Medical image-based spinal cord injury analysis method and system

Abstract

The invention discloses a medical image-based spinal cord injury analysis method and system, and belongs to the technical field of image analysis, and the method comprises the steps: obtaining MRI images of a target object at a plurality of time points, carrying out the registration and normalization processing, and constructing a time sequence image set; extracting a spinal cord and an injury area by adopting a dynamic segmentation network guided by a spinal cord structure; constructing a time sequence deformation path of the damaged area based on a form tracking algorithm; extracting high-order morphological characteristics including a curvature change rate, an edge contour jump value and a spinal cord compression rate; performing double-domain modeling on the features and neural function scores, and training a structure-function coupling model; the model is used for performing functional state prediction, risk area identification and rehabilitation suggestion output on a new patient, and the method has the advantages of being high in predictability, high in accuracy, remarkable in clinical auxiliary value and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of image analysis, and in particular to a spinal cord injury analysis method and system based on medical images. Background Art

[0002] Spinal cord injury, due to its irreversibility and complexity, remains a significant challenge in neurosurgery and rehabilitation medicine. Existing technologies primarily utilize static single MRI images for structural observation and interpretation, relying on subjective expert experience to identify lesion location and extent. These techniques lack quantitative standards and fail to reflect the biomechanical behavior of lesions as they evolve over time.

[0003] Some methods attempt to extract spinal cord morphology through image segmentation and three-dimensional reconstruction, but there are still the following limitations: First, the image data is limited to a certain time section, ignoring the deformation trajectory of the damaged area at different time points; second, it is impossible to extract the structural-functional change patterns implicit in the image, and cannot provide accurate decision-making support for clinical intervention; third, there is currently no systematic method that can integrate multi-time series image information and jointly model the dynamic structural response and functional change trends of spinal cord tissue.

[0004] Therefore, there is an urgent need for a spinal cord injury analysis method that combines time-series image analysis and structure-function coupling modeling. It can not only automatically extract the multidimensional features of the injured area, but also capture the deformation evolution behavior of the tissue during the course of the disease, further realize the prediction of the patient's neurological function recovery potential and intervention recommendations, and significantly improve the accuracy and timeliness of diagnosis. Summary of the Invention

[0005] The purpose of the present invention is to provide a spinal cord injury analysis method and system based on medical images to address the deficiencies in the background technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a method for analyzing spinal cord injury based on medical images, comprising: S100: Acquire a spinal MRI image sequence of the target object at multiple different time points , t represents the total number of time points of image acquisition, and registration and normalization preprocessing are performed to construct the time series medical image set I; S200, using the dynamic segmentation network guided by the spinal cord structure, extract the spinal cord tissue area and the suspected injury area from the image set I frame by frame to obtain the injury area sequence ; S300, constructing a temporal deformation path M(t) of the damage region sequence R based on a morphology tracking algorithm, wherein M(t) is used to describe the continuous evolution process of the damage morphology in the time dimension; S400, extracting the damaged area at each time point The high-order morphological eigenvectors of , including local curvature change rate, edge contour jump value and spinal cord compression rate; S500, collect clinical neurological function scores at corresponding time points , the morphological characteristics and functional score Perform dual-domain mapping and train to obtain the structure-function coupling model G; S600, based on model G, performs predictive analysis on the MRI image sequences of new patients and outputs the future evolution trend of their spinal cord injury, potential high-risk areas, and rehabilitation recommendations.

[0007] Preferably, the construction of the time-sequential deformation path M(t) of the damaged area includes: Perform non-rigid temporal registration on the damaged area sequences at different time points to align all areas to the coordinate reference system of the initial time point; Extracting a set of morphological characteristic parameters of the damaged area of each frame after registration, wherein the parameters include area, volume, boundary curvature, principal axis direction, center of gravity coordinates and geometric moments; The Euclidean distance or Mahalanobis distance of morphological features between adjacent time points is used for regional matching to construct a logically coherent damage evolution path; Based on the matching results, a continuous morphological change function M(t) is constructed. Polynomial fitting or spline interpolation is used to generate an evolution function f(t) that expresses the damage characteristics over time. Based on this function, the volume growth rate and boundary diffusion rate are calculated. The changing trends in different stages are classified and labeled according to the set threshold, and spatially visualized in the image in the form of an evolution trajectory diagram.

[0008] Preferably, the method for extracting the local curvature change rate, edge contour jump value and spinal cord compression rate is: Sample the boundary curve of the damaged area and obtain the equidistant point set P on the boundary. At each point Calculate its curvature , find the standard deviation of all curvature values, and use it as the curvature change rate indicator after standardization. The expression is: Where, represents the rate of change of local curvature at time t; is the mean of all curvature values; m is the total number of boundary points; Select the pixel-level contour line of the damaged area boundary and calculate the grayscale gradient between adjacent boundary points. Perform statistics; calculate the ratio of local mutation points or change intensity as the jump value, the expression is: Where, is the edge contour jump value at the tth moment; Represents boundary points The grayscale value of the image; Represents boundary points The grayscale value of the image; Extract the contour of the spinal cord cross-section area; calculate the minimum diameter of the compressed area Average diameter of healthy spinal cord Comparison, expression: ; represents the spinal cord compression rate at time t.

[0009] Preferably, the construction of the structure-function coupling model G includes: During the image sequence acquisition process, the patient's neurological function score at each time point was recorded synchronously ; The image morphological feature vector at each time point Corresponding functional scores Align on the time axis to form a training sample pair set ; Constructing mapping functions based on training samples ; Where: G represents the structure-function mapping model; represents the morphological feature vector at the tth moment; represents the corresponding functional score; ε represents the error term.

[0010] Preferably, the predictive analysis based on model G includes: Receive MRI image sequences collected from new patients at multiple time points and extract the morphological feature vectors of the damaged area at each time point ; Call the trained structure-function coupling model G, for each feature vector Perform forward reasoning to obtain function prediction values ; Based on the prediction sequence The functional evolution trend curve is fitted and combined with M(t) to evaluate the expansion speed of the damaged area and abnormal boundary changes, identify potential high-risk areas and output rehabilitation suggestions.

[0011] The present invention also provides a spinal cord injury analysis system based on medical images, comprising: Time series image preprocessing module, obtains the target object's spinal MRI image sequence at multiple different time points , t represents the total number of time points of image acquisition, and registration and normalization preprocessing are performed to construct the time series medical image set I; The structure-guided segmentation module uses a dynamic segmentation network guided by the spinal cord structure to extract the spinal cord tissue area and the suspected injury area from the image set I frame by frame to obtain the injury area sequence ; The morphological evolution modeling module constructs the temporal deformation path M(t) of the damage region sequence R based on the morphological tracking algorithm. The M(t) is used to describe the continuous evolution of the damage morphology in the time dimension. High-order morphological feature extraction module, extracting the damage area at each time point The high-order morphological eigenvectors of , including local curvature change rate, edge contour jump value and spinal cord compression rate; Coupled modeling module, collecting clinical neurological function scores at corresponding time points , the morphological characteristics and functional score Perform dual-domain mapping and train to obtain the structure-function coupling model G; The predictive analysis module, based on model G, performs predictive analysis on the MRI image sequences of new patients and outputs the future evolution trend of their spinal cord injury, potential high-risk areas and rehabilitation suggestions.

[0012] In the above technical solution, the technical effects and advantages provided by the present invention are: 1. By introducing a dynamic segmentation network guided by spinal cord structure and a multi-time series image fusion mechanism, this invention accurately extracts spinal cord injury regions and reconstructs their morphological evolution path. By combining high-order structural features such as local curvature change rate, edge jump value, and compression rate, this method enables dynamic tracking and quantitative analysis of injury morphology. Compared with traditional single-frame image observation methods, this invention can continuously display lesion evolution trends, improving visualization and quantitative understanding of the injury evolution process.

[0013] 2. This invention constructs a structure-function coupling model, mapping image features to clinical neurological function scores in a dual-domain manner. This allows for a predictive assessment of a patient's functional status and intelligently identifies potential high-risk areas in the future. This method not only significantly improves the accuracy of score predictions (with an error of less than ±1 point), but also provides personalized rehabilitation recommendations and auxiliary diagnostic reports. It boasts strong foresight, a high level of intelligence, and outstanding clinical practical value, making it suitable for a variety of medical applications, including neuroimaging analysis and intelligent rehabilitation assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments described in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0015] Figure 1 This is a mind map of the method of the present invention.

[0016] Figure 2This is a mind map of the system modules of the present invention. DETAILED DESCRIPTION

[0017] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0018] Example 1, please refer to Figure 1 As shown, the spinal cord injury analysis method based on medical images described in this embodiment includes: S100: Acquire a spinal MRI image sequence of the target object at multiple different time points , t represents the total number of time points of image acquisition, and registration and normalization preprocessing are performed to construct the time series medical image set I; S200, using the dynamic segmentation network guided by the spinal cord structure, extract the spinal cord tissue area and the suspected injury area from the image set I frame by frame to obtain the injury area sequence ; S300, constructing a temporal deformation path M(t) of the damage region sequence R based on a morphology tracking algorithm, wherein M(t) is used to describe the continuous evolution process of the damage morphology in the time dimension; S400, extracting the damaged area at each time point The high-order morphological eigenvectors of , including local curvature change rate, edge contour jump value and spinal cord compression rate; S500, collect clinical neurological function scores at corresponding time points , the morphological characteristics and functional score Perform dual-domain mapping and train to obtain the structure-function coupling model G; S600, based on model G, performs predictive analysis on the MRI image sequences of new patients and outputs the future evolution trend of their spinal cord injury, potential high-risk areas, and rehabilitation recommendations.

[0019] In an embodiment of the present invention, step S100 includes acquiring and preprocessing medical image data of a target object to construct a standardized image set for time-series lesion analysis.

[0020] First, a spinal MRI image sequence of the target subject is acquired at different time points, including but not limited to the initial visit, mid-treatment, late treatment, or reexamination time. The image sequence can be represented as , where t represents the total number of time points for image acquisition, Represents the MRI image data corresponding to the t-th time point.

[0021] The MRI images may include axial, sagittal, or coronal images, and preferably use a T2-weighted sequence to enhance the contrast of the spinal cord soft tissue and the injured area. After image acquisition is completed, the following preprocessing operations are performed on each image: Image Registration: Use rigid or non-rigid image registration algorithm to perform spatial alignment on the image sequence. As a reference template, the images at subsequent time points Mapping to a unified coordinate system eliminates spatial offsets caused by patient position, scanning parameters, or machine errors, ensuring structural consistency between time-series images.

[0022] Image normalization: Normalize the image grayscale values to ensure uniform contrast and brightness between images at different time points. Z-score normalization or histogram matching are often used to enhance the stability of subsequent feature extraction.

[0023] Denoising and Resampling: Apply isotropic filtering algorithms or convolution smoothing operations to reduce MRI noise, while standardizing the voxel size of the image to ensure consistency in spatial resolution, providing a basis for three-dimensional modeling and morphological analysis.

[0024] Constructing a temporal image set I: Image sequence after registration and normalization The temporal medical image set I is organized according to the acquisition time sequence, in which each image data is in a unified spatial framework and intensity standard and has temporal dimension comparability.

[0025] The image set I constructed through these steps provides a standardized, traceable data foundation for subsequent damage region extraction, temporal morphological modeling, and structure-function coupling analysis. This step significantly improves the accuracy and stability of temporal analysis, making it particularly suitable for dynamic monitoring of damage evolution and the development of intervention strategies.

[0026] In the present invention, step S200 is to obtain the temporal medical image set I= constructed in step S100. In the process, the spinal cord tissue area and the suspected injury area within it are automatically extracted to form a structured injury area sequence for subsequent dynamic modeling and evaluation analysis.

[0027] Specifically, the extraction process uses a dynamic segmentation network (SSD-Net) guided by spinal cord anatomical features. This network comprehensively utilizes the image's spatial context, texture gradient features, and a standard anatomical model of the spinal cord to achieve frame-by-frame segmentation of the target area in time-series images. The process includes the following steps: Network input and structure guidance: Each frame of MRI image in the image set The dynamic segmentation network is fed with a spinal cord structural template map Sref as an auxiliary feature map to enhance the model's ability to discern spinal cord tissue boundaries and injury features. The spinal cord structural template map Sref is an anatomical reference atlas constructed by registering and averaging multiple healthy volunteer MRI image samples (from public datasets or historical clinical data). Essentially, it serves as a standard morphological template map of the spinal cord region. This map provides prior structural information during network training, improving segmentation stability in the presence of blurred boundaries or signal noise.

[0028] The specific construction method is as follows: Select several groups of healthy sample MRI images; After registration, voxel-level averaging was performed to form a spinal cord morphology map in a unified coordinate system; The template image is embedded in the auxiliary input channel of the network in grayscale form to guide structure-aware feature extraction.

[0029] Multi-scale feature extraction: The segmentation network adopts a multi-scale encoder structure to extract local texture features and global spatial information in the image, while integrating long- and short-range dependencies to address the challenges of blurred spinal cord region boundaries and injury diversity.

[0030] Spinal cord region segmentation: In the first stage, the network identifies the overall spinal cord tissue region , and outputs a spinal cord mask for spatial restriction.

[0031] Damage area identification: Based on the spinal cord mask, the network further detects areas with abnormal signal intensity, morphological mutations, or texture disorder in the tissue to generate a candidate map of the damage area. and optimize the boundaries through post-processing operations (such as morphological filtering and small area elimination).

[0032] Sequence output: Repeat the above steps for each frame in image set I , extract the corresponding spinal cord region and damaged areas , and finally a set of time series structured outputs are obtained: damage area sequence .

[0033] Compared with traditional static image segmentation methods, the dynamic segmentation method has stronger temporal consistency and spatial robustness. It can automatically adapt to the changes in tissue morphology caused by the evolution of the disease in images at different time points, ensuring the continuity and accuracy of the damage area extraction results in time sequence.

[0034] Through this step, the obtained damage area sequence not only retains the spatial geometric information in the original image, but also has the time series index characteristics, laying a key data foundation for subsequent damage deformation modeling and functional coupling analysis.

[0035] In the present invention, step S300 is to process the damaged region sequence obtained in step S200. A time series analysis was performed, and a continuous evolution path M(t) of the injured area in the time dimension was constructed based on the morphological tracking algorithm to quantitatively describe the deformation behavior of spinal cord injury as the disease progressed.

[0036] The implementation process of the morphology tracking algorithm includes the following sub-steps: S301, in order to ensure that the damaged areas at different time points have a consistent spatial reference system, first the damaged area sequence Perform temporal registration processing. Use non-rigid registration algorithm based on B-spline or affine transformation to Align to initial time point coordinate system, thereby eliminating spatial offsets caused by patient position differences or scanning errors.

[0037] S302, damage area of each frame after registration Extract the morphological feature parameter set The parameters include but are not limited to: area and volume (A / V); principal axis length and direction (Orientation); boundary curvature distribution (Boundary Curvature); local shape complexity (such as Fractal Dimension or Shape Irregularity); centroid coordinates (Centroid) and geometric moments (Geometric Moments).

[0038] The above characteristics, as basic data reflecting the changes in damage morphology, will be used to construct a time series tracking model.

[0039] S303, using the Euclidean distance or Mahalanobis distance of the morphological features between adjacent time points to calculate the similarity score, and Perform one-to-one matching to ensure the logical coherence of the time series data chain. In the case of multiple regions or complex branches, the Max-Flow Matching algorithm can be introduced to establish the global optimal matching path.

[0040] S304: Based on the above matching results, a continuous change function M(t) of the damage morphology on the time axis is constructed. This path can be modeled in the following two forms: Discrete time series form: M(t)= , used to analyze the increase and decrease trends of features; Continuous change curve form: Use spline interpolation, polynomial fitting or time series regression to continuously model the changes of features over time, and generate The morphological evolution function is used to predict future trends. f(t) represents the morphological characteristic value of the damaged area at time point t, which represents the evolution trend function of a specific feature (such as volume, boundary length, compression rate, etc.) over time; is a constant term, which represents the initial characteristic value of the damage morphology at the initial time t=0 or t=1; is the linear coefficient, which represents the linear growth rate or downward trend of damage morphological characteristics over time; is the quadratic coefficient, describing the changing acceleration or curvature, and is used to reflect the nonlinear dynamics during injury aggravation or relief. In addition, the deformation path can also be spatially visualized as an "evolution trajectory diagram" in the spinal cord image to intuitively display the temporal changes of the injury boundary.

[0041] S305. Based on the deformation path M(t), the morphological change rate (such as volume growth rate, center of gravity drift speed, edge diffusion rate, etc.) is further calculated, and the change trends in different stages are classified and labeled according to the set threshold (such as "stable period", "acute expansion period", "potential deterioration period") to provide timely risk warning support for clinical decision-making.

[0042] Through the above steps, the morphological tracking algorithm not only achieves accurate tracking and continuous modeling of the spinal cord injury area in the time dimension, but also improves the quantifiable understanding of the dynamic evolution process of the disease, which is conducive to the diagnosis, prediction and formulation of individualized intervention strategies.

[0043] In the embodiment of the present invention, step S400 is to analyze the damaged area at each time point t. , extracting its high-order morphological features to quantify the geometric complexity and biomechanical response of structural changes. It mainly includes the following three key features: The local curvature change rate is extracted by sampling the boundary curve of the damaged area and obtaining the equidistant point set P on the boundary. ; Represents the i-th sampling point on the boundary of the damaged area, usually a coordinate point in two-dimensional or three-dimensional space; i∈{1,2,…,m} represents the index of the sampling point, which represents the points numbered sequentially along the boundary; m is the total number of points sampled on the boundary, that is, the boundary sampling resolution. At each point Calculate its curvature , circle fitting or differential method can be used; calculate the standard deviation of all curvature values or the difference between the maximum and minimum values; after standardization, it is used as the curvature change rate indicator, and the expression is: Where, represents the rate of change of local curvature at time t; is the mean of all curvature values; m is the total number of boundary points.

[0044] The edge contour jump value is extracted by selecting the pixel-level contour line of the damage area boundary; the grayscale gradient between adjacent boundary points Perform statistics; calculate the proportion of local mutation points or change intensity as the jump value, expression (gradient method): Where, is the edge contour jump value at the tth moment; Represents boundary points The grayscale value of the image; Represents boundary points The image gray value; m is the total number of boundary points; Spinal cord compression rate, the extraction method is: extract the outline of the spinal cord cross-section area; calculate the minimum diameter of the compressed area Average diameter of healthy spinal cord Comparison, expression: ; represents the spinal cord compression rate at time t; A value closer to 1 indicates severe compression, and a value closer to 0 indicates no compression.

[0045] By jointly extracting the above features, a high-dimensional feature vector can be formed , providing quantitative input for subsequent structure-function coupling modeling. This metric comprehensively considers boundary geometric complexity, tissue continuity, and the degree of mechanical compression, significantly outperforming existing static analysis methods based solely on area or signal intensity.

[0046] In one embodiment of the present invention, step S500 aims to perform dual-domain correlation modeling on the morphological features of spinal cord injury in the image domain and the neurological function score in the clinical domain, and construct a structure-function coupling model G to reveal the intrinsic mapping relationship between image structural changes and the degree of functional impairment.

[0047] This step specifically includes the following sub-steps: S501, obtaining an image sequence At the same time, record the patient's neurological function score at each corresponding time point t The scoring can be performed using internationally accepted standardized assessment scales, such as: ASIA score (American Spinal Injury Association): assesses motor function (0–100) and sensory function; Frankel grade: a five-level classification used to roughly grade the severity of injury; JOA score (Japanese Orthopaedic Association): used to assess cervical spinal cord function; other quantitative neurological function scoring scales. Each time point t corresponds to a score value, forming a time series score set .

[0048] S502: The morphological feature vector sequence extracted in step S400 With scoring sequence Correspond point by point on the time axis to form a training data pair set: ; Among them, each pair of samples describes the structural state and clinical functional performance at a certain time point, providing dual-domain input for subsequent modeling.

[0049] S503, based on the training data set , constructing a structure-function coupling model G to learn the mapping relationship between high-order morphological features and neural functional performance. Optional model structures include, but are not limited to: multivariate nonlinear regression models; support vector regression (SVR); multilayer perceptron (MLP) or other neural network structures; radial basis function (RBF) networks, etc.

[0050] Mapping function form: ; Where: G represents the structure-function mapping model; represents the morphological feature vector at the tth moment; represents the corresponding functional score; ε represents the error term, which is used to describe the residual error of the model fitting.

[0051] S504. Use evaluation indicators such as cross-validation, mean square error (MSE) or mean absolute error (MAE) to tune the parameters and test the generalization ability of model G to ensure its prediction accuracy and robustness.

[0052] Through this structure-function coupling model, the following technical effects can be achieved: Multi-dimensional structural feature fusion: Unify multiple high-level image domain indicators (such as curvature, jump, compression, etc.) to functional performance to avoid misleading single indicators; Personalized functional prediction: can be used to assess the patient's current condition and future functional evolution trend; Auxiliary treatment evaluation: assists doctors in judging the effectiveness of interventions or the room for improvement in prognosis.

[0053] The coupled modeling process is significantly different from the traditional judgment method that relies solely on the doctor's subjective experience, and provides an intelligent solution for accurate assessment and decision support of spinal cord injuries.

[0054] In an embodiment of the present invention, step S600 calls the aforementioned trained structure-function coupling model G for the MRI image sequence input by a new patient to realize predictive analysis of the future evolution trend of spinal cord injury, identify potential high-risk areas and output personalized rehabilitation intervention suggestions, providing clinical decision-making support.

[0055] The steps include: S601: Receive a new patient's spinal MRI image sequence acquired at different time points , proceed as follows according to steps S100 to S400: Image registration and normalization; dynamic segmentation network to extract damaged areas ; Calculate the corresponding morphological feature vector .

[0056] This process ensures that the feature extraction method is consistent with the training model and avoids input domain shift.

[0057] S602, call the trained model G, and input feature vector at each time point t Perform forward reasoning to obtain the corresponding neural function prediction value : If the input is only the initial frame image (such as t=1), the morphological evolution model M(t) can be combined to predict the future feature trend, and then input into G to achieve long-term function prediction.

[0058] S603, time series prediction results based on the output of model G , performing trend fitting and rate of change analysis to generate a spinal cord functional evolution trend curve. Simultaneously, combined with the deformation path in the structural model M(t), the expansion direction, rate, and degree of morphological abnormality of the injured area are evaluated, identifying high-risk areas (Rrisk) that may worsen in the future.

[0059] S604: Based on the prediction results and risk assessment, the following personalized rehabilitation advice information is output: Functional level warning: If the predicted score drops beyond the set threshold, it will be automatically marked as a risk of functional deterioration; Key time node prompts: predict the time window for injury aggravation and provide timing guidance for clinical intervention; Target area annotation: Highlight potential high-risk areas in the image to assist image navigation; Individualized rehabilitation program recommendations: Based on the degree of structural risk and functional prediction, intensive treatment or regular review cycles are recommended.

[0060] The above suggestions can be presented in the form of visual reports, including trend curves, image heat maps and score change tables, to facilitate doctors' quick understanding and decision-making.

[0061] In summary, this step uses model G to achieve individualized structure-function prediction and risk warning for new patients, which has the following innovations and practicality: Leverage historical structure-function coupling patterns for generalized predictions to improve diagnostic foresight; Combined with the morphological evolution trajectory, the future structural anomaly trend can be quantified and visualized; Provide clinicians with quantitative and actionable rehabilitation intervention recommendations, significantly improving diagnosis and treatment efficiency.

[0062] The system constitutes an end-to-end intelligent analysis framework that automatically completes the continuous assessment loop of spinal cord injury from image acquisition to function prediction to recommendation output.

[0063] Example 2: To verify the effectiveness of the spinal cord injury analysis method of the present invention, real MRI case data from a hospital was selected and the experimental process was constructed as follows: Ten patients diagnosed with spinal cord injury were selected. T2-weighted MRI images of the spine were acquired before treatment, during treatment, and during follow-up. Each patient underwent three time points, resulting in a total of 30 image sets. The images were 512 × 512 pixels, with a slice thickness of 5 mm and a 1 mm interval. The JOA and ASIA scores were also recorded for each patient at the three time points as a reference for functional status.

[0064] Registration and normalization: B-spline registration was performed using the Elastix tool to align the T2 and T3 time point images to the T1 image coordinate system; the Z-score was used to normalize the image grayscale values.

[0065] Dynamic segmentation: Based on an improved U-Net (introducing an attention mechanism and multi-scale residual blocks), the spinal cord and injury area were segmented for each set of images, with an average Dice coefficient of 0.912.

[0066] Feature extraction: The following features are extracted from the damaged area of each frame: Regional volume change rate (unit: mm³); Standard deviation of edge curvature change (mean approximately 0.32, anomaly threshold 0.5); Spinal cord compression rate (the highest value is 0.36, and the mild standard is <0.15).

[0067] The morphological feature vector at each time point Corresponding JOA score Pair them to form a training set, with a total of 30 groups of samples.

[0068] Model structure: A three-layer MLP neural network (input dimension = 3, hidden layer 64-32, output = 1) is used, with ReLU activation function; Training strategy: 80% training, 20% validation, Adam optimizer, learning rate 0.001, batch size 4; Loss function: mean square error (MSE); Prediction results: Mean absolute error MAE = 0.84 points, R² = 0.91.

[0069] The image of a patient who did not participate in the training (number P09) was input into the system: The system automatically extracts morphological features at three time points; The model predicted JOA scores as [8.6, 10.2, 12.1], which were very close to the actual scores [8, 10, 13]; The system marks the marginal expansion segment between T2 and T3 in the neck 5-6 region, with a risk warning of "moderate evolution"; System recommendation: Recheck within three weeks + physical decompression intervention.

[0070] Comparison with traditional methods Summary of beneficial effects: The accuracy of score prediction is significantly improved (the error is controlled within ±1 point); the system can identify the evolution trend of potential structures in advance, which has foresight and intervention value; auxiliary reports are automatically generated, saving doctors' judgment time and improving clinical efficiency.

[0071] It should be noted that the purpose of this experiment is to initially verify the feasibility of the structure-function coupling model, not to definitively define the model's performance. To ensure training stability under limited samples, the following technical measures were taken during model training: The input dimension of the MLP network used is only 3 (corresponding to three morphological features), and the network size is small, avoiding the risk of overfitting caused by high-dimensional input; An 80 / 20 training / validation split is used, and early stopping and mean square error evaluation are introduced to prevent overfitting. The extracted volume change rate, curvature change, and compression rate are all indicators with strong structural significance and clear correlation with neural function, which helps improve prediction efficiency; The mean absolute error (MAE) was 0.84 and R² was 0.91 in the validation set, and the predicted results were highly consistent with the actual scores, showing good mapping ability.

[0072] In addition, the present invention is not limited to this initial sample size, and the training set can be expanded in combination with continuous clinical data collection to further enhance the generalization ability.

[0073] Example 3, please refer to Figure 2 As shown, the spinal cord injury analysis system based on medical images described in this embodiment includes: Time series image preprocessing module, obtains the target object's spinal MRI image sequence at multiple different time points , t represents the total number of time points of image acquisition, and registration and normalization preprocessing are performed to construct the time series medical image set I; The structure-guided segmentation module uses a dynamic segmentation network guided by the spinal cord structure to extract the spinal cord tissue area and the suspected injury area from the image set I frame by frame to obtain the injury area sequence ; The morphological evolution modeling module constructs the temporal deformation path M(t) of the damage region sequence R based on the morphological tracking algorithm. The M(t) is used to describe the continuous evolution of the damage morphology in the time dimension. High-order morphological feature extraction module, extracting the damage area at each time point The high-order morphological eigenvectors of , including local curvature change rate, edge contour jump value and spinal cord compression rate; Coupled modeling module, collecting clinical neurological function scores at corresponding time points , the morphological characteristics and functional score Perform dual-domain mapping and train to obtain the structure-function coupling model G; The predictive analysis module, based on model G, performs predictive analysis on the MRI image sequences of new patients and outputs the future evolution trend of their spinal cord injury, potential high-risk areas and rehabilitation suggestions.

[0074] The above is only a specific implementation method of the present application, but the scope of protection of the present application is not limited thereto. Any technician familiar with this technical field can easily think of changes or replacements within the technical scope disclosed in this application, which should be covered by the scope of protection of the present application.

Claims

1. A method for analyzing spinal cord injury based on medical images, characterized by: include: S100: Acquire a spinal MRI image sequence of the target object at multiple different time points , t represents the total number of time points of image acquisition, and registration and normalization preprocessing are performed to construct the time series medical image set I; S200, using the dynamic segmentation network guided by the spinal cord structure, extract the spinal cord tissue area and the suspected injury area from the image set I frame by frame to obtain the injury area sequence ; S300, constructing a temporal deformation path M(t) of the damage region sequence R based on a morphology tracking algorithm, wherein M(t) is used to describe the continuous evolution process of the damage morphology in the time dimension; S400, extracting the damaged area at each time point The high-order morphological eigenvectors of , including local curvature change rate, edge contour jump value and spinal cord compression rate; S500, collect clinical neurological function scores at corresponding time points , the morphological characteristics and functional score Perform dual-domain mapping and train to obtain the structure-function coupling model G; S600, based on model G, performs predictive analysis on the MRI image sequences of new patients and outputs the future evolution trend of their spinal cord injury, potential high-risk areas, and rehabilitation recommendations.

2. The spinal cord injury analysis method based on medical images according to claim 1, characterized in that: The temporal deformation path M(t) of constructing the damage area includes: Perform non-rigid temporal registration on the damaged area sequences at different time points to align all areas to the coordinate reference system of the initial time point; Extracting a set of morphological characteristic parameters of the damaged area of each frame after registration, wherein the parameters include area, volume, boundary curvature, principal axis direction, center of gravity coordinates and geometric moments; The Euclidean distance or Mahalanobis distance of morphological features between adjacent time points is used for regional matching to construct a logically coherent damage evolution path; Based on the matching results, a continuous morphological change function M(t) is constructed. Polynomial fitting or spline interpolation is used to generate an evolution function f(t) that expresses the damage characteristics over time. Based on this function, the volume growth rate and boundary diffusion rate are calculated. The changing trends in different stages are classified and labeled according to the set threshold, and spatially visualized in the image in the form of an evolution trajectory diagram.

3. The method for analyzing spinal cord injury based on medical images according to claim 1, characterized in that: The extraction method of the local curvature change rate, edge contour jump value and spinal cord compression rate is: Sample the boundary curve of the damaged area and obtain the equidistant point set P on the boundary. At each point Calculate its curvature , find the standard deviation of all curvature values, and use it as the curvature change rate indicator after standardization. The expression is: Where, represents the rate of change of local curvature at time t; is the mean of all curvature values; m is the total number of boundary points; Select the pixel-level contour line of the damaged area boundary and calculate the grayscale gradient between adjacent boundary points. Conduct statistics; Calculate the local mutation point ratio or change intensity as the jump value, the expression is: Where, is the edge contour jump value at the tth moment; Represents boundary points Gray value of the image; Represents boundary points Gray value of the image; Extract the contour of the spinal cord cross-section area; calculate the minimum diameter of the compressed area Average diameter of healthy spinal cord Comparison, expression: ; represents the spinal cord compression rate at time t.

4. The method for analyzing spinal cord injury based on medical images according to claim 1, characterized in that: The construction of the structure-function coupling model G includes: During the image sequence acquisition process, the patient's neurological function score at each time point was recorded synchronously ; The image morphological feature vector at each time point Corresponding functional scores Align on the time axis to form a training sample pair set ; Constructing mapping functions based on training samples ; Where: G represents the structure-function mapping model; represents the morphological feature vector at the tth moment; represents the corresponding functional score; ε represents the error term.

5. The method for analyzing spinal cord injury based on medical images according to claim 1, characterized in that: The prediction analysis based on model G includes: Receive MRI image sequences collected from new patients at multiple time points and extract the morphological feature vectors of the damaged area at each time point ; Call the trained structure-function coupling model G, for each feature vector Perform forward reasoning to obtain function prediction values ; Based on the prediction sequence The functional evolution trend curve is fitted and combined with M(t) to evaluate the expansion speed of the damaged area and abnormal boundary changes, identify potential high-risk areas and output rehabilitation suggestions.

6. A medical image-based spinal cord injury analysis system, configured to implement the medical image-based spinal cord injury analysis method according to any one of claims 1 to 5, characterized in that: include: Time series image preprocessing module, obtains the target object's spinal MRI image sequence at multiple different time points , t represents the total number of time points of image acquisition, and registration and normalization preprocessing are performed to construct the time series medical image set I; The structure-guided segmentation module uses a dynamic segmentation network guided by the spinal cord structure to extract the spinal cord tissue area and the suspected injury area from the image set I frame by frame to obtain the injury area sequence ; The morphological evolution modeling module constructs the temporal deformation path M(t) of the damage region sequence R based on the morphological tracking algorithm. The M(t) is used to describe the continuous evolution process of the damage morphology in the time dimension. High-order morphological feature extraction module, extracting the damage area at each time point The high-order morphological eigenvectors of , including local curvature change rate, edge contour jump value and spinal cord compression rate; Coupled modeling module, collecting clinical neurological function scores at corresponding time points , the morphological characteristics and functional score Perform dual-domain mapping and train to obtain the structure-function coupling model G; The predictive analysis module, based on model G, performs predictive analysis on the MRI image sequences of new patients and outputs the future evolution trend of their spinal cord injury, potential high-risk areas and rehabilitation suggestions.

Images