Big data-based specific disease rehabilitation effect evaluation method and system
By constructing a QUBO model and an LSTM-ODE hybrid neural network, combined with multiple data sources, dynamic and accurate evaluation of osteoarthritis rehabilitation effects was achieved. This solves the problem of lack of dynamic closed-loop feedback and multi-objective optimization in existing technologies, and provides personalized treatment plans and reliable evaluation basis.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- NANJING MEDICAL UNIV
- Filing Date
- 2025-06-30
- Publication Date
- 2026-05-19
AI Technical Summary
Existing rehabilitation assessment methods lack dynamic closed-loop feedback and multi-objective optimization efficiency, making them unable to adapt to the complex dynamic changes in the individualized rehabilitation process of osteoarthritis patients. They also have large temporal matching errors and are difficult to generate Pareto solutions within a reasonable time frame.
By collecting gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data, and WOMAC score data from patients with osteoarthritis, the data were standardized and then input into a quantum annealing processor to construct a QUBO model. This model was then combined with an LSTM-ODE hybrid neural network to predict changes in cartilage thickness and generate an interactive rehabilitation assessment panel.
It enables dynamic and precise assessment of osteoarthritis rehabilitation effects, provides personalized treatment plans through quantum annealing optimization, and LSTM-ODE models quantify the synergistic effects of treatment and mechanical load on cartilage degeneration, providing reliable clinical decision-making basis.
Smart Images
Figure CN120766982B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical and health information technology, and in particular to a method and system for evaluating the rehabilitation effects of specific diseases based on big data. Background Technology
[0002] In recent years, big data technology has been increasingly widely applied in the field of healthcare, showing significant potential, especially in the evaluation of rehabilitation effects for chronic diseases. In existing technologies, by integrating patients' clinical scores and imaging data, machine learning models are used to predict rehabilitation trends. This technology uses traditional regression algorithms to analyze static data indicators, which can initially achieve quantitative evaluation of rehabilitation effects and reduce the subjectivity of manual evaluation to a certain extent. Some studies have attempted to introduce gait analysis sensors and joint range of motion monitoring devices, further enriching the data dimensions and providing more objective evidence for rehabilitation evaluation. Existing technologies mainly rely on data snapshots at a single time point, and the modules in the data processing flow are relatively independent, lacking the ability to perform dynamic time-series correlation analysis.
[0003] Existing methods typically separate rehabilitation program optimization from efficacy evaluation, using linear regression or simple clustering algorithms to optimize treatment programs separately, without dynamically feeding back the actual effects of the treatment programs into the evaluation model. This fragmented approach results in rehabilitation assessment lacking real-time performance and closed-loop optimization capabilities, particularly failing to adapt to the complex dynamic changes during individualized rehabilitation for osteoarthritis patients. Furthermore, the spatiotemporal alignment accuracy for multimodal data is insufficient, and the temporal matching error between gait data and imaging data may exceed clinically acceptable limits, further restricting the reliability of evaluation results. Moreover, emerging technologies such as quantum computing have not been fully utilized to address the high-dimensional nonlinear constraints in rehabilitation program combinatorial optimization, making it difficult to generate Pareto solutions within a reasonable timeframe. Summary of the Invention
[0004] In view of the aforementioned existing problems, the present invention is proposed.
[0005] Therefore, this invention provides a method for evaluating the rehabilitation effect of specific diseases based on big data, which solves the core problems of the lack of dynamic closed-loop feedback and insufficient efficiency of multi-objective optimization in existing rehabilitation assessments.
[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:
[0007] In a first aspect, the present invention provides a method for evaluating the rehabilitation effect of specific diseases based on big data, which includes collecting gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data and WOMAC score data of patients with osteoarthritis, and performing standardization processing to obtain a standardized dataset.
[0008] The standardized dataset is input into the quantum annealing processor to construct a QUBO model for optimizing rehabilitation program combinations. The Pareto solution set is obtained by solving the quantum annealing algorithm, and the rehabilitation program combination is output.
[0009] Baseline cartilage thickness data of patients is obtained through MRI imaging. Gait cycle data and joint range of motion data of osteoarthritis patients are analyzed to obtain temporal mechanical load data. The rehabilitation program combination, patient baseline cartilage thickness data and temporal mechanical load data are input into LSTM-ODE hybrid neural network to predict the cartilage thickness change curve and output a three-dimensional cartilage degeneration thermogram.
[0010] By comparing the predicted curves of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients, the rehabilitation deviation index was calculated.
[0011] By integrating standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices, an interactive rehabilitation assessment panel is generated.
[0012] As a preferred embodiment of the big data-based method for evaluating the rehabilitation effect of specific diseases described in this invention, the method includes: collecting gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data, and WOMAC score data of patients with osteoarthritis, and performing standardization processing to obtain a standardized dataset, including the following steps.
[0013] Patients' plantar pressure distribution data were collected during walking. Step length asymmetry and gait cycle time were processed to obtain gait cycle data. Knee flexion and extension angles were recorded to obtain joint range of motion data. Serum COMP concentration data, urinary CTX-II level data, and WOMAC score data were obtained through laboratory testing and Z-score standardization to generate a standardized dataset.
[0014] As a preferred embodiment of the big data-based method for evaluating the rehabilitation effect of specific diseases described in this invention, the method includes the following steps: inputting a standardized dataset into a quantum annealing processor, constructing a QUBO model for optimizing rehabilitation program combinations, solving for the Pareto solution set using the quantum annealing algorithm, and outputting the rehabilitation program combination.
[0015] Efficacy prediction features are extracted from the standardized dataset, and the efficacy prediction features are normalized to obtain a structured feature vector.
[0016] The efficacy and cost are calculated based on structured feature vectors, generating a bi-objective optimization function;
[0017] Construct a conflict matrix based on clinical contraindication rules;
[0018] By merging the bi-objective optimization function and the conflict matrix, a QUBO model for optimizing the combination of rehabilitation programs is obtained.
[0019] Based on the efficacy, cost, and conflict matrix, a sparse QUBO matrix is constructed.
[0020] The QUBO matrix is processed using the LeapHybridSampler solver to obtain the original solution set;
[0021] Perform Pareto rank classification on the original solution set and output the combination of rehabilitation plans.
[0022] As a preferred embodiment of the big data-based method for evaluating the rehabilitation effect of specific diseases described in this invention, the method includes: acquiring baseline cartilage thickness data of patients through MRI imaging, analyzing gait cycle data and joint range of motion data of osteoarthritis patients to obtain temporal mechanical load data, comprising the following steps.
[0023] The patient's knee joint 3D-SPGR sequence image was acquired using a 3T MRI scanner, the raw image data was output, and the raw image data was subjected to isotropic resampling and N4 bias field correction to obtain standardized images.
[0024] Based on standardized images, the cartilage region is segmented to calculate the average baseline thickness, thus obtaining the patient's baseline cartilage thickness data.
[0025] Fourier transform and principal component analysis were performed on gait cycle data and joint range of motion data, and time-series mechanical load data were generated after dimensionality reduction.
[0026] As a preferred embodiment of the big data-based method for evaluating the rehabilitation effect of specific diseases described in this invention, the method includes the following steps: inputting a combination of rehabilitation programs, baseline cartilage thickness data of the patient, and temporal mechanical load data into an LSTM-ODE hybrid neural network to predict the cartilage thickness change curve and output a three-dimensional cartilage degeneration thermogram.
[0027] The rehabilitation program combination embedding vector is converted into a rehabilitation program vector;
[0028] Perform 3D convolution dimensionality reduction on the temporal mechanical load data to output the temporal mechanical load data feature tensor;
[0029] The patient's baseline cartilage thickness matrix is Z-score normalized to output the standardized patient baseline cartilage thickness.
[0030] The rehabilitation program vector, the patient's baseline cartilage thickness, and the temporal mechanical load data feature tensor are input into the LSTM-ODE hybrid neural network to predict changes in cartilage thickness. The predicted cartilage thickness time series is output, and random forward propagation is performed on the predicted cartilage thickness time series to obtain the thickness time series prediction result.
[0031] The thickness time-series prediction results are denormalized and interpolated back to the original MRI spatial coordinate system to obtain a three-dimensional cartilage degeneration thermogram.
[0032] As a preferred embodiment of the big data-based method for evaluating the rehabilitation effect of specific diseases described in this invention, the following steps are included: comparing the predicted curve of a three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients to calculate a rehabilitation deviation index.
[0033] By re-examining MRI images, the measured cartilage thickness data was obtained and spatiotemporally registered and aligned with the baseline cartilage thickness data to obtain the actual follow-up data of the patients.
[0034] The timestamps of the 3D cartilage degeneration thermograms are aligned with the examination dates of the patients' actual follow-up data through linear interpolation, and time-synchronized prediction data pairs are output.
[0035] The SyN algorithm of ANTs is used to register the actual follow-up data of patients to the baseline MRI space and output a space-aligned cartilage thickness change matrix.
[0036] Extract the predicted thickness sequence and actual thickness sequence of the same anatomical region from the spatially aligned cartilage thickness variation matrix, and output the time-thickness curve;
[0037] The minimum cumulative distance between the predicted thickness sequence and the actual thickness sequence is obtained by using a dynamic time warping algorithm. The minimum cumulative distance is then normalized into the predicted thickness to obtain the recovery deviation index.
[0038] As a preferred embodiment of the big data-based method for evaluating the rehabilitation effect of specific diseases described in this invention, the method integrates standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices to generate an interactive rehabilitation assessment panel, including the following steps:
[0039] The standardized dataset, Pareto solution set, three-dimensional cartilage degeneration heat map, and rehabilitation deviation index were standardized and unified to obtain a unified standardized dataset.
[0040] A unified and standardized dataset is integrated using a web visualization engine based on React, Three.js, and D3.js to generate an interactive rehabilitation assessment panel.
[0041] Secondly, the present invention provides a specific disease rehabilitation effect evaluation system based on big data, including a data collection module that collects gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data and WOMAC score data of osteoarthritis patients, and performs standardization processing to obtain a standardized dataset.
[0042] The quantum optimization module inputs a standardized dataset into the quantum annealing processor, constructs a QUBO model for optimizing rehabilitation program combinations, obtains the Pareto solution set through the quantum annealing algorithm, and outputs the rehabilitation program combinations.
[0043] The cartilage prediction module obtains the patient's baseline cartilage thickness data through MRI imaging, analyzes the gait cycle data and joint range of motion data of osteoarthritis patients to obtain temporal mechanical load data, inputs the rehabilitation program combination, the patient's baseline cartilage thickness data and temporal mechanical load data into the LSTM-ODE hybrid neural network, predicts the cartilage thickness change curve, and outputs a three-dimensional cartilage degeneration thermogram.
[0044] The comparison module compares the predicted curve of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients to calculate the rehabilitation deviation index.
[0045] The integration module integrates standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices to generate an interactive rehabilitation assessment panel.
[0046] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the method for evaluating the rehabilitation effect of a specific disease based on big data as described in the first aspect of the present invention.
[0047] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the method for evaluating the rehabilitation effect of a specific disease based on big data as described in the first aspect of the present invention.
[0048] The beneficial effects of this invention are as follows: Through quantum annealing rehabilitation program optimization and LSTM-ODE cartilage degeneration prediction, dynamic and accurate evaluation of osteoarthritis rehabilitation effects is achieved. Quantum annealing optimization constructs a QUBO model and utilizes the parallelism of quantum computing to efficiently solve the Pareto optimal solution set. The dynamic weight adjustment mechanism and hard conflict avoidance ensure the safety and personalization of the treatment plan. The LSTM-ODE prediction model couples rehabilitation program, baseline cartilage thickness, and temporal mechanical load data to quantify the synergistic effect of treatment and mechanical load on cartilage degeneration in the form of differential equations. Its cross-scale coupling and confidence interval quantification provide a reliable basis for clinical decision-making. Attached Figure Description
[0049] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0050] Figure 1 This is a flowchart of a method for evaluating the rehabilitation effects of specific diseases based on big data.
[0051] Figure 2 This is a schematic diagram of a big data-based system for evaluating the rehabilitation effects of specific diseases.
[0052] Figure 3 This is a schematic diagram of a standardized dataset.
[0053] Figure 4 This is a schematic diagram of a rehabilitation combination program. Detailed Implementation
[0054] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.
[0055] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.
[0056] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.
[0057] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for evaluating the rehabilitation effect of a specific disease based on big data, including the following steps:
[0058] S1. Collect gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data, and WOMAC score data of patients with osteoarthritis, and perform standardization processing to obtain a standardized dataset.
[0059] S1.1 Collect plantar pressure distribution data of patients while walking, process the plantar pressure distribution data for stride asymmetry and gait cycle time to obtain gait cycle data, record the knee flexion and extension angles to obtain joint range of motion data, and generate a standardized dataset by using laboratory-detected serum COMP concentration data, urinary CTX-II level data and WOMAC score data after Z-score standardization.
[0060] S2. Input the standardized dataset into the quantum annealing processor, construct the QUBO model for optimizing the combination of rehabilitation programs, solve the Pareto solution set through the quantum annealing algorithm, and output the combination of rehabilitation programs.
[0061] S2.1 Extract efficacy prediction features from the standardized dataset, normalize the efficacy prediction features, and obtain a structured feature vector.
[0062] Furthermore, efficacy prediction features are extracted from the standardized dataset, including serum COMP concentration data, urinary CTX-II level data, and pain score, stiffness score, and functional score from the WOMAC score data. The extracted efficacy prediction features are processed using the Min-Max normalization method, which linearly transforms each feature value to the interval between 0 and 1. The normalized serum COMP concentration data, urinary CTX-II level data, WOMAC pain score data, WOMAC stiffness score data, and WOMAC functional score data are then concatenated in a fixed order to generate a structured feature vector.
[0063] S2.2 Calculate efficacy and cost based on structured feature vectors, and generate a bi-objective optimization function.
[0064] Specifically, the expression is,
[0065]
[0066] Where E represents the therapeutic effect, e i Let x be the efficacy score of regimen i, N be the total sample size of regimen i, and x be the value of x. i Let i be the binary decision variable for scheme i, where i is the scheme index.
[0067] Specifically, the expression is,
[0068]
[0069] Where C is the cost, c i Let be the average monthly cost of scheme i;
[0070] Furthermore, using serum COMP concentration data, urinary CTX-II level data, WOMAC pain score data, WOMAC stiffness score data, and WOMAC functional score data from the structured feature vector, the efficacy was calculated through linear weighted summation. The weight coefficients for serum COMP concentration data and urinary CTX-II level data were 0.3, and the weight coefficients for WOMAC pain score data, WOMAC stiffness score data, and WOMAC functional score data were 0.2. The monthly average cost data corresponding to each treatment plan was retrieved from the treatment plan database. A bi-objective optimization function was constructed, using the efficacy score and monthly average cost as two optimization objectives.
[0071] S2.3 Construct a conflict matrix based on clinical contraindication rules.
[0072] Furthermore, a drug interaction contraindication table was extracted from clinical guidelines, including contraindications for the combined use of nonsteroidal anti-inflammatory drugs and anticoagulants, and contraindications for the combined use of glucocorticoids and immunosuppressants; each rehabilitation treatment regimen was coded as a unique identifier; a D×D matrix was established, where D is the total number of treatment regimens; in the conflict matrix, the position corresponding to the treatment regimen combination with clinical contraindications was assigned a value of 10, and the position corresponding to the combination without contraindications was assigned a value of 0.
[0073] S2.4. Combine the bi-objective optimization function and the conflict matrix to obtain the QUBO model for optimizing the combination of rehabilitation programs;
[0074] Furthermore, the efficacy maximization objective function is transformed into a minimization problem form and multiplied by a weight coefficient of -0.7; the cost minimization objective function is multiplied by a weight coefficient of 0.3; the taboo penalty term in the conflict matrix is multiplied by a penalty coefficient of 10; the transformed efficacy objective function, cost objective function, and conflict matrix penalty term are added together to construct the complete QUBO model expression.
[0075] S2.5. Construct a sparse QUBO matrix based on efficacy, cost, and conflict matrix;
[0076] Furthermore, the linear terms in the efficacy objective function are converted into diagonal matrix elements, where each diagonal element corresponds to the efficacy score of a treatment plan multiplied by a weighting factor of -0.7; the linear terms in the cost objective function are converted into diagonal matrix elements, where each diagonal element corresponds to the average monthly cost of a treatment plan multiplied by a weighting factor of 0.3; the non-zero elements in the conflict matrix are multiplied by a penalty factor of 10 and used as the off-diagonal elements of the QUBO matrix; the diagonal matrix elements of the efficacy objective function, the diagonal matrix elements of the cost objective function, and the off-diagonal elements of the conflict matrix are added together to form the complete QUBO matrix.
[0077] S2.6. Use the LeapHybridSampler solver to process the QUBO matrix and obtain the original solution set.
[0078] Furthermore, the sparse QUBO matrix is input into the LeapHybridSampler solver of the D-Wave quantum computing platform; the solution parameters are set to include an annealing time of 200 microseconds, a chain strength of 2.0, and 1000 sampling times; the quantum annealing calculation process is executed to solve the QUBO problem on the CPU-QPU hybrid architecture; and the original solution set is output.
[0079] S2.7. Perform Pareto rank division on the original solution set and output the combination of rehabilitation plans.
[0080] Furthermore, the treatment plan selection variables, efficacy scores, and cost data for each solution group are extracted from the original solution set. The solution set is then processed hierarchically using the Non-Dominated Ranking Algorithm (NSGA-II), dividing the solution set into Pareto ranks (Rank 1 being the optimal non-dominated solution set). Solutions that violate clinical contraindications (combinations marked as contraindicated in the conflict matrix) are removed. The solutions at the Pareto front (Rank 1) are sorted according to the efficacy-cost ratio, and the sorted rehabilitation plan combinations are output.
[0081] S3. Baseline cartilage thickness data of patients are obtained through MRI imaging. Gait cycle data and joint range of motion data of osteoarthritis patients are analyzed to obtain time-series mechanical load data.
[0082] S3.1. Use a 3T MRI scanner to acquire 3D-SPGR sequence images of the patient's knee joint, output the raw image data, and perform isotropic resampling and N4 bias field correction on the raw image data to obtain standardized images.
[0083] Furthermore, a 3T MRI scanner was used to acquire 3D-SPGR sequence images of the patient's knee joint. Scanning parameters included a slice thickness of 1 mm, TR / TE = 20 / 5 ms, and a flip angle of 15 degrees to obtain raw DICOM format image data. The SimpleITK library was used to perform isotropic resampling of the raw image data, and the voxel size was uniformly adjusted to 1×1×1 mm. 3 The N4 bias field correction algorithm from the ANTs toolkit was used to eliminate magnetic field inhomogeneity artifacts in MRI images. The number of iterations was set to 100, and the convergence threshold was set to 0.001. The processed image data was converted to NIfTI format for storage to obtain standardized image data.
[0084] S3.2. Based on standardized images, segment the cartilage region to calculate the average baseline thickness and obtain the patient's baseline cartilage thickness data.
[0085]
[0086] Where B is the average baseline thickness, M is the total number of pixels in the cartilage region, T(x,y) is the two-dimensional thickness map, x is the coronal coordinate of the MRI image, and y is the sagittal coordinate of the MRI image.
[0087] Furthermore, a pre-trained 3D U-Net model is used to segment the cartilage region from the standardized image data, with an input of 1×1×1mm. 3 The NIfTI format image of voxels is output as a three-class probability map containing cartilage, bone, and background. The probability map is thresholded to generate a binary cartilage mask. The thickness value of each coronal-sagittal position (x, y) in the cartilage mask is calculated. The thickness matrix T(x, y) is obtained by measuring the number of continuous voxels of the cartilage mask along the axial direction (z-axis). The arithmetic mean of all valid thickness values is taken.
[0088] S3.3 Perform Fourier transform and principal component analysis on gait cycle data and joint range of motion data, and generate time-series mechanical load data after dimensionality reduction.
[0089] Furthermore, Fast Fourier Transform (FFT) was performed on the plantar pressure time-series signal in the gait cycle data and the flexion-extension angle time-series signal in the joint range of motion data, respectively. The top 10 frequency components with the largest amplitudes were extracted as frequency domain features. The frequency domain features of the gait cycle data and the frequency domain features of the joint range of motion data were aligned by timestamp and then concatenated to form an initial high-dimensional feature matrix. Principal component analysis was performed on the initial high-dimensional feature matrix, and the top 3 principal components with a cumulative variance contribution rate of more than 95% were retained. The principal component coefficient matrix was multiplied by the original feature matrix to obtain the dimensionality-reduced time-series mechanical load data.
[0090] S4. Input the rehabilitation program combination, patient baseline cartilage thickness data and temporal mechanical load data into the LSTM-ODE hybrid neural network to predict the cartilage thickness change curve and output a three-dimensional cartilage degeneration thermogram.
[0091] S4.1. Embed the rehabilitation program combination vector and convert it into a rehabilitation program vector.
[0092] Furthermore, the embedding vector of each treatment plan included in the rehabilitation plan combination is queried from the embedding matrix of the rehabilitation plan group. The mean pooling operation is then performed on the embedding vectors of all treatment plans in the rehabilitation plan combination to obtain the rehabilitation plan vector.
[0093] S4.2 Perform 3D convolution dimensionality reduction on the temporal mechanical load data and output the temporal mechanical load data feature tensor.
[0094] Furthermore, the temporal mechanical load data is resampled into a four-dimensional tensor, and this tensor is processed using a 3D convolutional neural network. A convolution operation with a stride of 2 is performed using a convolution kernel, compressing the number of channels from 3 to 16. The dimensionality is further reduced by using the ReLU activation function and a max pooling layer to obtain the feature tensor of the temporal mechanical load data.
[0095] S4.3. Normalize the patient's baseline cartilage thickness matrix using Z-score and output the standardized patient baseline cartilage thickness.
[0096] Furthermore, the thickness matrix is read from the patient's baseline cartilage thickness data to obtain the mean and standard deviation of the entire matrix thickness. The thickness value at each position in the thickness matrix is then subjected to Z-score transformation. The normalized thickness matrix is then bound to the original spatial coordinate information to generate the standardized patient baseline cartilage thickness.
[0097] S4.4 Input the rehabilitation program vector, the patient's baseline cartilage thickness, and the temporal mechanical load data feature tensor into the LSTM-ODE hybrid neural network to predict cartilage thickness changes, output the predicted cartilage thickness time series, and perform random forward propagation on the predicted cartilage thickness time series to obtain the thickness time series prediction result.
[0098] Furthermore, the rehabilitation program vector and the temporal mechanical load data feature tensor are concatenated in the time dimension to form a fused feature sequence. The fused feature sequence is then input into an LSTM layer to extract temporal features and output a hidden state vector. The hidden state vector and the standardized patient baseline cartilage thickness are input into an ODE layer to obtain changes in cartilage thickness. The Runge-Kutta fourth-order method is used to numerically solve the ODE and output the cartilage thickness prediction time series. The prediction results are then subjected to 50 Monte Carlo Dropout random forward propagations to obtain the thickness time series prediction results.
[0099] S4.5. Denormalize the thickness time-series prediction results and interpolate them back to the original MRI spatial coordinate system to obtain a three-dimensional cartilage degeneration thermogram.
[0100] Furthermore, the mean predicted thickness data in the thickness time-series prediction results are read, and the actual thickness value is restored by inverse Z-score transformation. The downsampled thickness data is restored to the spatial resolution of the original MRI image by a trilinear interpolation algorithm. The thickness matrix at each time point is aligned with the spatial coordinates of the baseline MRI to obtain a three-dimensional cartilage degeneration thermogram.
[0101] S5. Compare the predicted curve of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of the patient to calculate the rehabilitation deviation index.
[0102] S5.1. The measured cartilage thickness data obtained by reviewing MRI images is spatiotemporally registered and aligned with the baseline cartilage thickness data to obtain the actual follow-up data of the patient.
[0103] Furthermore, 3D-SPGR sequence images of the knee joint were acquired using a 3T MRI scanner during patient follow-up, with parameters consistent with the baseline scan. A pre-trained 3D U-Net model was used to segment the cartilage region in the follow-up MRI, generating a binary cartilage mask. The thickness values of the follow-up cartilage mask at each spatial location were obtained, forming a follow-up thickness matrix. The follow-up MRI was registered to the baseline MRI space with parameters set to a deformation field resolution of 0.5 mm and mutual information similarity measurement. The registered follow-up thickness matrix was aligned with the baseline thickness matrix on the time axis to obtain the actual follow-up data of the patient.
[0104] S5.2 Align the timestamp of the three-dimensional cartilage degeneration thermogram with the examination date of the patient's actual follow-up data through linear interpolation, and output time-synchronized prediction data pairs.
[0105] Furthermore, the predicted thickness time series is extracted from the three-dimensional cartilage degeneration thermogram, and the actual thickness measurement value corresponding to the follow-up examination date is read from the patient's actual follow-up data. With the treatment start date as the reference point (t=0), the predicted thickness series is linearly interpolated to obtain the predicted value at the follow-up examination time point. The interpolated predicted thickness is matched with the measured thickness one by one according to the spatial coordinates, and the time-synchronized predicted data pair is output.
[0106] S5.3. Using ANTs' SyN algorithm, the actual follow-up data of patients is registered to the baseline MRI space, and the spatially aligned cartilage thickness change matrix is output.
[0107] Furthermore, follow-up MRI images and cartilage thickness distribution data are read from the actual follow-up data of patients; the antsRegistration function of the ANTs toolkit is called, the registration type is set to SyN[0.1], the similarity measure is mutual information, and the number of iterations is 100x100x50, and the follow-up MRI images are nonlinearly registered with the baseline MRI images to generate a deformation field from the follow-up space to the baseline space. The deformation field is used to perform spatial transformation on the follow-up cartilage thickness data to maintain anatomical consistency with the baseline cartilage thickness data, and a spatially aligned cartilage thickness change matrix is generated.
[0108] S5.4 Extract the predicted thickness sequence and actual thickness sequence of the same anatomical region from the spatially aligned cartilage thickness variation matrix, and output the time-thickness curve.
[0109] Furthermore, the anatomical region of the medial femoral condyle is located in the spatially aligned cartilage thickness variation matrix. The thickness sequence of this region in the predicted heatmap and the thickness sequence in the actual follow-up data are extracted. For each time point, the average thickness of all pixels in the anatomical region is used to generate a predicted thickness curve and an actual thickness curve. The time axes of the two curves are aligned uniformly. The treatment start date is used as the reference point. Data for missing time points are filled in by linear interpolation, and the time-thickness curve is output.
[0110] S5.5. The minimum cumulative distance between the predicted thickness sequence and the actual thickness sequence is obtained by using the dynamic time warping algorithm. The minimum cumulative distance is then normalized into the predicted thickness to obtain the recovery deviation index.
[0111] Specifically, the expression is,
[0112]
[0113] Among them, REI base For the recovery deviation index, d DTW The minimum cumulative distance is given by max(P(t)), which is the maximum predicted thickness sequence.
[0114] S6. Integrate standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices to generate an interactive rehabilitation assessment panel.
[0115] S6.1 Standardize and unify the standardized dataset, Pareto solution set, three-dimensional cartilage degeneration heat map, and rehabilitation deviation index to obtain a unified standardized dataset.
[0116] Furthermore, the patient characteristic data, treatment efficacy-cost data from the Pareto solution set, time-series thickness data from the 3D cartilage degeneration heatmap, and rehabilitation deviation index values from the standardized dataset are converted into a unified JSON format. A MongoDB document structure with the patient ID as the primary key is established, containing four nested sub-documents: basic information, treatment plan, prediction results, and evaluation indicators. The time-series data is uniformly formatted using ISO 8601 timestamps, and the spatial coordinate data uses the DICOM standard coordinate system. The data integrity is verified using the SHA-256 hash algorithm to ensure that no information is lost during the conversion process, resulting in a unified and standardized dataset.
[0117] S6.2. Use a web visualization engine based on React+Three.js+D3.js to integrate the unified and standardized dataset and generate an interactive rehabilitation assessment panel.
[0118] Furthermore, the system reads patient basic information, treatment plan data, 3D cartilage degeneration heatmap data, and rehabilitation deviation index data from a unified and standardized dataset. It uses the React framework to build the panel's basic architecture, dividing it into three functional areas: data dashboard, 3D view, and trend analysis. Three.js is used to load the 3D cartilage degeneration heatmap data, constructing a rotatable and scalable 3D model of the knee joint, enabling the overlay display of thickness change heatmaps. D3.js is used to draw a Pareto solution set efficacy-cost scatter plot, and click events are added to update the 3D view. A rehabilitation deviation index warning module is integrated; when the index exceeds 0.15, a red warning border is triggered in the panel, enabling multi-view linkage. Clicking on a treatment plan point in the scatter plot automatically updates the corresponding predicted effect display in the 3D view. Finally, an interactive rehabilitation assessment panel is generated.
[0119] This embodiment also provides a specific disease rehabilitation effect evaluation system based on big data, including: a data collection module that collects gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data and WOMAC score data of osteoarthritis patients, and performs standardization processing to obtain a standardized dataset;
[0120] The quantum optimization module inputs a standardized dataset into the quantum annealing processor, constructs a QUBO model for optimizing rehabilitation program combinations, obtains the Pareto solution set through the quantum annealing algorithm, and outputs the rehabilitation program combinations.
[0121] The cartilage prediction module obtains the patient's baseline cartilage thickness data through MRI imaging, analyzes the gait cycle data and joint range of motion data of osteoarthritis patients to obtain temporal mechanical load data, inputs the rehabilitation program combination, the patient's baseline cartilage thickness data and temporal mechanical load data into the LSTM-ODE hybrid neural network, predicts the cartilage thickness change curve, and outputs a three-dimensional cartilage degeneration thermogram.
[0122] The comparison module compares the predicted curve of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients to calculate the rehabilitation deviation index.
[0123] The integration module integrates standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices to generate an interactive rehabilitation assessment panel.
[0124] This embodiment also provides a computer device applicable to the evaluation method of rehabilitation effect of specific diseases based on big data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the evaluation method of rehabilitation effect of specific diseases based on big data as proposed in the above embodiment.
[0125] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.
[0126] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the method for evaluating the rehabilitation effect of a specific disease based on big data as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.
[0127] In summary, this invention achieves dynamic and accurate evaluation of osteoarthritis rehabilitation effects through quantum annealing rehabilitation program optimization and LSTM-ODE cartilage degeneration prediction. Quantum annealing optimization utilizes the parallelism of quantum computing to efficiently solve for Pareto optimal solutions by constructing a QUBO model. The dynamic weight adjustment mechanism and hard conflict avoidance ensure the safety and personalization of the treatment plan. The LSTM-ODE prediction model quantifies the synergistic effect of treatment and mechanical load on cartilage degeneration in the form of differential equations by coupling rehabilitation program, baseline cartilage thickness, and temporal mechanical load data. Its cross-scale coupling and confidence interval quantification provide a reliable basis for clinical decision-making.
[0128] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.
Claims
1. A method for evaluating the rehabilitation effect of specific diseases based on big data, characterized in that: include, Gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data, and WOMAC score data of patients with osteoarthritis were collected and standardized to obtain a standardized dataset. The standardized dataset is input into the quantum annealing processor to construct a QUBO model for optimizing rehabilitation program combinations. The Pareto solution set is obtained by solving the model using the quantum annealing algorithm, and the rehabilitation program combination is output. This includes the following steps: Efficacy prediction features are extracted from the standardized dataset, and the efficacy prediction features are normalized to obtain a structured feature vector. The efficacy and cost are calculated based on structured feature vectors, generating a bi-objective optimization function; Construct a conflict matrix based on clinical contraindication rules; By merging the bi-objective optimization function and the conflict matrix, a QUBO model for optimizing the combination of rehabilitation programs is obtained. Based on the efficacy, cost, and conflict matrix, a sparse QUBO matrix is constructed. The QUBO matrix is processed using the LeapHybridSampler solver to obtain the original solution set; Perform Pareto rank classification on the original solution set and output the combination of rehabilitation plans; Baseline cartilage thickness data was obtained from MRI imaging. Gait cycle data and joint range of motion data of osteoarthritis patients were analyzed to obtain temporal mechanical load data. Includes the following steps, The patient's knee joint 3D-SPGR sequence image was acquired using a 3T MRI scanner, the raw image data was output, and the raw image data was subjected to isotropic resampling and N4 bias field correction to obtain standardized images. Based on standardized images, the cartilage region is segmented to calculate the average baseline thickness, thus obtaining the patient's baseline cartilage thickness data. Fourier transform and principal component analysis were performed on gait cycle data and joint range of motion data, and time-series mechanical load data were generated after dimensionality reduction. The rehabilitation program combination, patient baseline cartilage thickness data, and temporal mechanical load data are input into an LSTM-ODE hybrid neural network to predict cartilage thickness change curves and output a three-dimensional cartilage degeneration thermogram. This includes the following steps: The rehabilitation program combination embedding vector is converted into a rehabilitation program vector; Perform 3D convolution dimensionality reduction on the temporal mechanical load data to output the feature tensor of the temporal mechanical load data; The patient's baseline cartilage thickness matrix is Z-score normalized to output the standardized patient baseline cartilage thickness. The rehabilitation program vector, the patient's baseline cartilage thickness, and the temporal mechanical load data feature tensor are input into the LSTM-ODE hybrid neural network to predict changes in cartilage thickness. The predicted cartilage thickness time series is output, and random forward propagation is performed on the predicted cartilage thickness time series to obtain the thickness time series prediction result. The thickness time-series prediction results were denormalized and interpolated back to the original MRI spatial coordinate system to obtain a three-dimensional cartilage degeneration thermogram. By comparing the predicted curves of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients, the rehabilitation deviation index was calculated. By integrating standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices, an interactive rehabilitation assessment panel is generated.
2. The method for evaluating the rehabilitation effect of a specific disease based on big data as described in claim 1, characterized in that: Gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data, and WOMAC score data of patients with osteoarthritis were collected and standardized to obtain a standardized dataset, including the following steps. Patients' plantar pressure distribution data were collected during walking. Step length asymmetry and gait cycle time were processed to obtain gait cycle data. Knee flexion and extension angles were recorded to obtain joint range of motion data. Serum COMP concentration data, urinary CTX-II level data, and WOMAC score data were obtained through laboratory testing and Z-score standardization to generate a standardized dataset.
3. The method for evaluating the rehabilitation effect of a specific disease based on big data as described in claim 2, characterized in that: By comparing the predicted curves of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients, the rehabilitation deviation index is calculated, including the following steps: By re-examining MRI images, the measured cartilage thickness data was obtained and spatiotemporally registered and aligned with the baseline cartilage thickness data to obtain the actual follow-up data of the patients. The timestamps of the 3D cartilage degeneration thermograms are aligned with the examination dates of the patients' actual follow-up data through linear interpolation, and time-synchronized prediction data pairs are output. The SyN algorithm of ANTs is used to register the actual follow-up data of patients to the baseline MRI space and output a space-aligned cartilage thickness change matrix. Extract the predicted thickness sequence and actual thickness sequence of the same anatomical region from the spatially aligned cartilage thickness variation matrix, and output the time-thickness curve; The minimum cumulative distance between the predicted thickness sequence and the actual thickness sequence is obtained by using a dynamic time warping algorithm. The minimum cumulative distance is then normalized into the predicted thickness to obtain the recovery deviation index.
4. The method for evaluating the rehabilitation effect of a specific disease based on big data as described in claim 3, characterized in that: By integrating standardized datasets, Pareto solutions, 3D cartilage degeneration heatmaps, and rehabilitation deviation indices, an interactive rehabilitation assessment panel is generated, including the following steps. The standardized dataset, Pareto solution set, three-dimensional cartilage degeneration heat map, and rehabilitation deviation index were standardized and unified to obtain a unified standardized dataset. A unified and standardized dataset is integrated using a web visualization engine based on React, Three.js, and D3.js to generate an interactive rehabilitation assessment panel.
5. A system for evaluating the rehabilitation effect of specific diseases based on big data, based on the method for evaluating the rehabilitation effect of specific diseases based on big data as described in any one of claims 1 to 4, characterized in that: include, The data collection module collects gait cycle data, joint range of motion data, serum COMP concentration data, urinary CTX-II level data, and WOMAC score data from patients with osteoarthritis, and performs standardization processing to obtain a standardized dataset. The quantum optimization module inputs a standardized dataset into the quantum annealing processor, constructs a QUBO model for optimizing rehabilitation program combinations, obtains the Pareto solution set through the quantum annealing algorithm, and outputs the rehabilitation program combinations. The cartilage prediction module obtains the patient's baseline cartilage thickness data through MRI imaging, analyzes the gait cycle data and joint range of motion data of osteoarthritis patients to obtain temporal mechanical load data, inputs the rehabilitation program combination, the patient's baseline cartilage thickness data and temporal mechanical load data into the LSTM-ODE hybrid neural network, predicts the cartilage thickness change curve, and outputs a three-dimensional cartilage degeneration thermogram. The comparison module compares the predicted curve of the three-dimensional cartilage degeneration thermogram with the actual follow-up data of patients to calculate the rehabilitation deviation index. The integration module integrates standardized datasets, Pareto solutions, three-dimensional cartilage degeneration heatmaps, and rehabilitation deviation indices to generate an interactive rehabilitation assessment panel.
6. 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, it implements the steps of the method for evaluating the rehabilitation effect of a specific disease based on big data as described in any one of claims 1 to 4.
7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the big data-based method for evaluating the rehabilitation effect of specific diseases as described in any one of claims 1 to 4.