Sports injury risk intelligent prediction method based on sole mode

Abstract

The invention specifically discloses an intelligent sports injury risk prediction method based on a plantar modal, and the method comprises the steps: obtaining plantar pressure data and foot thermal imaging data, and generating a corresponding plantar pressure matrix and a temperature matrix according to the plantar pressure data and the foot thermal imaging data; performing plantar pressure field reconstruction according to the plantar pressure matrix and the temperature matrix, and extracting plantar multi-modal features from the reconstructed pressure field; and establishing a stress injury risk prediction model, and inputting the plantar multi-modal characteristics into the stress injury risk prediction model to perform sports injury risk prediction. The accuracy of sports injury risk prediction can be improved, and the effectiveness of athlete training monitoring, rehabilitation evaluation and public fitness risk prevention and control is improved.

Description

Technical Field

[0001] This invention relates to the technical field of sports medicine and injury prevention, and more specifically, to an intelligent prediction method for sports injury risk based on plantar modality. Background Technology

[0002] Statistics show that as many as 88.3% of runners have experienced or are currently experiencing running injuries due to various reasons, including muscle strains, ankle sprains, Achilles tendinitis, knee joint wear and tear, and O- or X-shaped legs. Currently, sports injury screening relies heavily on gait observation, univariate plantar pressure analysis, or subjective diagnosis using infrared thermography. Traditional plantar pressure assessment focuses on mechanical load but fails to reveal tissue metabolic status; infrared thermography can capture temperature abnormalities but is not sensitive to instantaneous impact force changes, resulting in a high rate of missed diagnoses in the early stages of sports injuries. Therefore, accurately predicting the risk of sports injuries to provide a scientific basis for the rehabilitation of patients with sports function impairments and for athletes' training is of great significance. Summary of the Invention

[0003] This invention provides an intelligent prediction method for sports injury risk based on plantar modality, which solves the problem that existing sports injury screening mainly relies on subjective diagnosis, which easily leads to a high rate of early missed diagnosis. It can improve the accuracy of sports injury risk prediction and increase the effectiveness of athlete training monitoring, rehabilitation assessment and public fitness risk prevention and control.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A smart prediction method for sports injury risk based on plantar modality includes:

[0006] Acquire plantar pressure data and foot thermal imaging data, and generate corresponding plantar pressure matrix and temperature matrix based on the plantar pressure data and foot thermal imaging data;

[0007] The plantar pressure field is reconstructed based on the plantar pressure matrix and the temperature matrix, and multimodal features of the plantar pressure field are extracted from the reconstructed pressure field.

[0008] A stress injury risk prediction model is established, and the plantar multimodal features are input into the stress injury risk prediction model to predict sports injury risk.

[0009] Preferably, the reconstruction of the plantar pressure field includes:

[0010] The plantar pressure field is reconstructed based on normalized spatial location mapping using an adaptive radial basis function multi-scale kernel fitting algorithm.

[0011] Preferably, the extraction of plantar multimodal features from the reconstructed pressure field includes:

[0012] The reconstructed pressure field is discretized and projected onto the finite element node mesh using a finite element mechanical model. A structural regularization optimization objective function is constructed that couples the pressure field with the finite element mechanical model. The characteristics of plantar pressure transmission and mechanical response are obtained by solving the objective function.

[0013] Preferably, the step of extracting multimodal features of the foot from the reconstructed pressure field further includes:

[0014] By combining geometrical statistical moment calculation, dynamic stability algorithm, spectral feature algorithm, and complexity entropy algorithm, multimodal feature vectors of the foot are extracted from the reconstructed pressure field data.

[0015] Preferably, the establishment of the stress damage risk prediction model includes:

[0016] The first layer is the feature embedding layer, which performs nonlinear mapping on the input plantar multimodal feature vector through a dual-hidden-layer fully connected network;

[0017] The second layer is the prediction layer, which uses a Soft-GradientBoost ensemble model and combines LightGBM, support vector machine and sparse regression classifier. It uses Soft-GradientBoost weighted fusion to predict risk probability.

[0018] Preferably, the establishment of the stress damage risk prediction model further includes:

[0019] The algorithm was optimized using the Focal-Tversky hybrid loss function, and the probability prediction of the model output was optimized using the Beta-three-parameter probability calibration method.

[0020] Preferred options also include:

[0021] Based on the predicted risk probability, the prediction results are divided into four levels: high risk, medium risk, low risk, and healthy. When the predicted risk probability is greater than a set threshold, it is judged as high risk.

[0022] This invention provides an intelligent prediction method for sports injury risk based on plantar modality. It reconstructs the plantar pressure field based on plantar pressure data and foot thermal imaging data, and extracts multimodal features of the plantar surface from the reconstructed pressure field. These features are then input into a stress injury risk pre-model to predict sports injury risk. This method addresses the problem that existing sports injury screening mainly relies on subjective diagnosis, which easily leads to a high rate of early missed diagnoses. It can improve the accuracy of sports injury risk prediction and increase the effectiveness of athlete training monitoring, rehabilitation assessment, and public fitness risk prevention and control. Attached Figure Description

[0023] To more clearly illustrate the specific embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below.

[0024] Figure 1 This diagram illustrates a method for intelligent prediction of sports injury risk based on plantar modality, as provided by the present invention. Detailed Implementation

[0025] To enable those skilled in the art to better understand the embodiments of the present invention, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings and implementation methods.

[0026] To address the high rate of early missed diagnoses in current sports injury screening, this invention provides an intelligent prediction method for sports injury risk based on plantar modality. This method solves the problem that existing sports injury screening mainly relies on subjective diagnosis, which easily leads to a high rate of early missed diagnoses. It can improve the accuracy of sports injury risk prediction and increase the effectiveness of athlete training monitoring, rehabilitation assessment, and public fitness risk prevention and control.

[0027] like Figure 1 As shown, a smart prediction method for sports injury risk based on plantar modality includes:

[0028] S1: Acquire plantar pressure data and foot thermal imaging data, and generate corresponding plantar pressure matrix and temperature matrix based on the plantar pressure data and the foot thermal imaging data.

[0029] S2: Reconstruct the plantar pressure field based on the plantar pressure matrix and the temperature matrix, and extract plantar multimodal features from the reconstructed pressure field.

[0030] S3: Establish a stress injury risk prediction model and input the plantar multimodal features into the stress injury risk prediction model to predict sports injury risk.

[0031] In practical applications, during real-world sports activities, the soft tissues of the foot are constantly subjected to a dual stress environment of high impact and high metabolism. Any early micro-injuries will leave extremely small and transient signal disturbances in the pressure and temperature fields. Therefore, a 64 × 64 piezoresistive-film hybrid array can be embedded in sports insoles to acquire plantar pressure data. Simultaneously, composite infrared luminescent fibers are spun into the sock surface, forming a temperature pixel array with coated thermistors to collect thermal imaging data of the foot. By reconstructing the plantar pressure field using the acquired plantar pressure and thermal imaging data, and extracting multimodal features from the reconstructed pressure field, these features are input into a stress injury risk pre-model for sports injury risk prediction. This addresses the problem that current sports injury screening relies mainly on subjective diagnosis, which easily leads to a high rate of early missed diagnoses. It can improve the accuracy of sports injury risk prediction and increase the effectiveness of athlete training monitoring, rehabilitation assessment, and public fitness risk prevention and control.

[0032] Furthermore, the reconstruction of the plantar pressure field includes:

[0033] The plantar pressure field is reconstructed based on normalized spatial location mapping using an adaptive radial basis function multi-scale kernel fitting algorithm.

[0034] Specifically, the distribution of the plantar pressure field is characterized by the following formula:

[0035] ;

[0036] , v): The reconstructed pressure field distribution.

[0037] L: Total number of kernel function scales.

[0038] Q: Number of sensor units.

[0039] Kernel function fitting coefficients.

[0040] Furthermore, the extraction of plantar multimodal features from the reconstructed pressure field includes:

[0041] The reconstructed pressure field is discretized and projected onto the finite element node mesh using a finite element mechanical model. A structural regularization optimization objective function is constructed that couples the pressure field with the finite element mechanical model. The characteristics of plantar pressure transmission and mechanical response are obtained by solving the objective function.

[0042] Specifically, optimize the objective function: J(u) is the objective function, K is the local stiffness matrix, u is the nodal displacement vector, f is the nodal force vector generated by the projection of the reconstructed pressure field, L is the higher-order smoothing operator, and β is the regularization adjustment coefficient.

[0043] Furthermore, the extraction of plantar multimodal features from the reconstructed pressure field also includes:

[0044] By combining geometrical statistical moment calculation, dynamic stability algorithm, spectral feature algorithm, and complexity entropy algorithm, multimodal feature vectors of the foot are extracted from the reconstructed pressure field data.

[0045] Furthermore, the establishment of the stress damage risk prediction model includes:

[0046] The first layer is the feature embedding layer, which performs nonlinear mapping on the input plantar multimodal feature vector through a dual-hidden-layer fully connected network;

[0047] The second layer is the prediction layer, which uses a Soft-GradientBoost ensemble model and combines LightGBM, support vector machine and sparse regression classifier. It uses Soft-GradientBoost weighted fusion to predict risk probability.

[0048] Furthermore, the establishment of the stress damage risk prediction model also includes:

[0049] The algorithm was optimized using the Focal-Tversky hybrid loss function, and the probability prediction of the model output was optimized using the Beta-three-parameter probability calibration method.

[0050] The method also includes: classifying the prediction results into four levels: high risk, medium risk, low risk, and healthy, based on the predicted risk probability; when the predicted risk probability is greater than a set threshold, it is judged as high risk.

[0051] In practical applications, the formulas for measuring healthy baseline, chronic fatigue stress, and acute hypertension injury adopt normalized cosine similarity: S AB ∈ [0, 1],

[0052] Where Φ(A) is a high-dimensional representation of a certain risk template, and Φ(B) is the current subject vector; the square root term in the denominator ensures that the metric only focuses on "direction deviation" rather than "amplitude scaling".

[0053] If we take the threshold values ​​τ1 = 0.80, τ2 = 0.65, and τ3 = 0.50, then:

[0054] S AB ≥ τ1 is considered "high risk" and requires immediate intervention.

[0055] τ2 ≤ S AB <τ1 is classified as "medium risk" and requires close monitoring.

[0056] τ3 ≤ S AB <τ2 is judged as "low risk", and it is recommended to reduce the amount of training.

[0057] S AB <τ3 is considered a healthy baseline.

[0058] Therefore, this invention provides an intelligent prediction method for sports injury risk based on plantar modality. It reconstructs the plantar pressure field based on plantar pressure data and foot thermal imaging data, and extracts multimodal features of the plantar surface from the reconstructed pressure field. These features are then input into a stress injury risk pre-model to predict sports injury risk. This addresses the problem that existing sports injury screening mainly relies on subjective diagnosis, which easily leads to a high rate of early missed diagnoses. It can improve the accuracy of sports injury risk prediction and increase the effectiveness of athlete training monitoring, rehabilitation assessment, and public fitness risk prevention and control.

[0059] The structure, features, and effects of the present invention have been described in detail above with reference to the embodiments shown in the figures. The above description is only a preferred embodiment of the present invention, but the present invention is not limited to the scope of implementation shown in the figures. Any changes made in accordance with the concept of the present invention, or equivalent embodiments modified to have equivalent changes, shall be within the protection scope of the present invention as long as they do not exceed the spirit covered by the specification and figures.

Claims

1. A smart prediction method for sports injury risk based on plantar modality, characterized in that, include: Acquire plantar pressure data and foot thermal imaging data, and generate corresponding plantar pressure matrix and temperature matrix based on the plantar pressure data and foot thermal imaging data; The plantar pressure field is reconstructed based on the plantar pressure matrix and the temperature matrix, and multimodal features of the plantar pressure field are extracted from the reconstructed pressure field. A stress injury risk prediction model is established, and the plantar multimodal features are input into the stress injury risk prediction model to predict sports injury risk.

2. The intelligent prediction method for sports injury risk based on plantar modality according to claim 1, characterized in that, The reconstruction of the plantar pressure field includes: The plantar pressure field is reconstructed based on normalized spatial location mapping using an adaptive radial basis function multi-scale kernel fitting algorithm.

3. The intelligent prediction method for sports injury risk based on plantar modality according to claim 2, characterized in that, The extraction of plantar multimodal features from the reconstructed pressure field includes: The reconstructed pressure field is discretized and projected onto the finite element node mesh using a finite element mechanical model. A structural regularization optimization objective function is constructed that couples the pressure field with the finite element mechanical model. The characteristics of plantar pressure transmission and mechanical response are obtained by solving the objective function.

4. The intelligent prediction method for sports injury risk based on plantar modality according to claim 3, characterized in that, The extraction of plantar multimodal features from the reconstructed pressure field also includes: By combining geometrical statistical moment calculation, dynamic stability algorithm, spectral feature algorithm, and complexity entropy algorithm, multimodal feature vectors of the foot are extracted from the reconstructed pressure field data.

5. The intelligent prediction method for sports injury risk based on plantar modality according to claim 4, characterized in that, The establishment of the stress damage risk prediction model includes: The first layer is the feature embedding layer, which performs nonlinear mapping on the input plantar multimodal feature vector through a dual-hidden-layer fully connected network; The second layer is the prediction layer, which uses a Soft-GradientBoost ensemble model and combines LightGBM, support vector machine and sparse regression classifier. It uses Soft-GradientBoost weighted fusion to predict risk probability.

6. The intelligent prediction method for sports injury risk based on plantar modality according to claim 5, characterized in that, The establishment of the stress damage risk prediction model also includes: The algorithm was optimized using the Focal-Tversky hybrid loss function, and the probability prediction of the model output was optimized using the Beta-three-parameter probability calibration method.

7. The intelligent prediction method for sports injury risk based on plantar modality according to claim 6, characterized in that, Also includes: Based on the predicted risk probability, the prediction results are divided into four levels: high risk, medium risk, low risk, and healthy. When the predicted risk probability is greater than a set threshold, it is judged as high risk.

Images