, the scoliosis angle, and the characteristic parameters of the plantar pressure distribution.

[0016] In one embodiment, the deformity index is calculated according to the key geometric features using the formula, including:

[0017] The deformity index is calculated using the following formula:

[0018]

[0019] Among them, H represents the comprehensive arch height index, n represents the number of measurement points, h i represents the arch height of the i-th measurement point, L i represents the distance from the i-th measurement point to the heel, L max represents the foot length, S represents the scoliosis degree index, m represents the number of spinal measurement segments, α t and β t represent the weight coefficients of the t-th segment, θ t represents the scoliosis angle of the t-th segment, P represents the plantar pressure distribution uniformity index, k represents the number of pressure sensors, p j represents the pressure value measured by the j-th sensor, p avg represents the average pressure value, p max represents the maximum pressure value.

[0020] In one embodiment, input the potential deformity feature values into a three-dimensional convolutional neural network for spatial feature analysis to obtain the deformity type classification labels, including:

[0021] Use the multi-scale feature fusion algorithm to calculate the potential deformity feature values to obtain the multi-modal feature tensor.

[0022] Input the multi-modal feature tensor into a three-dimensional convolutional neural network to extract spatial topological features.

[0023] Generate an activation map sequence based on the spatial topological features and perform feature fusion to obtain a probability distribution vector.

[0024] Judge the maximum probability value in the probability distribution vector to obtain a preliminary classification label corresponding to the type of deformity.

[0025] Among them, if the preliminary classification label matches the preset threshold, the neural network is used to adjust the input of the feature values to obtain optimized spatial topological features.

[0026] Group the preliminary classification labels according to the optimized spatial topological features using a clustering algorithm to obtain a set of classification labels.

[0027] Judge whether there are overlapping types in the set of classification labels. If so, compare the differences in spatial topological features to obtain the final classification label for the type of deformity.

[0028] In one embodiment, the deformity degree score value is calculated by the following formula:

[0029]

[0030] Among them, D represents the personalized deformity deviation degree, n represents the number of feature parameters, ω i represents the weight coefficient of the i-th feature, y i represents the current measured parameter value, ref i represents the standard reference value, σ represents the correction coefficient, R represents the growth parameter correlation degree, m represents the number of growth parameters, μ i represents the influence factor of the i-th growth parameter, τ i represents the measured value of the i-th growth parameter, G represents the deformity degree score value, k represents the number of deformity type classifications, C i represents the set of samples of the i-th type of deformity, x represents the sample feature vector, δ i represents the center point of the i-th type.

[0031] In one embodiment, input the deformity degree score value into a trained three-dimensional foot arch model to obtain a foot advice report, including:

[0032] Obtain the simulation intervention parameters corresponding to the deformity degree score value; the simulation intervention parameters are used to drive the simulation of the foot arch shape change.

[0033] Generate dynamic grid data of the arch height distribution and spinal curvature distribution according to the simulation intervention parameters.

[0034] Map the dynamic grid data into the trained three-dimensional foot and spine model to obtain a morphological feature matrix including the trend of pressure distribution change.

[0035] Extract the curvature parameters of the arch height gradient and spinal curvature based on the morphological feature matrix; the curvature parameters are used to update the topological structure of the dynamic grid data.

[0036] Based on the curvature parameters, simulate the dynamic changes of the foot and spine under different time nodes and different intervention measures to generate three-dimensional visualization information; the three-dimensional visualization information includes the arch height distribution, spinal curvature distribution, and the trend of pressure distribution change.

[0037] Perform fusion analysis on the three-dimensional visualization information to generate a foot advice report; the foot advice report includes corrective measures, rehabilitation training plans, recommendations for suitable insoles or assistive devices, and time suggestions for regular reexaminations for different types and degrees of deformities.

[0038] In a second aspect, the present application also provides a computer device, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the foregoing method is implemented.

[0039] In a third aspect, the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the foregoing method is implemented.

[0040] The above-mentioned method for screening foot and spine deformities in children and adolescents based on artificial intelligence and computer vision technologies, first, construct three-dimensional foot data by obtaining the heel images and footprint images of children and adolescents, and based on the data, obtain the plantar pressure distribution and spinal curvature, so as to construct a three-dimensional foot and spine model. Then, use the voxel segmentation algorithm to partition the data of the three-dimensional foot and spine model to obtain key geometric features, including characteristic parameters such as arch height, scoliosis angle, and plantar pressure distribution. Calculate the deformity index according to the key geometric features using a specific formula, and compare it with the normal parameter threshold range. If it exceeds the threshold, it is marked as a potential deformity feature value. Subsequently, input the potential deformity feature value into a three-dimensional convolutional neural network for spatial feature analysis to obtain a deformity type classification label. Then, based on this label, analyze the matching relationship between the characteristic parameters and growth parameters with the help of a clustering algorithm to generate a deformity degree score value. Finally, input this score value into the trained three-dimensional foot and spine model to obtain a foot advice report including targeted suggestions. The above method can more accurately identify the potential features of foot and spine deformities, greatly improve the accuracy of screening, effectively improve the screening efficiency, help formulate personalized correction plans, and provide strong support for the early intervention and improvement of the foot and spine health of children and adolescents. Brief Description of the Drawings

[0041] To more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following will briefly introduce the drawings required for use in the description of the embodiments or related technologies. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0042] Figure 1 Flowchart of the method for screening foot and spinal deformities in children and adolescents based on artificial intelligence and computer vision technologies provided by the embodiments of the present invention;

[0043] Figure 2 Flowchart of partitioning data using a voxel segmentation algorithm based on a three-dimensional foot and spine model provided by the embodiments of the present invention to obtain key geometric features;

[0044] Figure 3 Flowchart of inputting potential deformity feature values into a three-dimensional convolutional neural network for spatial feature analysis to obtain deformity type classification labels provided by the embodiments of the present invention;

[0045] Figure 4 Flowchart of inputting the deformity degree score value into a trained three-dimensional foot and spine model to obtain a foot advice report provided by the embodiments of the present invention. Detailed implementation manners

[0046] In order to make the objectives, technical solutions and advantages of the present application more clear and understandable, the following further details the present application in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0047] First, the implementation environment of the embodiments of the present application is described. Exemplarily, the implementation environment includes an image acquisition device, a data processing device, and a data storage device.

[0048] In the method for screening foot and spinal deformities in children and adolescents based on artificial intelligence and computer vision technologies, the image acquisition device captures image data of the foot and spine, is connected to the data processing device through a high-speed data cable, and transmits the collected original image data to the latter in real time. The data processing device runs relevant artificial intelligence algorithms to analyze and process the images, and the processed result data and the original image data are then orderly stored in the data storage device through an internal high-speed data bus or network connection.

[0049] Combined with the above implementation environment, the application scenarios of the embodiments of the present application are described.

[0050] The method for screening foot and spinal deformities in children and adolescents based on artificial intelligence and computer vision technology provided by the embodiments of the present application uses a professional image acquisition device to capture the images of the feet and spines of children and adolescents from multiple angles, comprehensively obtaining their morphological and depth information. The collected original image data is transmitted to a data processing device through a high-speed network. The device runs an artificial intelligence algorithm trained with a large number of samples to accurately identify bone feature points and calculate key parameters such as the arch height and scoliosis angle to analyze the foot and spinal conditions. The screening results and the original image data are stored in an orderly manner in a data storage device, providing support for doctors' diagnosis, condition tracking, and medical research, and realizing early and accurate screening and prevention and control of foot and spinal deformities. Exemplarily, the method for screening foot and spinal deformities in children and adolescents based on artificial intelligence and computer vision technology provided by the embodiments of the present application can be applied to at least one of the following scenarios including but not limited to the following scenarios.

[0051] First, the method for screening foot and spinal deformities in children and adolescents based on artificial intelligence and computer vision technology is applied to the school scenario. In the school environment, a combination of a professional high-definition camera and a depth camera quickly captures multi-angle images of the feet and spines. The depth camera simultaneously obtains depth information, recording the three-dimensional contour of the feet and the morphological curve of the spine. The collected image data is transmitted in real time through the high-speed network within the campus to a high-performance computer in the school data processing center. The computer quickly processes the images, uses a pre-trained model to identify bone feature points, and calculates key parameters such as the arch height and scoliosis angle. The processed screening results and the original image data are stored together in a school-specific storage server. Doctors in the school infirmary can retrieve the data at any time to further diagnose students suspected of having foot and spinal deformities, and at the same time, these data also provide a basis for the school to formulate targeted health intervention measures.

[0052] Second, the screening method for children and adolescents' foot and spinal deformities based on artificial intelligence and computer vision technology is applied to the hospital scenario. In the children's health care department or the rehabilitation medicine center of the hospital, the configuration of the image acquisition equipment is more professional and accurate. High-precision medical cameras and advanced three-dimensional reconstruction depth cameras can obtain extremely clear and detailed foot and spinal images. Under the guidance of medical staff, patients assume standard postures on a specific shooting platform to ensure the accuracy of image acquisition. The collected image data is quickly transmitted to the large server cluster in the data processing computer room through the high-speed and stable medical dedicated network within the hospital. The servers are equipped with powerful computing chips and a large amount of memory, running complex artificial intelligence algorithms to deeply analyze the images, which can not only accurately judge the types of foot and spinal deformities, but also predict the development trend of the condition. The analyzed result data is securely stored in the enterprise-level storage array of the hospital, and multiple redundant backup technologies are used to ensure data security. Doctors can retrieve the screening data of patients at any time through the hospital information system, combine clinical symptoms for comprehensive diagnosis, and formulate personalized treatment plans. At the same time, these data also provide valuable clinical case resources for medical research.

[0053] In one embodiment, as Figure 1 shown, the present application provides a screening method for children and adolescents' foot and spinal deformities based on artificial intelligence and computer vision technology, which may include the following steps:

[0054] Step S101, obtaining the three-dimensional data of the feet of children and adolescents; obtaining the plantar pressure distribution and spinal curvature according to the three-dimensional data of the feet, and constructing a three-dimensional foot and spinal model; the three-dimensional data of the feet includes the heel image and the footprint image.

[0055] The three-dimensional data of the feet of children and adolescents is obtained through specific computer vision devices and technologies. High-precision imaging systems are used to collect the heel image and the footprint image, which contain rich information such as the shape, contour, and texture of the feet. Based on the collected image data, advanced image processing algorithms and three-dimensional reconstruction technologies are used to generate a three-dimensional model of the feet. At the same time, with the help of pressure sensing devices and data analysis methods, the plantar pressure distribution is extracted from the three-dimensional data of the feet, and the pressure values of each part of the sole under different stress states are accurately measured. And, through the recognition and calculation of the characteristic points of the relevant parts of the feet and the spine, combined with the biomechanical principle and the human body structure model, the spinal curvature data is obtained. The plantar pressure distribution, spinal curvature data, and the three-dimensional foot and spinal model are fused to finally construct a three-dimensional foot and spinal model that comprehensively reflects the morphological and mechanical characteristics of children and adolescents' feet and spines.

[0056] Step S102: Based on the three-dimensional foot and spine model, use the voxel segmentation algorithm to partition the data and obtain key geometric features. Calculate the deformity index using a formula based on the key geometric features, and compare the deformity index with the normal parameter threshold range. If it exceeds the threshold range, it is marked as a potential deformity feature value.

[0057] Specifically, the voxel segmentation algorithm is used to perform refined processing on the model data. By analyzing the attributes of each voxel, the algorithm divides the three-dimensional model into multiple regions with specific meanings according to the characteristics and differences of the foot and spine structures. For example, it accurately distinguishes different structural regions such as the bones, muscles, and soft tissues of the foot, as well as each segment of the spine. After the partitioning is completed, key geometric features are extracted from these different regions, such as the height of the foot arch, the scoliosis angle of the spine, and the characteristic parameters of the plantar pressure distribution. These key geometric features can intuitively reflect the morphology and mechanical state of the foot and spine. Subsequently, a specially designed formula is used to calculate the key geometric features to obtain the deformity index. These formulas are derived based on a large amount of medical research, clinical data, and biomechanical principles, and have scientific basis and accuracy. The calculated deformity index is compared with the pre-set normal parameter threshold range. The normal parameter threshold range is determined through statistical analysis of the foot and spine data of a large number of healthy children and adolescents, representing the standard interval of normal foot and spine development. Once the deformity index exceeds this normal threshold range, it means that the development of the foot and spine may be abnormal. At this time, these out-of-range indicators are marked as potential deformity feature values. These potential deformity feature values become important bases for subsequent further diagnosis and analysis.

[0058] Step S103: Input the potential deformity feature values into a three-dimensional convolutional neural network for spatial feature analysis to obtain the deformity type classification label. Analyze the matching relationship between the feature parameters and the growth parameters using a clustering algorithm based on the deformity type classification label to generate the deformity degree score value.

[0059] The previously marked potential deformity feature values are used as input data and sent to the trained three-dimensional convolutional neural network for progressive abstraction and feature extraction, to explore the correlation and uniqueness of the potential deformity feature values in the spatial dimension, and then to obtain the key features for differentiating the deformity types, and finally to determine the deformity type classification label, so as to clarify what specific type of foot and spine deformity the potential deformity belongs to, such as flat feet, high arches, scoliosis, etc. After obtaining the deformity type classification label, a clustering algorithm is used to deeply analyze the matching relationship between the feature parameters and the growth parameters. The feature parameters include the data reflecting the foot and spine deformity conditions such as the previously extracted arch height, scoliosis angle, etc., while the growth parameters cover the factors related to growth and development such as the age, height, weight of children and adolescents. Through this analysis, considering various factors comprehensively, a deformity degree score value that can quantitatively reflect the severity of the foot and spine deformity is generated. This score value provides a key quantitative basis for subsequent assessment of the condition and formulation of personalized intervention plans, making the assessment of foot and spine deformities in children and adolescents more scientific and accurate.

[0060] Step S104, input the deformity degree score value into the trained three-dimensional foot and spine model to obtain a foot advice report.

[0061] The generated deformity degree score value is used as the key input information and sent to the three-dimensional foot and spine model that has been pre-trained with a large number of samples. This model has learned the characteristics, development laws and corresponding effective intervention strategies of many different foot and spine deformity cases. After receiving the deformity degree score value, the model combines the foot and spine biomechanical principles, medical knowledge and a large amount of clinical data stored internally to simulate the mechanical changes and growth trends of the foot and spine in daily activities under different deformity degrees. Through in-depth analysis of these simulated data, the model can predict the further impacts that the foot and spine deformity may bring, as well as the effects of different intervention measures on the improvement of the foot and spine.

[0062] The above-mentioned method for screening foot and spinal deformities in children and adolescents based on artificial intelligence and computer vision technology. First, by obtaining the heel images and footprint images of children and adolescents, three-dimensional foot data is constructed. Based on the data, the plantar pressure distribution and spinal curvature are obtained, thereby constructing a three-dimensional foot and spinal model. Then, the voxel segmentation algorithm is used to partition the data of the three-dimensional foot and spinal model to obtain key geometric features, including the arch height, scoliosis angle, and characteristic parameters of the plantar pressure distribution, etc. According to the key geometric features, the deformity index is calculated using a specific formula and compared with the normal parameter threshold range. If it exceeds the threshold, it is marked as a potential deformity eigenvalue. Subsequently, the potential deformity eigenvalue is input into a three-dimensional convolutional neural network for spatial feature analysis to obtain the deformity type classification label. Then, based on this label, the matching relationship between the feature parameters and growth parameters is analyzed using a clustering algorithm to generate the deformity degree score value. Finally, this score value is input into the trained three-dimensional foot and spinal model to obtain a foot advice report including targeted suggestions. The above method can more accurately identify the potential features of foot and spinal deformities, greatly improve the accuracy of screening, effectively improve the screening efficiency, help formulate personalized correction plans, and provide strong support for the early intervention and improvement of the foot and spinal health of children and adolescents.

[0063] In one of the embodiments, as Figure 2 shown, based on the three-dimensional foot and spinal model, using the voxel segmentation algorithm to partition the data to obtain key geometric features may include the following steps:

[0064] Step S201, obtain the voxel grid data of the three-dimensional foot and spinal model; the voxel grid data includes the bone structure point cloud and the pressure distribution information.

[0065] Step S202, perform region growing segmentation on the voxel grid data according to a preset density threshold to obtain the plantar pressure gradient partition.

[0066] Step S203, perform morphological closing operation on the plantar pressure gradient partition to generate the boundary of the plantar contact area.

[0067] Step S204, extract the pressure peak coordinates in the bone structure point cloud based on the boundary of the plantar contact area and combine its curvature distribution to obtain the arch vertex coordinates.

[0068] Step S205, perform spatial vector matching on the arch vertex coordinates and the spinal midline point cloud to calculate the foot and spinal biomechanical coupling parameters.

[0069] Step S206, use the random forest algorithm to perform feature fusion on the foot and spinal biomechanical coupling parameters to obtain the key geometric features reflecting the foot and spinal morphological features; the key geometric features include the arch height, scoliosis angle, and characteristic parameters of the plantar pressure distribution.

[0070] Specifically, first, obtain the voxel grid data of the three-dimensional foot and spine model, which contains the bone structure point cloud and pressure distribution information. Then, perform region growing segmentation on the voxel grid data according to a preset density threshold to obtain the plantar pressure gradient partition. Subsequently, perform morphological closing operation on the plantar pressure gradient partition to generate the boundary of the plantar contact area. After that, based on this boundary of the plantar contact area, extract the pressure peak coordinates in the bone structure point cloud, and combine its curvature distribution to determine the coordinates of the arch vertex. Then, perform spatial vector matching between the coordinates of the arch vertex and the midline point cloud of the spine, and calculate the foot and spine biomechanical coupling parameters. Finally, use the random forest algorithm to perform feature fusion on the foot and spine biomechanical coupling parameters, and then obtain the key geometric features that can reflect the morphological characteristics of the foot and spine. These key geometric features include the characteristic parameters of the arch height, scoliosis angle, and plantar pressure distribution.

[0071] In this embodiment, through steps such as region growing segmentation and morphological closing operation, the plantar pressure gradient and contact area can be accurately divided and defined. In terms of feature extraction, by combining operations such as pressure peak coordinates, curvature distribution, and spatial vector matching, the geometric features of the foot and spine are comprehensively and meticulously explored. In particular, the acquisition of the coordinates of the arch vertex and the foot and spine biomechanical coupling parameters provides key information for accurately evaluating the foot and spine morphology. Finally, using the random forest algorithm for feature fusion effectively integrates the information of multiple parameters, improving the representativeness and discrimination of the features. Overall, this method can more accurately and comprehensively reflect the morphological characteristics of the foot and spine, providing a solid and reliable basis for the early diagnosis, classification, and formulation of personalized treatment plans for foot and spine deformities, helping to improve the accuracy and effectiveness of foot and spine deformity screening, and being of great significance for ensuring the foot and spine health of children and adolescents.

[0072] In one of the embodiments, the deformity index can be calculated using a formula based on the key geometric features, and the following steps may be included:

[0073] The deformity index is calculated using the following formula:

[0074]

[0075] where H represents the comprehensive arch height index, n represents the number of measurement points, h i represents the arch height of the i-th measurement point, L i represents the distance from the i-th measurement point to the heel, L max represents the foot length, S represents the scoliosis degree index, m represents the number of spinal measurement segments, α t and β t represent the weight coefficients of the t-th segment, θ t represents the scoliosis angle of the t-th segment, P represents the plantar pressure distribution uniformity index, k represents the number of pressure sensors, pj represents the pressure value measured by the j-th sensor, p avg represents the average pressure value, p max represents the maximum pressure value.

[0076] This method comprehensively considers three key foot and spine feature dimensions: the arch of the foot, the spine, and the plantar pressure distribution. By calculating the comprehensive height index of the arch of the foot, the degree index of scoliosis, and the uniformity index of plantar pressure distribution, it can comprehensively and meticulously depict the morphology and functional status of the foot and spine, avoid the limitations of single-index evaluation, and make the evaluation of foot and spine deformities more comprehensive and accurate. These deformity indexes provide a quantitative basis for the early diagnosis and condition assessment of foot and spine deformities. Doctors can quickly and accurately judge the type and degree of foot and spine deformities based on these indexes, and then formulate personalized treatment plans to improve the pertinence and effectiveness of treatment. At the same time, it is also convenient to track and evaluate the treatment effect, provide strong support for subsequent treatment adjustment, contribute to improving the foot and spine health status of children and adolescents, and has important clinical practical significance.

[0077] In one embodiment, as Figure 3 shown, inputting the potential deformity feature values into a three-dimensional convolutional neural network for spatial feature analysis to obtain the deformity type classification labels may include the following steps:

[0078] Step S301, using a multi-scale feature fusion algorithm to calculate the potential deformity feature values to obtain a multi-modal feature tensor.

[0079] Step S302, inputting the multi-modal feature tensor into a three-dimensional convolutional neural network to extract the spatial topological features.

[0080] Step S303, generating an activation map sequence based on the spatial topological features and performing feature fusion to obtain a probability distribution vector.

[0081] Step S304, judging the maximum probability value in the probability distribution vector to obtain a preliminary classification label corresponding to the deformity type.

[0082] Among them, if the preliminary classification label matches the preset threshold, the neural network is used to adjust the feature value input to obtain the optimized spatial topological features.

[0083] Step S305, grouping the preliminary classification labels according to the optimized spatial topological features using a clustering algorithm to obtain a classification label set.

[0084] Step S306, judging whether there are overlapping types in the classification label set. If so, the final deformity type classification label is obtained by comparing the spatial topological feature differences.

[0085] First, use the multi-scale feature fusion algorithm to calculate the potential malformation eigenvalue. By fusing the feature information at different scales, a multi-modal feature tensor containing rich spatial and semantic information is obtained. Then, input this multi-modal feature tensor into a three-dimensional convolutional neural network to mine the spatial topological features in the data. Subsequently, generate a sequence of activation maps based on the obtained spatial topological features, and perform feature fusion on these sequences to obtain a probability distribution vector reflecting the possibilities of different malformation types. After that, judge the maximum probability value in the probability distribution vector to obtain the preliminary classification label corresponding to the malformation type. If the preliminary classification label matches the preset threshold, use the neural network to adjust the eigenvalue input to further optimize the spatial topological features. Then, based on the optimized spatial topological features, use the clustering algorithm to group the preliminary classification labels to form a classification label set. Finally, perform an overlapping type judgment on the classification label set. If there is an overlapping situation, compare the differences in spatial topological features to obtain the final accurate malformation type classification label.

[0086] In one of the embodiments, the malformation degree score value can be calculated by the following formula:

[0087]

[0088] where D represents the personalized malformation deviation degree, n represents the number of feature parameters, ω i represents the weight coefficient of the i-th feature, y i represents the current measured parameter value, ref i represents the standard reference value, σ represents the correction coefficient, R represents the growth parameter correlation degree, m represents the number of growth parameters, μ i represents the influence factor of the i-th growth parameter, τ i represents the measured value of the i-th growth parameter, G represents the malformation degree score value, k represents the number of malformation type classifications, C i represents the i-th set of malformation samples, x represents the sample feature vector, δ i represents the center point of the i-th class.

[0089] Among them, the personalized deformity deviation degree combines the number of characteristic parameters, the weight coefficient of each feature, the current measured parameter value and the standard reference value, which can accurately measure the deviation degree of the individual foot arch characteristics from the normal standard, and fully reflects the differential importance of different characteristics in deformity judgment. The correction coefficient further optimizes the calculation of this deviation degree to make the evaluation more in line with the actual situation. At the same time, the growth parameter correlation incorporates the number of growth parameters, the influence factor of each growth parameter and the measured value, fully considering the characteristics of the growth and development stage of children and adolescents, quantifying the influence of growth factors on foot arch deformity into the score, and ensuring that the score can dynamically reflect the deformity conditions at different growth stages. From the perspective of accuracy and scientificity, the classification calculation is carried out for the number of deformity types, considering different deformity type sample sets, sample feature vectors and their center points, so that the score can more accurately reflect the deformity degree under different deformity categories, avoiding the problem of confusing evaluations of different deformity types. Overall, through the scientific integration of multi-dimensional parameters, this formula provides a comprehensive, accurate and quantifiable evaluation index that can adapt to the growth changes of children and adolescents for the degree of foot arch deformity, which helps doctors more accurately judge the condition, formulate more targeted and effective intervention treatment plans, greatly improving the quality of foot arch deformity screening and diagnosis for children and adolescents, and is of great significance for ensuring the healthy development of their foot arches.

[0090] In one of the embodiments, as Figure 4 shown, inputting the deformity degree score value into the trained three-dimensional foot arch model to obtain a foot advice report may include the following steps:

[0091] Step S401, obtaining the simulation intervention parameters corresponding to the deformity degree score value; the simulation intervention parameters are used to drive the simulation of foot arch shape changes.

[0092] Step S402, generating dynamic grid data of the arch height distribution and the spinal curvature distribution according to the simulation intervention parameters.

[0093] Step S403, mapping the dynamic grid data into the trained three-dimensional foot arch model to obtain a morphological feature matrix including the change trend of the pressure distribution.

[0094] Step S404, extracting the arch height gradient and the curvature parameter of the spinal curvature based on the morphological feature matrix; the curvature parameter is used to update the topological structure of the dynamic grid data.

[0095] Step S405, simulating the dynamic changes of the foot arch at different time nodes and under different intervention measures based on the curvature parameter to generate three-dimensional visualization information; the three-dimensional visualization information includes the arch height distribution, the spinal curvature distribution and the change trend of the pressure distribution.

[0096] Step S406, perform fusion analysis on the three-dimensional visualization information to generate a foot advice report; the foot advice report includes correction measures, rehabilitation training plans, recommendations for suitable insoles or assistive devices, and time suggestions for regular reviews for different types and degrees of deformities.

[0097] First, obtain the simulation intervention parameters corresponding to the deformity degree score value, and generate dynamic grid data on the arch height distribution and spinal curvature distribution based on these simulation intervention parameters. Subsequently, map the generated dynamic grid data into the trained three-dimensional foot and spine model to obtain a morphological feature matrix containing the pressure distribution change trend. Extract the arch height gradient and the curvature parameters of the spinal curvature from this morphological feature matrix, and these curvature parameters will be used to update the topological structure of the dynamic grid data. Based on the updated curvature parameters, simulate the dynamic change process of the foot and spine at different time nodes and under different intervention measures, and then generate three-dimensional visualization information covering the arch height distribution, spinal curvature distribution, and pressure distribution change trend. Finally, perform fusion analysis on the generated three-dimensional visualization information to generate a foot advice report, and the report content includes correction measures, rehabilitation training plans, recommendations for suitable insoles or assistive devices, and time suggestions for regular reviews for different types and degrees of deformities.

[0098] In this embodiment, by obtaining the simulation intervention parameters and simulating the morphological changes of the foot and spine, it can intuitively show the impact of different intervention measures on the foot and spine at different times, provide a visual reference basis for doctors, help doctors more accurately judge the development trend of the condition, and thus formulate a more targeted treatment plan. For example, according to the dynamic changes of the arch height and spinal curvature, doctors can accurately determine the key points and directions of correction. In terms of patient rehabilitation, the detailed foot advice report provides clear rehabilitation guidance for patients and their families. The correction measures and rehabilitation training plans can help patients gradually improve the foot and spine deformity condition; the recommendations for suitable insoles or assistive devices help relieve foot pressure and promote the recovery of foot and spine function; the time suggestions for regular reviews ensure the scientificity and safety of the treatment process, can timely adjust the treatment plan, and ensure the foot and spine health of patients. This comprehensive and scientific process significantly improves the diagnosis and treatment effects of foot and spine deformities and is of great significance for improving the foot and spine health of children and adolescents.

[0099] It should be understood that although the steps in the flowcharts involved in the above-described embodiments are sequentially shown according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless there is a clear description in this article, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above-described embodiments may include multiple steps or multiple stages. These steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be executed alternately or in turn with at least a part of other steps or steps or stages in other steps.

[0100] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the method for screening children and adolescents' foot and spine deformities based on artificial intelligence and computer vision technology as described above are implemented.

[0101] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0102] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to the partial descriptions of the method embodiments. The device embodiments described above are only illustrative. The components described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules can be selected according to actual needs to achieve the purpose of the present disclosure solution. Those of ordinary skill in the art can understand and implement it without creative efforts.

[0103] The above-described embodiments only represent several implementation manners of the embodiments of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the application embodiments. It should be noted that for those of ordinary skill in the art, without departing from the concept of the embodiments of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the embodiments of the present application.

Claims

1. A method for screening foot ridge deformity in children and adolescents based on artificial intelligence and computer vision technology, characterized in that: The method comprises: Acquire three-dimensional foot data of children and adolescents; acquire plantar pressure distribution and spinal curvature according to the three-dimensional foot data, and construct a three-dimensional foot spine model; the three-dimensional foot data includes a heel image and a footprint image; Based on the three-dimensional foot spine model, the data is partitioned using a voxel segmentation algorithm to obtain key geometric features; the deformity index is calculated using a formula based on the key geometric features, and the deformity index is compared with a normal parameter threshold range. If it exceeds the threshold range, it is marked as a potential deformity feature value; The potential deformity feature value is input into a three-dimensional convolutional neural network for spatial feature analysis to obtain a deformity type classification label; a clustering algorithm is used to analyze the matching relationship between the feature parameter and the growth parameter according to the deformity type classification label to generate a deformity degree score value; The deformity degree score value is input into the trained three-dimensional foot spine model to obtain a foot recommendation report.

2. The method according to claim 1, characterized in that: The data is partitioned using a voxel segmentation algorithm based on the three-dimensional foot spine model to obtain key geometric features, including: Acquire voxel grid data of the three-dimensional foot spine model; the voxel grid data includes bone structure point cloud and pressure distribution information; Performing region growing segmentation on the voxel grid data according to a preset density threshold to obtain plantar pressure gradient partitions; Performing morphological closing operation on the plantar pressure gradient partitions to generate plantar contact area boundaries; Extracting the pressure peak coordinates in the bone structure point cloud based on the boundary of the sole contact area and combining it with its curvature distribution to obtain the arch vertex coordinates; Performing spatial vector matching of the arch vertex coordinates with the spine midline point cloud to calculate the biomechanical coupling parameters of the foot spine; The random forest algorithm is used to perform feature fusion on the biomechanical coupling parameters of the foot spine to obtain key geometric features that reflect the morphological characteristics of the foot and spine; the key geometric features include characteristic parameters of arch height, scoliosis angle and plantar pressure distribution.

3. The method according to claim 2, characterized in that The deformity index is calculated by using a formula according to the key geometric features, including: The deformity index is calculated using the following formula: Among them, H represents the comprehensive height index of the arch, n represents the number of measurement points, and h i represents the arch height of the i-th measurement point, L i represents the distance from the ith measurement point to the heel, L max represents foot length, S represents the degree of scoliosis, m represents the number of measured segments of the spine, and α t and β t represents the weight coefficient of the tth segment, θ t represents the lateral bending angle of the tth segment, P represents the uniformity index of plantar pressure distribution, k represents the number of pressure sensors, and p j represents the pressure value measured by the jth sensor, p avg Indicates the average pressure value, p max Indicates the maximum pressure value.

4. The method according to claim 1, characterized in that: The step of inputting the potential deformity feature value into a three-dimensional convolutional neural network for spatial feature analysis to obtain a deformity type classification label includes: The potential deformity feature value is calculated using a multi-scale feature fusion algorithm to obtain a multi-modal feature tensor; Inputting the multimodal feature tensor into a three-dimensional convolutional neural network to extract spatial topological features; Generate an activation map sequence according to the spatial topological features and perform feature fusion to obtain a probability distribution vector; Determining the maximum probability value in the probability distribution vector to obtain a preliminary classification label corresponding to the deformity type; Wherein, if the preliminary classification label matches the preset threshold, the feature value input is adjusted using a neural network to obtain an optimized spatial topological feature; Using a clustering algorithm to group the preliminary classification labels according to the optimized spatial topological features to obtain a classification label set; The classification label set is judged to determine whether there is overlapping type. If there is, the final deformity type classification label is obtained by comparing the differences in the spatial topological features.

5. The method according to claim 1, characterized in that The deformity score is calculated by the following formula: Where D represents the individualized deformity deviation, n represents the number of characteristic parameters, ω i represents the weight coefficient of the i-th feature, y i Indicates the current measurement parameter value, ref i represents the standard reference value, σ represents the correction coefficient, R represents the correlation of growth parameters, m represents the number of growth parameters, μ i represents the influencing factor of the i-th growth parameter, τ i represents the measured value of the i-th growth parameter, G represents the deformity score, k represents the number of deformity type classifications, C i represents the i-th type of deformed sample set, x represents the sample feature vector, δ i represents the center point of the i-th class.

6. The method according to claim 1, characterized in that The deformity degree score is input into the trained three-dimensional foot spine model to obtain a foot recommendation report, including: Acquiring simulation intervention parameters corresponding to the deformity degree score value; the simulation intervention parameters are used to drive the simulation of foot spine morphological changes; Generate dynamic mesh data of arch height distribution and spinal curvature distribution according to the simulation intervention parameters; Mapping the dynamic grid data to a trained three-dimensional foot spine model to obtain a morphological feature matrix including a pressure distribution change trend; Extracting curvature parameters of arch height gradient and spinal curvature based on the morphological feature matrix; the curvature parameters are used to update the topological structure of the dynamic mesh data; Based on the curvature parameters, dynamic changes of the foot spine under different time nodes and different intervention measures are simulated to generate three-dimensional visualization information; the three-dimensional visualization information includes the distribution of arch height, the distribution of spinal curvature and the change trend of pressure distribution; The three-dimensional visualization information is fused and analyzed to generate a foot recommendation report; the foot recommendation report includes corrective measures for different types and degrees of deformity, rehabilitation training plans, recommendations for suitable insoles or assistive devices, and time recommendations for regular review.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images