Dynamic follow-up system for maxillofacial deformity orthognathic surgery based on intelligent nursing platform

Abstract

The present invention discloses a dynamic follow-up system for orthognathic surgery of maxillofacial deformities based on an intelligent nursing platform. The system comprises: an image input and preprocessing module for acquiring images to generate a continuous time-series image sequence; a geometric flow modeling module for constructing a nonparametric partial differential geometric flow network model; a feature encoding module for extracting features of structural parameters; a parameter optimization module for executing the Kelvin-snake swarm resonance optimization algorithm; an inference and scoring module for executing inference, outputting a representation of the geometric flow evolution, generating a recovery status score and abnormality prompt information; and an intelligent nursing platform integration module for deploying the final nonparametric partial differential geometric flow network model, receiving patient images, and completing recovery monitoring, score output, abnormality marking, and follow-up information push. The present invention automates postoperative recovery modeling, improves assessment accuracy and follow-up efficiency, and facilitates precise management of intelligent nursing care.

Description

Technical Field

[0001] The present invention relates to the field of medical artificial intelligence technology, and in particular to a dynamic follow-up system for maxillofacial deformity orthognathic surgery based on an intelligent nursing platform. Background Art

[0002] Currently, in the rehabilitation management of maxillofacial deformity orthognathic surgery, the traditional follow-up model mainly relies on manual regular return visits, doctors' subjective judgments, and image comparison and analysis to monitor and guide patients' postoperative recovery. This method has obvious limitations such as long cycles, reliance on experience, and delayed intervention. Especially in the early postoperative recovery stage, the facial structure changes frequently, and traditional methods are difficult to achieve continuous modeling of the recovery process and regional differentiated management. Some hospitals and rehabilitation centers have tried to introduce digital platforms to assist in follow-up, such as electronic questionnaire systems and image upload interfaces, but they can only achieve data collection and archiving, lack intelligent evaluation and recovery trend analysis functions, and are difficult to meet the needs of refined and personalized rehabilitation.

[0003] In terms of modeling methods, existing studies have mostly used static deep neural networks or time series prediction models, using discriminative classifiers to roughly classify recovery states. These methods often ignore the structural evolution of facial regions in the temporal dimension and fail to establish coupled correlations between regions, resulting in a lack of continuity and spatial specificity when processing facial data from different time periods. In addition, the dynamic changes in facial recovery exhibit significant regional heterogeneity. For example, the maxillary and zygomatic bones, as well as facial soft tissues, exhibit significant differences in postoperative recovery speed and evolution direction. Existing technologies, which mostly use a unified parameter modeling approach, have difficulty capturing the temporal response differences between regions.

[0004] Traditional models often use grid search or gradient-based optimization algorithms for structural parameter optimization. These methods are prone to falling into local optima in high-dimensional spaces. This is particularly true when dealing with nonlinear evolving features and complex facial curvature variations. Their optimization results are unstable, convergence is slow, and model generalization is poor. Furthermore, existing optimization strategies typically lack adaptive perturbation mechanisms and are unable to dynamically adjust the optimization step size based on the current state of the model. This results in structural parameter convergence results being highly sensitive to initial values, increasing the difficulty of model training and the reliance on manual intervention.

[0005] Taking all of the above into account, existing technologies for postoperative recovery monitoring generally have the following shortcomings: First, they lack the ability to continuously model the dynamic process of maxillofacial surgery; second, they cannot establish a differentiated recovery parameter system based on different facial regions; and third, they suffer from poor optimization results and lack robustness in high-dimensional parameter spaces. Therefore, an intelligent follow-up system that can integrate time-series images and regional features, construct an interpretable geometric evolution model, and possess adaptive optimization capabilities is urgently needed to improve the accuracy and efficiency of postoperative recovery assessment.

[0006] Therefore, how to provide a dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on an intelligent nursing platform is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0007] One purpose of the present invention is to propose a dynamic follow-up system for orthognathic surgery of maxillofacial deformities based on non-parametric partial differential geometric flow modeling and Kelvin-snake resonance optimization algorithm. The present invention fully combines continuous time series image modeling, facial region structure parameterization and bio-inspired optimization methods, and describes in detail the whole process of continuous prediction and personalized scoring of postoperative facial recovery status. It has the advantages of visualization of recovery trends, automatic identification of abnormal areas and high evaluation accuracy.

[0008] The dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on an intelligent nursing platform according to an embodiment of the present invention includes:

[0009] Image input and preprocessing module, used to collect preoperative and postoperative images, perform regional calibration and standardization processing, and generate continuous time-series image sequences;

[0010] The geometric flow modeling module is used to build a non-parametric partial differential geometric flow network model, receive continuous time series image sequences and extract regional structure parameters;

[0011] Feature encoding module, used to extract the characteristics of regional structural parameters and generate regional weight parameters, evolution direction parameters and time scale control factors;

[0012] Parameter optimization module, used to execute the Kelvin-Snake Resonance optimization algorithm to optimize regional structural parameters;

[0013] The inference and scoring module is used to perform inference using the optimized non-parametric partial differential geometric flow network model, output the geometric flow evolution representation, and generate the recovery state score and anomaly prompt information;

[0014] The intelligent nursing platform integration module is used to deploy the final non-parametric partial differential geometric flow network model, receive the patient's postoperative images, and complete recovery monitoring, score output, abnormality marking and follow-up information push.

[0015] Optionally, modules can be connected using the following methods:

[0016] S1. Obtain preoperative images and images of several time periods after surgery, perform regional calibration and standardization processing, and generate a continuous time-series image sequence;

[0017] S2. Based on the continuous time-series image sequence, a non-parametric partial differential geometric flow network model is constructed. The non-parametric partial differential geometric flow network model includes a regional weight parameter, an evolution direction parameter, and a time scale control factor;

[0018] S3. Input the continuous time-series image sequence into the nonparametric partial differential geometric flow network model, encode the maxillofacial region through the local geometric kernel function, and generate a geometric flow evolution representation sequence containing regional weight parameters, evolution direction parameters and time scale control factors;

[0019] S4. Construct a Kelvin-snake swarm resonance optimization algorithm and initialize a population parameter set including a multi-frequency perturbation mechanism and a curvature adjustment factor;

[0020] S5. Using the Kelvin-snake swarm resonance optimization algorithm, we jump out of local extreme points under the action of the multi-frequency perturbation mechanism, adaptively adjust the step size under the control of the curvature adjustment factor, iteratively optimize the geometric flow evolution representation sequence, and generate a non-parametric partial differential geometric flow network model after structural optimization;

[0021] S6. Deploy the optimized nonparametric partial differential geometric flow network model to the intelligent nursing platform, receive the patient's postoperative images, perform geometric flow reasoning, and output the recovery status score and abnormal prompt information;

[0022] S7. Generate a follow-up feedback information package based on the recovery status score and abnormal prompt information, and complete archiving and push.

[0023] Optionally, the S2 specifically includes:

[0024] S21, construct a nonparametric partial differential geometric flow network model based on a continuous time series image sequence, wherein the continuous time series image sequence is a set of preoperative images and postoperative two-dimensional facial images at several time periods arranged in chronological order, and is set as ,in, Indicates the The 2D facial image submanifold of time steps, is the total number of time steps;

[0025] S22, in each two-dimensional facial image submanifold Upper Division facial regions, defining a set of regional weight parameters ,in, Represents the time step Time The weight coefficient of each region;

[0026] S23. Define the evolution direction parameter set ,in, Represents the time step Time The two-dimensional evolution direction vector of each region is used to represent the spatial guidance feature of the morphological changes of the facial region;

[0027] S24. Define a set of time scale control factors ,in, Indicates the The time response factor of each facial region during the restoration process is used to control the difference in geometric evolution speed between different regions;

[0028] S25, based on regional weight parameters , evolution direction parameters and time scale control factors Construct nonparametric partial differential evolution and transform , As a trainable structural parameter of the non-parametric partial differential geometric flow network model, the initial structure setting of the model is completed.

[0029] Optionally, the S3 specifically includes:

[0030] S31. For each two-dimensional facial image submanifold Perform region division to obtain a facial region set ,in Represents the time step Next facial subregions;

[0031] S32. In each facial sub-region A local geometric kernel function is constructed on the surface of the image. The local geometric kernel function is an embedded computing unit that extracts the geometric structural features of the curvature change, boundary direction and texture density in the region, and is used to map the two-dimensional facial image submanifold into a structural feature vector. The generated structural feature vector is recorded as ;

[0032] S33, according to the structural feature vector Initializing the regional weight parameters in the nonparametric partial differential geometric flow network model , evolution direction parameters and time scale control factors ,in:

[0033] Regional weight parameter By structural feature vector Estimation of the mean geometric gradient modulus in , where the gradient modulus reflects the local curvature and deformation intensity of the region;

[0034] Evolution direction parameter By structural feature vector The main direction characteristic component in is calculated, and the direction is the vector direction corresponding to the main component;

[0035] Time scale control factor By facial The deformation degree of the region in the full time series image is calculated, and the standard deviation of the change of the structural characteristics of the region at different time steps is used as the estimated value of the response factor;

[0036] S34, based on , and Constructing a geometric flow evolution representation ,in, Indicates that at time step Next, The geometric evolution intensity of the region is used to characterize the local morphological changes of the face during the restoration process, and the geometric flow evolution representation set Structural output as a nonparametric partial differential geometric flow network model.

[0037] Optionally, the S4 specifically includes:

[0038] S41. Construct a Kelvin-snake swarm resonance optimization algorithm to optimize the regional weight parameters in the nonparametric partial differential geometric flow network model. , evolution direction parameters and time scale control factors , structural perturbation and convergent search are achieved by simulating the coordinated action of snakes in a multi-frequency resonance field;

[0039] S42. Setting the optimized population parameter set ,in, is the population size, each individual Contains a structure parameter ;

[0040] S43. For each individual Assign independent multi-frequency perturbation vectors , where each component Indicates the The resonance frequency corresponding to each structural parameter dimension, and the disturbance frequency is obtained by random sampling in the preset frequency range;

[0041] S44. Introduce a set of curvature adjustment factors in each structural parameter dimension , curvature adjustment factor Dynamically adjust the perturbation amplitude of each parameter dimension based on the gradient change rate calculated in the geometric flow evolution representation in the previous stage;

[0042] S45. Construct the multi-frequency resonance perturbation expression, and set Individuals in The perturbation vector of the structural parameter dimension is ;

[0043] S46, each initial individual The structural parameters and perturbation vector Binding to form structural parameters with perturbations , as the initial population input of the Kelvin-snake swarm resonance optimization algorithm.

[0044] Optionally, the S5 specifically includes:

[0045] S51. Based on the initial population input, in each round of iteration, the structural parameters are periodically perturbed according to the multi-frequency perturbation vector corresponding to the initial individual, so that the structural parameters jump out of the local extreme value region. At the same time, according to the curvature adjustment factor of the dimension of the structural parameter, the perturbation amplitude is adaptively adjusted, taking a preset small step size in the changing region and a preset large step size in the smooth region, dynamically controlling the update amplitude of the structural parameters, and generating new structural parameter candidate individuals;

[0046] S52, inputting the new structural parameter candidate individuals into the nonparametric partial differential geometric flow network model, executing the inference process, and obtaining the geometric flow evolution results of each facial region at each time step;

[0047] S53, comparing the geometric flow evolution result output by the model with the actual recovery state, calculating the fitting effect, and assigning a fitness score;

[0048] S54. Score and sort all individuals in the population, select the individuals with the best fitness for the next round of retention, and replace the individuals with low fitness based on the multi-frequency perturbation mechanism to generate new parameter combinations to expand the search space;

[0049] S55. Repeat the parameter update, model inference and population iteration process until the preset convergence conditions are met or the maximum number of rounds is reached. Finally, the regional weight parameters, evolution direction parameters and time scale control factors of the individuals with the best fitness are extracted to form a non-parametric partial differential geometric flow network model after structural optimization.

[0050] Optionally, the recovery status score in step S6 is generated based on the evolution amplitude and trend stability of each region, and the abnormal prompt information is used to identify the regional location and time period of the local recovery abnormality.

[0051] Optionally, the follow-up feedback information package of step S7 includes a recovery trend chart, rehabilitation suggestion text and interactive instructions, and archiving and pushing are completed based on the follow-up feedback information package.

[0052] The beneficial effects of the present invention are:

[0053] First, by constructing a non-parametric partial differential geometric flow network model, the present invention can simulate the continuous geometric evolution process of the facial area after maxillofacial surgery in the time dimension, realize dynamic modeling of postoperative recovery status, overcome the defect that traditional static neural networks cannot capture temporal changes, and effectively improve the continuity of recovery process modeling and regional expression accuracy.

[0054] Secondly, the present invention introduces three types of structural parameters: regional weight parameters, evolution direction parameters and time scale control factors, and independently models the differential recovery characteristics of each anatomical region of the face, realizing multi-dimensional control of recovery speed, dominant deformation direction and influence intensity, enhancing the model's ability to depict facial heterogeneity and regional differences, and meeting the needs of refined evaluation.

[0055] In addition, the present invention adopts the Kelvin-snake swarm resonance optimization algorithm to optimize the model structure parameters. Combined with the multi-frequency perturbation mechanism and the curvature adjustment factor, it can effectively jump out of the local extreme point and dynamically adjust the optimization step size, thereby improving the search efficiency and convergence stability in the high-dimensional structure space, and solving the problem that traditional optimization methods are prone to falling into local optimality. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0057] Figure 1 This is a schematic diagram of the module composition of the dynamic follow-up system for maxillofacial deformity orthognathic surgery based on the intelligent nursing platform proposed in the present invention;

[0058] Figure 2 This is a method flow chart of the dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on the intelligent nursing platform proposed by the present invention;

[0059] Figure 3 This is the initialization structure diagram of the Kelvin-snake swarm resonance optimization algorithm for the dynamic follow-up system after orthognathic surgery for maxillofacial deformity based on the intelligent nursing platform proposed in the present invention. DETAILED DESCRIPTION

[0060] The present invention will now be described in further detail with reference to the accompanying drawings, which are simplified schematic diagrams that illustrate the basic structure of the present invention in a schematic manner.

[0061] refer to Figure 1-3 , a dynamic follow-up system for maxillofacial deformity orthognathic surgery based on an intelligent nursing platform, including:

[0062] Image input and preprocessing module, used to collect preoperative and postoperative images, perform regional calibration and standardization processing, and generate continuous time-series image sequences;

[0063] The geometric flow modeling module is used to build a non-parametric partial differential geometric flow network model, receive continuous time series image sequences and extract regional structure parameters;

[0064] Feature encoding module, used to extract the characteristics of regional structural parameters and generate regional weight parameters, evolution direction parameters and time scale control factors;

[0065] Parameter optimization module, used to execute the Kelvin-Snake Resonance optimization algorithm to optimize regional structural parameters;

[0066] The inference and scoring module is used to perform inference using the optimized non-parametric partial differential geometric flow network model, output the geometric flow evolution representation, and generate the recovery state score and anomaly prompt information;

[0067] The intelligent nursing platform integration module is used to deploy the final non-parametric partial differential geometric flow network model, receive the patient's postoperative images, and complete recovery monitoring, score output, abnormality marking and follow-up information push.

[0068] The present invention clearly divides functional units such as image input, modeling, feature extraction, parameter optimization, inference scoring and platform integration. The logic between modules is clear and the data is closed-loop, which facilitates actual system deployment and functional expansion. Compared with existing closed algorithm process solutions, it has the advantages of being more structured, integrable and engineering feasible.

[0069] In this embodiment, the modules are connected through the following methods:

[0070] S1. Acquire preoperative images and images of several time periods after surgery, perform regional calibration and standardization processing, and generate a continuous time-series image sequence, specifically including: collecting preoperative images of the patient and images of multiple key time nodes after surgery to ensure that the structural evolution process during the entire recovery cycle is covered. After image acquisition, the system automatically calibrates the facial area of ​​the image, including the positioning of key anatomical areas such as the zygomatic bone, mandibular angle, maxilla, and chin. Subsequently, the image data will be processed through size normalization, brightness standardization, and posture correction to eliminate interference caused by shooting conditions and individual differences, ensuring that images at different time points are comparable. After processing, the image sequence is constructed into a continuous time-series image sequence in a unified format in chronological order;

[0071] S2. Based on the continuous time-series image sequence, a non-parametric partial differential geometric flow network model is constructed. The non-parametric partial differential geometric flow network model includes a regional weight parameter, an evolution direction parameter, and a time scale control factor;

[0072] S3. Input the continuous time-series image sequence into the nonparametric partial differential geometric flow network model, encode the maxillofacial region through the local geometric kernel function, and generate a geometric flow evolution representation sequence containing regional weight parameters, evolution direction parameters and time scale control factors;

[0073] S4. Construct a Kelvin-snake swarm resonance optimization algorithm and initialize a population parameter set including a multi-frequency perturbation mechanism and a curvature adjustment factor;

[0074] S5. Using the Kelvin-snake swarm resonance optimization algorithm, we jump out of local extreme points under the action of the multi-frequency perturbation mechanism, adaptively adjust the step size under the control of the curvature adjustment factor, iteratively optimize the geometric flow evolution representation sequence, and generate a non-parametric partial differential geometric flow network model after structural optimization;

[0075] S6. Deploy the optimized nonparametric partial differential geometric flow network model to the intelligent nursing platform, receive the patient's postoperative images, perform geometric flow reasoning, and output the recovery status score and abnormal prompt information;

[0076] S7. Generate a follow-up feedback information package based on the recovery status score and abnormal prompt information, and complete archiving and push.

[0077] The present invention constructs an intelligent follow-up system based on non-parametric partial differential geometric flow modeling and Kelvin-snake swarm resonance optimization algorithm, breaking the limitations of traditional postoperative rehabilitation methods that rely on static image comparison and manual experience evaluation, and realizing continuous modeling, regional difference analysis and parameter adaptive optimization of the postoperative recovery process, effectively improving the accuracy and efficiency of recovery prediction, and has the technical advantages of high modeling flexibility, strong regional resolution and intelligent system response.

[0078] In this embodiment, S2 specifically includes:

[0079] S21, construct a nonparametric partial differential geometric flow network model based on a continuous time series image sequence, wherein the continuous time series image sequence is a set of preoperative images and postoperative two-dimensional facial images at several time periods arranged in chronological order, and is set as ,in, Indicates the The 2D facial image submanifold of time steps, is the total number of time steps;

[0080] S22, in each two-dimensional facial image submanifold Upper Division facial regions, defining a set of regional weight parameters ,in, Represents the time step Time The weight coefficient of each region;

[0081] S23. Define the evolution direction parameter set ,in, Represents the time step Time The two-dimensional evolution direction vector of each region is used to represent the spatial guidance feature of the morphological changes of the facial region;

[0082] S24. Define a set of time scale control factors ,in, Indicates the The time response factor of each facial region during the restoration process is used to control the difference in geometric evolution speed between different regions;

[0083] S25, based on regional weight parameters , evolution direction parameters and time scale control factors Construct nonparametric partial differential evolution and transform , As a trainable structural parameter of the nonparametric partial differential geometric flow network model, it completes the setting of the initial structure of the model. The expression is as follows:

[0084] ;

[0085] in, Indicates the facial regions at time step The local geometric change rate under is the evolution direction parameter The divergence value is used to control the expansion or contraction trend of the area.

[0086] Through this formula, the present invention couples the regional weight parameter of the time scale control factor and the divergence term of the evolution direction as the driving core of the facial region morphological change rate, so that the dynamic changes of each region during the recovery process not only have spatial direction guidance, but also have the ability to regulate time and adjust importance weights. Compared with traditional models that only consider image pixel changes or unified evolution methods, this structure has the advantages of strong interpretability, fine control granularity, and obvious regional response differentiation. It can more realistically reflect the expansion or contraction trend of the facial region during postoperative recovery, thereby significantly improving the robustness and accuracy of the model in time series prediction and anomaly recognition. Especially in scenarios with strong regional heterogeneity and asynchronous recovery rhythms, this formula provides a precise, controllable, and dynamic evolutionary modeling method with strong clinical adaptability and intelligent analysis value.

[0087] The present invention innovatively introduces regional weight parameters, evolution direction parameters and time scale control factors to construct a multi-dimensional structural expression method for describing the evolutionary process of maxillofacial restoration. Different from the unified weight modeling strategy of existing models, it can realize the controllability of regional difference modeling and restoration rhythm, thereby enhancing the model's expression and prediction capabilities for complex facial deformation structures.

[0088] In this embodiment, S3 specifically includes:

[0089] S31. For each two-dimensional facial image submanifold The regional division is carried out, specifically including: firstly, extracting standard feature points in the image through the facial key point detection algorithm, such as brow bone, eye corner, nose wing, lip corner, zygomatic arch, mandibular margin, etc.; then constructing a facial geometric topology map based on these key points, dividing the entire image into several facial regions with physiological significance, such as zygomatic bone area, nose area, mandibular angle area and lower lip area, etc., to obtain the facial region set ,in Represents the time step Next facial subregions;

[0090] S32. In each facial sub-region A local geometric kernel function is constructed on the basis of the above. Constructing a local geometric kernel function means that in each divided facial sub-region, the local geometric structure information of the image is used to generate a kernel function mapping method for feature extraction. The process first extracts low-dimensional geometric descriptors representing the regional morphology based on geometric elements such as boundary curvature, grayscale gradient and texture direction in the region. Subsequently, these descriptors are mapped to a high-dimensional feature space using a kernel method. Common kernels include Gaussian kernels, Laplacian kernels or custom kernels based on structural tensors, the purpose of which is to amplify regional detail differences. The constructed local geometric kernel function has nonlinear mapping capabilities and can capture subtle deformations and structural changes in facial regions under different temporal states. Finally, each kernel function forms a set of structural feature vectors in the corresponding region, which serve as the basic input for subsequent structural parameter generation, thereby realizing a structured transformation from image to parameter space. The local geometric kernel function is an embedded computing unit that extracts geometric structural features such as curvature changes, boundary directions and texture density within the region, and is used to map the two-dimensional facial image submanifold into a structural feature vector. The generated structural feature vector is recorded as ;

[0091] S33, according to the structural feature vector Initializing the regional weight parameters in the nonparametric partial differential geometric flow network model , evolution direction parameters and time scale control factors ,in:

[0092] Regional weight parameter By structural feature vector The geometric gradient modulus mean estimation in specifically includes: first extracting the gradient vector of each pixel in the region, for example, by calculating using the Sobel operator or the structure tensor, then taking the modulus of each gradient vector to obtain the gradient intensity distribution, and finally calculating the mean of the gradient modulus values ​​of all pixels in the region, where the gradient modulus value reflects the local curvature and deformation intensity of the region;

[0093] Evolution direction parameter By structural feature vector The main direction feature components in the region are calculated, specifically including: first constructing a structural tensor matrix in the region, fusing horizontal and vertical gradient information, then performing eigenvalue decomposition on the tensor, extracting the eigenvector corresponding to the maximum eigenvalue as the main direction vector component of the region, and taking the vector direction corresponding to the principal component;

[0094] Time scale control factor By facial The calculation of the deformation degree of a region in the full time series image specifically includes: first, registering and tracking the same sub-region in the continuous time series image to extract its key geometric feature points; then calculating the spatial displacement or contour change amplitude of these feature points at different time points; finally, normalizing the change value during the entire time series process to obtain the cumulative deformation degree index of the region, and using the standard deviation of the structural feature change of the region at different time steps as the response factor estimate;

[0095] S34, based on , and Constructing a geometric flow evolution representation ,in, Indicates that at time step Next, The geometric evolution intensity of the region is used to characterize the local morphological changes of the face during the restoration process, and the geometric flow evolution representation set Structural output as a nonparametric partial differential geometric flow network model.

[0096] The present invention constructs a regional feature vector through a local geometric kernel function, and generates three types of structural parameters based on the vector: regional weight, evolution direction and time regulation, forming an initial structural representation that can be used for evolutionary modeling. It solves the problem that traditional image features are insufficiently expressive in the time dimension and cannot support time series recovery reasoning, and significantly improves the regional-level dynamic prediction effect.

[0097] In this embodiment, the S4 specifically includes:

[0098] S41. Construct a Kelvin-snake swarm resonance optimization algorithm to optimize the regional weight parameters in the nonparametric partial differential geometric flow network model. , evolution direction parameters and time scale control factors , structural perturbation and convergent search are achieved by simulating the coordinated action of snakes in a multi-frequency resonance field;

[0099] S42. Setting the optimized population parameter set ,in, is the population size, each individual Contains a structure parameter ;

[0100] S43. For each individual Assign independent multi-frequency perturbation vectors , where each component Indicates the The resonance frequency corresponding to each structural parameter dimension, and the disturbance frequency is obtained by random sampling in the preset frequency range;

[0101] S44. Introduce a set of curvature adjustment factors in each structural parameter dimension , curvature adjustment factor Dynamically adjust the perturbation amplitude of each parameter dimension based on the gradient change rate calculated in the geometric flow evolution representation in the previous stage;

[0102] S45. Construct the multi-frequency resonance perturbation expression, and set Individuals in The perturbation vector of the structural parameter dimension is :

[0103] ;

[0104] in, is the basic amplitude, is pi, is the iteration step, is the phase offset, is the curvature adjustment factor, Indicates the perturbation step size of the corresponding dimension;

[0105] The present invention realizes dynamic perturbation control of model parameters during the structural optimization process by constructing a multi-frequency resonance perturbation expression, combining the basic amplitude, multi-frequency perturbation frequency, phase offset and curvature adjustment factor. This formula not only allows each parameter dimension to be periodically perturbated in an independent frequency space, but also introduces a curvature adjustment mechanism so that the perturbation amplitude can be flexibly adjusted according to the gradient change of the current model state. Compared with traditional fixed step size or single frequency perturbation optimization algorithms, this mechanism significantly improves the flexibility of parameter space exploration and the ability to escape local optimality, effectively improves the optimization convergence speed and structural adaptability, and provides a refined adjustment basis for the efficient optimization of high-dimensional geometric parameters.

[0106] S46, each initial individual The structural parameters and perturbation vector Binding to form structural parameters with perturbations , as the initial population input of the Kelvin-snake swarm resonance optimization algorithm.

[0107] This method, by introducing a multi-frequency perturbation mechanism and a curvature adjustment factor, achieves intelligent perturbation control during the model structure parameter search process. This method effectively avoids the local optimum or parameter update rigidity issues that plague traditional optimization algorithms, enhancing the hops and convergence stability of high-dimensional parameter space searches. It offers significant performance advantages over genetic algorithms and particle swarm optimization in structural optimization tasks.

[0108] In this embodiment, the S5 specifically includes:

[0109] S51. Based on the initial population input, in each round of iteration, the structural parameters are periodically perturbed according to the multi-frequency perturbation vector corresponding to the initial individual, so that the structural parameters jump out of the local extreme value region. At the same time, according to the curvature adjustment factor of the dimension of the structural parameter, the perturbation amplitude is adaptively adjusted, taking a preset small step size in the changing region and a preset large step size in the smooth region, dynamically controlling the update amplitude of the structural parameters, and generating new structural parameter candidate individuals;

[0110] S52, inputting the new structural parameter candidate individuals into the nonparametric partial differential geometric flow network model, executing the inference process, and obtaining the geometric flow evolution results of each facial region at each time step;

[0111] S53. Compare the geometric flow evolution results output by the model with the actual restored state, calculate the fitting effect, and assign a fitness score. The specific process is as follows: first, extract the structural parameters of the key areas in the reconstructed image or feature curve output by the model; then, perform spatial alignment and feature correspondence with the real time series image at the same time point, and calculate the differences between the two in contour position, gradient direction, or geometric deformation amplitude; finally, quantify these differences as fitting error indicators to reflect the model's ability to restore postoperative facial changes at the current stage, providing a basis for further scoring and optimization;

[0112] S54. Score and sort all individuals in the population, select the individuals with the best fitness for the next round of retention, and replace the individuals with low fitness based on the multi-frequency perturbation mechanism to generate new parameter combinations to expand the search space;

[0113] S55. Repeat the parameter update, model inference and population iteration process until the preset convergence conditions are met or the maximum number of rounds is reached. Finally, the regional weight parameters, evolution direction parameters and time scale control factors of the individuals with the best fitness are extracted to form a non-parametric partial differential geometric flow network model after structural optimization.

[0114] The present invention constructs a parameter update process based on disturbance control and difference feedback evaluation. By dynamically adjusting the parameters under the action of multi-frequency disturbance and adaptive regulation mechanism, the model structure fine-tuning and fitting capability improvement required for high-precision recovery state scoring are achieved, solving the problem of insufficient model accuracy caused by fixed parameters or single training path in existing models.

[0115] In this embodiment, the recovery status score of step S6 is generated based on the evolution amplitude and trend stability of each region, and the abnormal prompt information is used to identify the regional location and time period of local recovery abnormality, specifically including: firstly analyzing the deformation trajectory of each facial sub-region in the full time series image, extracting its evolution amplitude change value and the change trend in the time series, the system calculates the size of the morphological change of each region at different time points, and evaluates the continuity and stability of these changes. When an abnormal mutation occurs in a certain area, the recovery speed is abnormally slow or fluctuates violently, the system marks the area and the corresponding time period as an abnormal point, and generates abnormal prompt information to prompt clinical attention.

[0116] In this embodiment, the follow-up feedback information package in step S7 includes a recovery trend chart, rehabilitation advice text, and interactive instructions. Archiving and push notifications are completed based on this follow-up feedback information package. Specifically, after the follow-up visit, the collected facial recovery images, scoring results, and abnormality prompt information are uniformly processed. Based on the recovery scores and trend changes in each area, the system generates a visual recovery trend chart; combines postoperative recovery rules with model evaluation results to automatically generate personalized rehabilitation advice text; and extracts interactive instructions based on user interaction behavior to guide the next step of nursing care. Ultimately, this information is packaged into a follow-up feedback information package, automatically archived in the user's nursing file, and pushed to medical staff and patients through the platform, achieving closed-loop follow-up management.

[0117] Example 1:

[0118] To verify the feasibility of this invention, it was applied to a remote rehabilitation management project for orthognathic surgery conducted by the Maxillofacial Surgery Department of a tertiary comprehensive hospital. This project aims to improve the efficiency of postoperative follow-up visits and the accuracy of recovery status assessments, alleviating the challenges of limited manual follow-up resources, subjective scoring, and delayed abnormality identification. In actual use, 60 patients undergoing correction of mandibular protrusion or maxillary underdevelopment were selected. All patients uploaded a total of approximately 3,400 facial images between the first and 90th days after surgery, covering daily changes during the critical postoperative recovery phase.

[0119] Before the system was deployed, traditional follow-up assessments used biweekly face-to-face consultations and postoperative satisfaction questionnaires, combined with visual evaluation by the physician. Scoring was primarily based on facial swelling, mouth cleft, and subjective bite feedback. Each follow-up visit required an average of 22 minutes of manual processing time per patient, and the subjective scoring consistency rate was only 73%. However, after the system was deployed, patients can automatically upload facial images via a mobile app. After the images are standardized through the image input and preprocessing modules, they are fed into a constructed nonparametric partial differential geometric flow network model. The system generates a real-time recovery evolution intensity score for each region based on the model's output of regional weight parameters, evolution direction parameters, and time scale control factors.

[0120] During the initial deployment, the system introduced expert-labeled samples to construct initial evolutionary reference values. The system then used the Kelvin-Snake Resonance optimization algorithm to automatically optimize the network structure parameters, achieving stable convergence after approximately 80 rounds of optimization iterations. The trained model achieved a mean recovery score error of ±3.7 points (out of 100) on real test data, outperforming the ±9.1 points obtained from traditional questionnaires and physician judgment. In samples with significant fluctuations in recovery scores within the 14th day after surgery, the system automatically identified 37 abnormal areas, 32 of which were confirmed by physicians to be genuine issues of persistent early edema, local infection, or occlusal deviation, achieving an abnormality identification accuracy rate of 86.5%. For the eight cases in which patients reported discomfort but showed no abnormalities in traditional scoring, the system was also able to identify deviations in the evolutionary trend of the zygomatic arch region, effectively assisting physicians in making early interventions.

[0121] The platform automatically generates recovery trend graphs, scoring reports, and abnormality alerts, which are aggregated and pushed to the attending physician and patient's mobile devices, and archived in the hospital's rehabilitation system. Actual deployment data shows that the system takes an average of no more than 1.8 seconds to score each patient, increasing the efficiency of assisted follow-up by nearly 12 times and raising the consistency rate of medical and nursing scores to 92.6%. In terms of questionnaire recovery rate and patient satisfaction, system tracking shows that the patient's active cooperation rate within 90 days has increased from the original 58% to 91%, and the average number of times patients actively upload images is 47 times per person, a 1.9-fold increase compared to the traditional model.

[0122] In practical applications, this invention significantly improves the efficiency of postoperative follow-up, the accuracy of scoring, and the timeliness of abnormal prompts, verifying its practical feasibility and promotion value in postoperative dynamic rehabilitation management.

[0123] The following table shows the key data statistics collected and evaluated during the project implementation process of this invention:

[0124] Table 1: Dynamic recovery modeling and scoring performance evaluation statistics after orthognathic surgery

[0125]

[0126] During the implementation of this dynamic follow-up system after orthognathic surgery, a comparative analysis was conducted using the "Statistical Table for Dynamic Recovery Modeling and Scoring Performance Evaluation After Orthognathic Surgery" to compare traditional follow-up methods with the proposed system. The average follow-up time required for each patient was 22 minutes for traditional methods, while this system only required 1.8 seconds. This demonstrates an order of magnitude improvement in follow-up efficiency and significantly reduces the workload for medical staff. Regarding recovery scoring accuracy, the traditional scoring error was approximately ±9.1 points, while this system, through model evolution output, reduced this error to ±3.7 points, effectively improving the accuracy of quantitative assessments.

[0127] In terms of anomaly identification, the system achieved an accuracy rate of 86.5%, significantly outperforming the 38.2% rate of traditional manual identification, demonstrating the significant impact of geometric flow modeling and structural parameter optimization. Furthermore, the consistency rate of medical and nursing scores increased from 73.0% to 92.6%, demonstrating the system's higher professionalism and consistency in output, making it more reliable in assisting physicians in making decisions. The patient cooperation rate also increased from 58.0% to 91.0%, demonstrating the intelligent platform's enhanced user experience and user-friendly feedback process.

[0128] In terms of the frequency of patient image uploads, the system guided an average of 47 uploads per patient, significantly higher than the 25 per patient achieved with traditional methods. This ensured the quality of the model's sequential input and the continuity of recovery trend modeling. Furthermore, during model training and deployment, an average of 80 rounds of structural optimization iterations were completed, ensuring the rational use of computing resources while enabling fine-tuning of the model's structure and supporting the efficient operation of the final scoring and prompting modules. The system's automatic scoring report generation time was controlled within 3 seconds, providing strong support for real-time feedback from the remote care platform.

[0129] Overall, the data in this table clearly demonstrates the technical advantages of the present invention in the postoperative recovery assessment process. From scoring accuracy, processing speed, user participation to abnormality detection capabilities, it is superior to traditional methods in all aspects, verifying the actual implementation value of this system in the dynamic follow-up scenario after maxillofacial deformity surgery.

[0130] The above description is only a preferred specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any technician familiar with the technical field, within the technical scope disclosed by the present invention, who makes equivalent replacements or changes based on the technical solution and inventive concept of the present invention, should be covered by the scope of protection of the present invention.

Claims

1. The dynamic follow-up system for maxillofacial deformity orthognathic surgery based on the intelligent nursing platform is characterized by: include: Image input and preprocessing module, used to collect preoperative and postoperative images, perform regional calibration and standardization processing, and generate continuous time-series image sequences; The geometric flow modeling module is used to build a non-parametric partial differential geometric flow network model, receive continuous time series image sequences and extract regional structure parameters; Feature encoding module, used to extract the characteristics of regional structural parameters and generate regional weight parameters, evolution direction parameters and time scale control factors; a parameter optimization module for executing a Kelvin-snake swarm resonance optimization algorithm to optimize regional structural parameters; wherein the Kelvin-snake swarm resonance optimization algorithm iteratively optimizes the regional structural parameters by assigning multi-frequency perturbation vectors to individuals in the optimization population and introducing a curvature adjustment factor to dynamically adjust the perturbation amplitude; The inference and scoring module is used to perform inference using the optimized non-parametric partial differential geometric flow network model, output the geometric flow evolution representation, and generate the recovery state score and anomaly prompt information; The intelligent nursing platform integration module is used to deploy the final nonparametric partial differential geometric flow network model, receive postoperative patient images, and complete recovery monitoring, score output, abnormality marking, and follow-up information push; The modules are implemented as follows: S1. Obtain preoperative images and images of several time periods after surgery, perform regional calibration and standardization processing, and generate a continuous time-series image sequence; S2. Based on the continuous time-series image sequence, a non-parametric partial differential geometric flow network model is constructed. The non-parametric partial differential geometric flow network model includes a regional weight parameter, an evolution direction parameter, and a time scale control factor; S3. Input the continuous time-series image sequence into the nonparametric partial differential geometric flow network model, encode the maxillofacial region through the local geometric kernel function, and generate a geometric flow evolution representation sequence containing regional weight parameters, evolution direction parameters and time scale control factors; S4. Construct a Kelvin-snake swarm resonance optimization algorithm and initialize a population parameter set including a multi-frequency perturbation mechanism and a curvature adjustment factor; S5. Using the Kelvin-snake swarm resonance optimization algorithm, we jump out of local extreme points under the action of the multi-frequency perturbation mechanism, adaptively adjust the step size under the control of the curvature adjustment factor, iteratively optimize the geometric flow evolution representation sequence, and generate a non-parametric partial differential geometric flow network model after structural optimization; S6. Deploy the optimized nonparametric partial differential geometric flow network model to the intelligent nursing platform, receive the patient's postoperative images, perform geometric flow reasoning, and output the recovery status score and abnormal prompt information; S7. Generate a follow-up feedback information package based on the recovery status score and abnormal prompt information, and complete archiving and push; The S4 specifically includes: S41. Construct a Kelvin-snake swarm resonance optimization algorithm to optimize the regional weight parameters in the nonparametric partial differential geometric flow network model. , evolution direction parameters and time scale control factors , structural perturbation and convergent search are achieved by simulating the coordinated action of snakes in a multi-frequency resonance field; S42. Setting the optimized population parameter set ,in, is the population size, each individual Contains a structure parameter ; S43. For each individual Assign independent multi-frequency perturbation vectors , where each component Indicates the The resonance frequency corresponding to each structural parameter dimension, and the disturbance frequency is obtained by random sampling in the preset frequency range; S44. Introduce a set of curvature adjustment factors in each structural parameter dimension , curvature adjustment factor Dynamically adjust the perturbation amplitude of each parameter dimension based on the gradient change rate calculated in the geometric flow evolution representation in the previous stage; S45. Construct the multi-frequency resonance perturbation expression, and set Individuals in The perturbation vector of the structural parameter dimension is ; S46, each initial individual The structural parameters and perturbation vector Binding to form structural parameters with perturbations , as the initial population input of the Kelvin-snake swarm resonance optimization algorithm.

2. The dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on the intelligent nursing platform according to claim 1 is characterized in that: The S2 specifically includes: S21, construct a nonparametric partial differential geometric flow network model based on a continuous time series image sequence, wherein the continuous time series image sequence is a set of preoperative images and postoperative two-dimensional facial images at several time periods arranged in chronological order, and is set as ,in, Indicates the The 2D facial image submanifold of time steps, is the total number of time steps; S22, in each two-dimensional facial image submanifold Upper Division facial regions, defining a set of regional weight parameters ,in, Represents the time step Time The weight coefficient of each region; S23. Define the evolution direction parameter set ,in, Represents the time step Time The two-dimensional evolution direction vector of each region is used to represent the spatial guidance feature of the morphological changes of the facial region; S24. Define a set of time scale control factors ,in, Indicates the The time response factor of each facial region during the restoration process is used to control the difference in geometric evolution speed between different regions; S25, based on regional weight parameters , evolution direction parameters and time scale control factors Construct nonparametric partial differential evolution and transform , As a trainable structural parameter of the non-parametric partial differential geometric flow network model, the initial structure setting of the model is completed.

3. The dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on the intelligent nursing platform according to claim 2 is characterized in that: The S3 specifically includes: S31. For each two-dimensional facial image submanifold Perform region division to obtain a facial region set ,in Represents the time step Next facial subregions; S32. In each facial sub-region A local geometric kernel function is constructed on the surface of the image. The local geometric kernel function is an embedded computing unit that extracts the geometric structural features of the curvature change, boundary direction and texture density in the region, and is used to map the two-dimensional facial image submanifold into a structural feature vector. The generated structural feature vector is recorded as ; S33, according to the structural feature vector Initializing the regional weight parameters in the nonparametric partial differential geometric flow network model , evolution direction parameters and time scale control factors ,in: Regional weight parameter By structural feature vector Estimation of the mean geometric gradient modulus in , where the gradient modulus reflects the local curvature and deformation intensity of the region; Evolution direction parameter By structural feature vector The main direction characteristic component in is calculated, and the direction is the vector direction corresponding to the main component; Time scale control factor By facial The deformation degree of the region in the full time series image is calculated, and the standard deviation of the change of the structural characteristics of the region at different time steps is used as the estimated value of the response factor; S34, based on , and Constructing a geometric flow evolution representation ,in, Indicates that at time step Next, The geometric evolution intensity of the region is used to characterize the local morphological changes of the face during the restoration process, and the geometric flow evolution representation set Structural output as a nonparametric partial differential geometric flow network model.

4. The dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on the intelligent nursing platform according to claim 3 is characterized in that: The S5 specifically includes: S51. Based on the initial population input, in each round of iteration, the structural parameters are periodically perturbed according to the multi-frequency perturbation vector corresponding to the initial individual, so that the structural parameters jump out of the local extreme value region. At the same time, according to the curvature adjustment factor of the dimension of the structural parameter, the perturbation amplitude is adaptively adjusted, taking a preset small step size in the changing region and a preset large step size in the smooth region, dynamically controlling the update amplitude of the structural parameters, and generating new structural parameter candidate individuals; S52, inputting the new structural parameter candidate individuals into the nonparametric partial differential geometric flow network model, executing the inference process, and obtaining the geometric flow evolution results of each facial region at each time step; S53, comparing the geometric flow evolution result output by the model with the actual recovery state, calculating the fitting effect, and assigning a fitness score; S54. Score and sort all individuals in the population, select the individuals with the best fitness for the next round of retention, and replace the individuals with low fitness based on the multi-frequency perturbation mechanism to generate new parameter combinations to expand the search space; S55. Repeat the parameter update, model inference and population iteration process until the preset convergence conditions are met or the maximum number of rounds is reached. Finally, the regional weight parameters, evolution direction parameters and time scale control factors of the individuals with the best fitness are extracted to form a non-parametric partial differential geometric flow network model after structural optimization.

5. The dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on the intelligent nursing platform according to claim 4 is characterized in that: The recovery status score in step S6 is generated based on the evolution amplitude and trend stability of each region, and the abnormal prompt information is used to identify the regional location and time period of local recovery abnormalities.

6. The dynamic follow-up system for orthognathic surgery of maxillofacial deformity based on the intelligent nursing platform according to claim 5 is characterized in that: The follow-up feedback information package of step S7 includes a recovery trend chart, rehabilitation suggestion text and interactive instructions, and archiving and pushing are completed based on the follow-up feedback information package.

Images