A Dynamic Grading Prediction and Intervention System and Method for Knee Osteoarthritis Based on a Large Model

Abstract

This invention discloses a dynamic grading prediction and intervention system and method for knee osteoarthritis based on a large model, relating to the field of medical image analysis. The system includes: acquiring a patient's knee MRI image sequence and knee joint angle time-series data; processing the MRI image sequence using a pre-trained joint cavity segmentation model to obtain a quantitative value of joint fluid volume; aligning the knee joint angle time-series to the image frame acquisition time point to generate a synchronized angle sequence; calculating the fluid volume change rate and angle change sequence, determining whether the average value of the fluid volume change rate sequence is lower than a preset stability threshold, and outputting a low-risk signal if it is; otherwise, calculating the correlation coefficient between the fluid volume change rate and angle change, and outputting a moderate or severe knee osteoarthritis prediction signal based on its value; further determining the mild or moderate risk of knee osteoarthritis based on the amplitude characteristics of the fluid volume change rate sequence. This invention can identify early knee osteoarthritis lesion characteristics and improve the accuracy of grading prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of medical image analysis, in particular to a knee osteoarthritis dynamic grading prediction and intervention system and method based on a large model. BACKGROUND

[0002] As a common orthopedic disease, knee osteoarthritis is usually accompanied by gradual degradation of joint function and pain; with the acceleration of the global aging process, the incidence of knee osteoarthritis is on the rise, which has a great impact on the quality of life of patients.

[0003] Early diagnosis of knee osteoarthritis relies on capturing the dynamic changes of the joint, and the existing knee osteoarthritis grading scheme based on the method of static image only captures the structural damage under the fixed posture, and cannot reflect the dynamic distribution rule of joint fluid in the movement process; a few technologies combining joint movement monitoring only record the movement trajectory, and do not correlate the movement angle change with the imaging characterization.

[0004] In view of the above problems, the prior art needs to be improved. SUMMARY

[0005] In view of the deficiencies in the prior art, the present application provides a knee osteoarthritis dynamic grading prediction and intervention system and method based on a large model.

[0006] In order to achieve the above purpose, the technical scheme of the present application is as follows:

[0007] In the first aspect, the present application discloses a knee osteoarthritis dynamic grading prediction and intervention method based on a large model, comprising the following steps:

[0008] Obtain the knee joint MRI image sequence and the knee joint angle time series data of the patient in the standard flexion and extension movement process;

[0009] Based on the knee joint MRI image sequence, process the joint cavity segmentation model to obtain the joint fluid volume quantization value corresponding to each image frame;

[0010] Align the knee joint angle time series data to the image frame acquisition time point to generate a synchronous joint angle sequence;

[0011] Calculate the liquid volume change rate sequence of adjacent time points according to the joint fluid volume quantization value sequence, and calculate the angle change amount sequence of adjacent time points according to the synchronous joint angle sequence;

[0012] Determine whether the average value of the absolute value of the liquid volume change rate sequence is lower than a preset stable threshold: if yes, output a knee osteoarthritis low-risk prediction signal, otherwise calculate the correlation coefficient of the liquid volume change rate sequence and the angle change amount sequence;

[0013] Determine whether the absolute value of the correlation coefficient is lower than a preset correlation threshold: if yes, output a moderate knee osteoarthritis prediction signal; otherwise, further determine whether the correlation coefficient is positive: if yes, output a severe knee osteoarthritis prediction signal; otherwise, output a mild or moderate knee osteoarthritis prediction signal based on the amplitude characteristics of the fluid volume change rate sequence.

[0014] Secondly, this invention discloses a dynamic grading prediction and intervention system for knee osteoarthritis based on a large model, comprising:

[0015] The image acquisition module is used to acquire MRI image sequences of the knee joint and time-series data of the knee joint angle during the patient's standard flexion and extension movements.

[0016] The joint cavity segmentation module is used to process the knee joint MRI image sequence based on a pre-trained joint cavity segmentation model to obtain the quantitative value of the joint fluid volume corresponding to each image frame.

[0017] The data synchronization module is used to align the knee joint angle time series data to the image frame acquisition time point to generate a synchronized joint angle sequence.

[0018] The fluid volume change rate calculation module is used to calculate the fluid volume change rate sequence at adjacent time points based on the joint fluid volume quantification value sequence;

[0019] An angle change calculation module is used to calculate the angle change sequence at adjacent time points based on the synchronous joint angle sequence.

[0020] The first judgment module is used to determine whether the average value of the absolute value of the fluid volume change rate sequence is lower than a preset stability threshold: if yes, output a low-risk prediction signal for knee osteoarthritis; otherwise, calculate the correlation coefficient between the fluid volume change rate sequence and the angle change sequence.

[0021] The second judgment module is used to determine whether the absolute value of the correlation coefficient is lower than a preset correlation threshold: if yes, it outputs a moderate knee osteoarthritis prediction signal; otherwise, it further determines whether the correlation coefficient is positive. If yes, it outputs a severe knee osteoarthritis prediction signal; otherwise, it outputs a mild or moderate knee osteoarthritis prediction signal based on the amplitude characteristics of the fluid volume change rate sequence.

[0022] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0023] 1. By dynamically capturing time-series data of knee joint fluid volume changes and joint motion angles, this invention solves the problem that traditional static imaging methods cannot reflect the dynamic distribution pattern of joint fluid. By combining imaging characteristics with joint motion angle data, this invention can more accurately identify early knee osteoarthritis lesion characteristics, improve the accuracy of grading prediction, and avoid missing early lesions or misjudging grading.

[0024] 2. By analyzing the correlation coefficient between the rate of change of synovial fluid volume and the amount of change of angle, the correlation pattern between the dynamic distribution of synovial fluid and the joint movement state can be revealed; a positive correlation reflects abnormal joint capsule compliance, while a negative correlation suggests abnormal synovial inflammation activity; by combining the analysis of the amplitude of synovial fluid volume change, physiological fluctuations and pathological abnormalities can be effectively distinguished, enhancing the clinical applicability of grading prediction.

[0025] 3. By accurately identifying the dynamic coupling relationship between knee joint fluid changes and movement angles, and based on indicators such as the magnitude of fluid changes and hysteresis effects, it provides specific angle range prompts and functional abnormality warnings, offering doctors targeted treatment guidance and patients personalized rehabilitation training programs. It can intervene in the progression of knee osteoarthritis in a timely manner and slow down the process of joint degenerative diseases. Attached Figure Description

[0026] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. 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.

[0027] Figure 1 This is an overall block diagram of the method in Embodiment 1 of the present invention;

[0028] Figure 2 This is a flowchart of the method according to Embodiment 1 of the present invention;

[0029] Figure 3 This is an overall block diagram of the system in Embodiment 2 of the present invention. Detailed Implementation

[0030] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0031] Application Overview: In traditional knee osteoarthritis grading methods, static imaging analysis can only capture joint structural damage in a fixed posture, failing to reflect the correlation between dynamic distribution of synovial fluid and changes in joint angle during movement. While motion monitoring technology can record joint movement trajectories, it lacks temporal alignment and correlation analysis between angle change data and imaging representations, resulting in insufficient extraction of dynamic pathological features and affecting the accuracy of grading prediction. This problem directly leads to the system's inability to identify the dynamic coupling relationship between changes in synovial fluid volume and movement angle, increasing the risk of missing early lesion features and misgrading.

[0032] In addressing the aforementioned issues, this application first recognizes the crucial role of the temporal correlation between dynamic changes in synovial fluid volume and the angle of motion in graded prediction. Existing technologies suffer from timestamp discrepancies between magnetic resonance imaging (MRI) images and angle sensor data, leading to inaccurate synchronization between the fluid volume change rate and the angle change, directly impacting the accuracy of correlation coefficient calculations. To resolve this, this application proposes establishing a precise correspondence between image frame acquisition time points and angle data, eliminating temporal misalignment caused by hardware delays through time series alignment. Furthermore, it finds that relying solely on the magnitude of fluid volume changes is insufficient to reflect abnormal joint function patterns; a dynamic coupling relationship between the direction of fluid volume change and the direction of joint motion is necessary for judgment. Analyzing the statistical correlation between the fluid volume change rate and the angle change effectively distinguishes between physiological fluctuations and pathological abnormalities, with a positive correlation reflecting abnormal joint capsule compliance and a negative correlation indicating abnormal synovial inflammation activity.

[0033] Example 1:

[0034] like Figures 1-2 As shown, the dynamic grading prediction and intervention method for knee osteoarthritis based on a large model includes the following steps:

[0035] Acquire knee MRI image sequences and knee joint angle time-series data during standard flexion and extension movements. The knee MRI image sequence refers to the continuous acquisition of knee magnetic resonance imaging data during standard flexion and extension movements, which can be achieved using time-series magnetic resonance scanning technology to dynamically capture changes in joint cavity morphology. The knee joint angle time-series data refers to the data on the changes in knee flexion and extension angles over time recorded by motion capture devices or inertial sensors, which can be achieved using optical motion capture systems or wearable sensors to quantify joint motion status.

[0036] Based on the knee joint MRI image sequence, the quantified value of the joint fluid volume corresponding to each image frame is obtained by processing the pre-trained joint cavity segmentation model. The quantified value of the joint fluid volume refers to the sum of the fluid signal intensity calculated after extracting the joint cavity region through image segmentation technology. Specifically, it can be achieved by using a pre-trained deep learning model to perform pixel-level segmentation of the magnetic resonance image, which is used to objectively reflect the dynamic changes in joint fluid volume.

[0037] The knee joint angle time series data is aligned to the image frame acquisition time points to generate a synchronized joint angle sequence. The fluid volume change rate sequence at adjacent time points is calculated based on the synovial fluid volume quantification value sequence, and the angle change sequence at adjacent time points is calculated based on the synchronized joint angle sequence. The fluid volume change rate sequence refers to the sequence of relative changes in the synovial fluid volume quantification values ​​at adjacent time points, which can be implemented using a difference calculation method to characterize the intensity of synovial fluid volume fluctuations. The angle change sequence refers to the sequence of absolute changes in the knee joint angle values ​​at adjacent time points, which can be implemented using a numerical differentiation method to quantify changes in joint motion amplitude.

[0038] Determine whether the average absolute value of the fluid volume change rate sequence is lower than the preset stability threshold: if yes, output a low-risk prediction signal for knee osteoarthritis; otherwise, calculate the correlation coefficient between the fluid volume change rate sequence and the angle change sequence.

[0039] The system determines whether the absolute value of the correlation coefficient is lower than a preset correlation threshold: if yes, it outputs a moderate prediction signal for knee osteoarthritis; otherwise, it further determines whether the correlation coefficient is positive. If yes, it outputs a severe prediction signal for knee osteoarthritis; otherwise, it outputs a mild or moderate prediction signal for knee osteoarthritis based on the amplitude characteristics of the fluid volume change rate sequence. The correlation coefficient refers to the statistical correlation between the fluid volume change rate and the angle change, specifically implemented using the Pearson correlation coefficient algorithm, to reveal the correlation pattern between the dynamic distribution of synovial fluid and the motion state. The preset stability threshold is a critical value used to determine whether the fluid volume change rate is within the physiological fluctuation range, specifically determined through statistical analysis of clinical data, used to distinguish between stable and unstable joint states. The preset correlation threshold is a critical value used to assess the significance of the fluid volume-motion correlation, specifically determined through hypothesis testing methods, used to classify different levels of pathological correlation strength.

[0040] As a preferred embodiment, the specific implementation of this application is as follows: Acquire a sequence of MRI images of the knee joint during standard flexion and extension movements, with an acquisition frequency of 10 frames / second, for a total of 300 frames. Simultaneously acquire knee joint angle time-series data, with a sampling frequency of 100Hz. Use a pre-trained U-Net model as the joint cavity segmentation model to process each frame of MRI image to obtain a sequence of quantified joint fluid volume values. Align the knee joint angle time-series data to the image frame acquisition time points using an interpolation method to generate a synchronized joint angle sequence. Calculate the fluid volume change rate sequence and angle change sequence at adjacent time points. Set a preset stability threshold of 0.05 and determine whether the average absolute value of the fluid volume change rate sequence is lower than this threshold. If it is lower than the threshold, output a low-risk prediction signal for knee osteoarthritis. Otherwise, calculate the Pearson correlation coefficient between the fluid volume change rate sequence and the angle change sequence. Set a preset correlation threshold of 0.3 and determine whether the absolute value of the correlation coefficient is lower than this threshold. If it is lower than the threshold, output a moderate prediction signal for knee osteoarthritis. If it is not lower than the threshold and the correlation coefficient is positive, output a severe prediction signal for knee osteoarthritis. If the value is not lower than the threshold and the correlation coefficient is negative, the maximum value of the absolute value of the fluid volume change rate sequence is calculated as the amplitude feature, and a prediction signal for mild or moderate knee osteoarthritis is output based on this feature.

[0041] Through the aforementioned approach, this application achieves graded prediction of knee osteoarthritis based on dynamic imaging sequences and joint angle data. By analyzing the dynamic correlation between changes in synovial fluid volume and joint angle, it overcomes the limitation of static imaging analysis in failing to reflect the dynamic distribution of synovial fluid during movement. This method can capture early lesion characteristics, improve the accuracy of graded prediction, provide a basis for the development of personalized intervention plans, help to promptly grasp the early treatment window, and slow down the progression of joint degenerative diseases. Furthermore, by quantitatively assessing key pathological indicators such as synovial inflammation activity and joint capsule compliance, this method promotes the development of knee osteoarthritis diagnosis and treatment technology towards precision.

[0042] This application further proposes to perform joint cavity region segmentation on each frame of MRI image; the segmentation process is implemented using a pre-trained deep learning model, which is trained and generated based on a dataset of knee arthritis MRI images with labeled joint cavity regions; the sum of pixel intensities that conform to the characteristics of fluid signals within the segmented region is calculated as the quantification value of joint fluid volume in the current frame.

[0043] The pre-trained deep learning model was trained using a dataset of magnetic resonance imaging of knee arthritis, which included image data of annotated joint cavity regions, ensuring that the model could identify the joint cavity boundaries under different pathological conditions. Pixels within the segmented region that meet the characteristics of fluid signals were filtered by intensity thresholds of specific MRI sequences. For example, in T2-weighted imaging, fluid signals are represented by bright areas. The calculation of the sum of pixel intensities was achieved by accumulating the gray values ​​of all pixels within the segmented region that meet the characteristics of fluid signals. This value is positively correlated with the volume of joint fluid.

[0044] Specifically, the pre-trained model processes the input MRI images frame by frame, extracts spatial features of the joint cavity region using a convolutional neural network, and outputs a corresponding segmentation mask. Within the area covered by the segmentation mask, fluid signal features are filtered through a preset intensity range to exclude interference signals from surrounding soft tissue. When calculating the sum of the filtered pixel intensities, a weighted summation method is used, with weights adjusted according to MRI imaging parameters to ensure a linear relationship between the quantified value and the actual fluid volume. In this way, the dynamic changes in joint fluid volume can be accurately captured, providing a reliable data foundation for subsequent calculations of the fluid volume change rate.

[0045] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0046] First, joint cavity region segmentation is performed on each frame of MRI image. The segmentation process is implemented using a pre-trained deep learning model. This deep learning model is trained on a dataset of knee osteoarthritis MRI images with labeled joint cavity regions. Specifically, segmentation network architectures such as U-Net can be used, trained on a large amount of labeled data, enabling the model to accurately identify and segment the joint cavity regions in MRI images.

[0047] Next, the sum of pixel intensities within the segmented region that conform to the characteristics of a fluid signal is calculated as the quantified value of the synovial fluid volume in the current frame. Since synovial fluid exhibits high signal characteristics in MRI T2-weighted sequences, a threshold can be set, and pixels exceeding this threshold can be considered as fluid signals. The intensity values ​​of these pixels are then summed to obtain a quantified index reflecting the amount of synovial fluid.

[0048] Through the above technical solution, this application achieves automated and quantitative analysis of synovial fluid volume in knee MRI images. The pre-trained deep learning model improves the accuracy and efficiency of joint cavity segmentation, avoiding the subjectivity and tediousness of manual segmentation. The quantitative value of synovial fluid volume calculated based on the segmentation results provides a reliable data foundation for subsequent dynamic analysis, helping to more accurately assess the severity and progression of knee osteoarthritis.

[0049] This application further proposes that when the correlation coefficient is negative, it indicates that the fluid volume decreases when the joint is extended or increases when the joint is flexed; it determines whether the amplitude feature is lower than the preset moderate fluctuation threshold, and if so, outputs a mild prediction signal for knee arthritis; otherwise, it outputs a moderate prediction signal for knee arthritis; the amplitude feature of the fluid volume change rate sequence is the maximum value of the absolute value of the fluid volume change rate.

[0050] The amplitude characteristic is achieved by calculating the largest absolute value in the fluid volume change rate sequence, which reflects the extreme fluctuations in fluid volume dynamics. A preset moderate fluctuation threshold is set based on clinical data statistics, for example, by analyzing the correspondence between fluid volume change rates and pathological grades in historical cases to determine the threshold range. When comparing the maximum absolute value of the fluid volume change rate sequence with the preset threshold, a direct numerical comparison is used. When the maximum value exceeds the threshold, the fluid volume fluctuation amplitude is determined to have reached a moderately abnormal level.

[0051] Specifically, when the correlation coefficient is negative, the system first extracts the value with the largest absolute value in the fluid volume change rate sequence as the key judgment criterion. This value is compared with a preset moderate fluctuation threshold. For example, when the threshold is 0.15 / s, if the maximum fluid volume change rate is 0.12 / s, a mild prediction signal is output; if the maximum fluid volume change rate is 0.18 / s, a moderate prediction signal is output. This process avoids misjudgment caused by relying solely on the sign of the correlation coefficient by quantifying the extreme fluctuation amplitude of fluid volume changes. Furthermore, the preset threshold can be adjusted according to different patient groups; for example, a lower threshold can be used for elderly patients to improve sensitivity. By combining the dual judgment of the direction and amplitude of fluid volume change, the degree of abnormal joint fluid dynamics can be more accurately distinguished, thereby improving the clinical applicability of the graded prediction.

[0052] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0053] When the correlation coefficient is negative, it indicates a decrease in fluid volume during joint extension or an increase in fluid volume during joint flexion. The system determines whether the amplitude characteristic is below a preset moderate fluctuation threshold; if so, it outputs a mild prediction signal for knee osteoarthritis; otherwise, it outputs a moderate prediction signal for knee osteoarthritis.

[0054] The amplitude feature of the liquid volume change rate sequence is the maximum value of the absolute value of the liquid volume change rate. Specifically, the absolute value of each element in the liquid volume change rate sequence is first calculated, and then the maximum value is selected as the amplitude feature. For example, for the liquid volume change rate sequence [-0.02, 0.03, -0.01, 0.04, -0.03], the calculated absolute value sequence is [0.02, 0.03, 0.01, 0.04, 0.03], where the maximum value of 0.04 is the amplitude feature.

[0055] Furthermore, the preset moderate fluctuation threshold can be set to 0.05. When the amplitude feature is below 0.05, a mild knee osteoarthritis prediction signal is output; when the amplitude feature is greater than or equal to 0.05, a moderate knee osteoarthritis prediction signal is output.

[0056] Therefore, by analyzing the amplitude characteristics of the fluid volume change rate sequence, mild and moderate knee osteoarthritis can be more accurately distinguished, thus improving the accuracy of grading prediction.

[0057] Through the above technical solution, this application can further distinguish between mild and moderate knee osteoarthritis when the correlation coefficient is negative, based on the amplitude characteristics of the fluid volume change rate sequence. This method considers the amplitude of joint fluid volume changes, improving the accuracy of grading prediction. Furthermore, because a preset moderate fluctuation threshold is used as the judgment criterion, this method has good adaptability and adjustability, allowing for threshold adjustments based on different patient groups or clinical needs, thereby achieving more personalized and accurate knee osteoarthritis grading prediction.

[0058] This application further proposes to use the Pearson correlation coefficient for the correlation coefficient, the calculation process of which includes: standardizing the liquid volume change rate sequence and the angle change sequence; calculating the sum of the products of corresponding elements of the two sequences; and normalizing the sum of the products according to the sequence length.

[0059] The standardization process eliminates dimensional differences by subtracting the mean from the liquid volume change rate sequence and dividing by the standard deviation from the angle change sequence; the sum of the products reflects the strength and direction of the linear relationship between the two sequences; the normalization process obtains the standardized correlation coefficient value by dividing the sum of the products by the sequence length minus one.

[0060] Specifically, standardization ensures that the rate of change in liquid volume and the change in angle are within the same dimension, avoiding calculation biases caused by unit differences. When calculating the sum of the products of corresponding elements in two sequences, point-by-point multiplication and summation are used to capture the coordinated trend of changes in liquid volume and angle. Normalization further eliminates the influence of sequence length on the calculation results, limiting the correlation coefficient to between -1 and 1, facilitating a direct assessment of the correlation strength. For example, when the rate of change in liquid volume and the change in angle are perfectly positively correlated, the product sum reaches its maximum value, and the normalized correlation coefficient is 1; when they are perfectly negatively correlated, the correlation coefficient is -1. Thus, through the step-by-step processing of standardization, product sum calculation, and normalization, the degree of correlation between dynamic changes in liquid volume and joint motion angle can be accurately quantified, providing a reliable basis for graded prediction.

[0061] This application further proposes a scheme to calculate an abnormal correlation strength index based on the correlation coefficient and the joint fluid volume change rate sequence, and to provide an early warning correction for the predicted signal when the index exceeds a preset pathological threshold. Specifically, the abnormal correlation strength index is obtained by multiplying the absolute value of the correlation coefficient by the logarithmic enhancement of the arithmetic mean of the absolute values ​​of the joint fluid volume change rates. When the abnormal correlation strength index exceeds the preset pathological threshold, if the original predicted signal is mild, a progression warning is added; if it is moderate, an activity warning is added.

[0062] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0063] Based on the correlation coefficient and the sequence of changes in synovial fluid volume, an abnormal correlation strength index is calculated through nonlinear combination. When the abnormal correlation strength index exceeds a preset pathological threshold, the predicted signal is corrected with an early warning. If the original predicted signal is mild, it is corrected to include a progression warning. If the original predicted signal is moderate, it is corrected to include an activity warning.

[0064] The abnormal association strength index is calculated as follows: The arithmetic mean of the absolute values ​​of the synovial fluid volume change rate sequence is calculated. The absolute value of the correlation coefficient is then multiplied by the logarithmic enhancement of this arithmetic mean to generate the abnormal association strength index. The calculation formula is as follows:

[0065]

[0066] in, The correlation coefficient is... This is the arithmetic mean of the absolute values ​​of the sequence of synovial fluid volume change rates. This is an index of abnormal correlation strength.

[0067] For example, the correlation coefficient calculated in a specific implementation. The arithmetic mean of the absolute values ​​of the series of synovial fluid volume change rates is 0.75. The value is 0.05, which is then substituted into the formula:

[0068] AAI=0.75×ln(1+0.05)=0.75×0.0488=0.0366

[0069] Assuming the preset pathological threshold is 0.03, since the calculated AAI value of 0.0366 is greater than the pathological threshold of 0.03, the predicted signal needs to be corrected for early warning. If the original predicted signal is mild, it is corrected to include a progression warning; if the original predicted signal is moderate, it is corrected to include an activity warning.

[0070] Through the aforementioned technical solution, this application can calculate a comprehensive index reflecting the severity of knee osteoarthritis based on the correlation between changes in synovial fluid volume and joint angle, combined with the magnitude of changes in synovial fluid volume. This method considers not only the degree of correlation between fluid volume and angle changes but also the absolute magnitude of fluid volume changes, thus providing a more comprehensive assessment. By introducing a logarithmic function, the scheme can maintain sensitivity while avoiding the excessive influence of abnormally large fluid volume changes on the results. Furthermore, by setting pathological thresholds and implementing early warning corrections, the scheme can promptly identify potential disease progression risks, providing more accurate reference information for clinical decision-making. This dynamic assessment and early warning mechanism helps to detect the changing trends of knee osteoarthritis earlier and more accurately, thereby supporting more targeted treatment interventions.

[0071] This application further proposes that when outputting a moderate or moderate graded prediction signal, the joint range of motion is divided into several intervals according to the angle, the arithmetic mean of the synovial fluid volume in each interval is calculated, the target interval with the largest arithmetic mean of fluid volume is identified, and the angle range corresponding to the target interval is added to the prediction signal.

[0072] Specifically, the angle division operation divides the joint range of motion into multiple continuous and uniform angle intervals, for example, dividing the flexion and extension range from 0° to 120° into five 24° intervals; the calculation of the arithmetic mean of fluid volume is achieved by summing the fluid volume quantization values ​​of all image frames within the same angle interval and dividing by the number of frames, thus eliminating the random fluctuations of single-frame data; the identification of the target interval is accomplished by comparing the magnitude relationship of the fluid volume average values ​​of each interval, for example, using a sorting algorithm to determine the interval corresponding to the maximum value; the addition of angle range prompts appends the start and end angle values ​​of the target interval to the prediction signal output in the form of numerical intervals.

[0073] Specifically, after obtaining synchronized joint angle sequences and fluid volume quantification value sequences, the system first determines the overall range based on the maximum and minimum angles of joint movement, for example, from 0° of full extension to 120° of maximum flexion. This range is then divided into a predetermined number of sub-intervals. The fluid volume quantification values ​​within each sub-interval are extracted and their arithmetic mean is calculated. For example, within the 24°-48° interval, the system statistically analyzes the fluid volume quantification values ​​corresponding to all image frames whose angle values ​​fall within this range and calculates their average. By traversing all intervals, the target interval with the highest average fluid volume is determined; for example, the average fluid volume peaks in the 72°-96° interval. This angle range is appended to the prediction signal; for example, a label such as "fluid volume abnormality concentrated in 72°-96° flexion angle" is added after a moderate prediction signal. This process provides an angle range reference for clinical intervention by locating the specific movement phase where fluid volume abnormalities accumulate, for example, guiding patients to perform rehabilitation training within a specific angle range or avoiding overload.

[0074] Through the aforementioned technical solution, this application can further provide information on the specific angular range of abnormal joint fluid volume, building upon the predicted results of moderate or severe knee osteoarthritis. This precise localization helps physicians develop more targeted treatment plans, such as limiting joint movement or strengthening exercises within a specific angular range. Simultaneously, patients can better understand their joint condition and avoid maintaining high-risk angles for extended periods during daily activities, thereby reducing the risk of joint damage.

[0075] This application further proposes to perform a time-shift operation on the synchronized joint fluid volume quantification value sequence to generate a time-shifted fluid volume sequence, calculate the correlation coefficient between the joint angle value sequence and the time-shifted fluid volume sequence at each time-shift step size, determine the optimal hysteresis step size when the absolute value of the correlation coefficient is the largest, calculate the hysteresis effect coefficient based on the optimal hysteresis step size, magnetic resonance sampling interval and the average amplitude of the joint angle change sequence, compare the hysteresis effect coefficient with a preset response threshold and add a functional abnormality prompt to the prediction signal.

[0076] The time-shifting operation replaces the i-th element in the quantified joint fluid volume sequence with the (i+d)-th element to generate a time-shifted fluid volume sequence with a time-shifting step size of d, where d is an integer greater than or equal to zero. The correlation coefficient is calculated using the standardized Pearson correlation coefficient, and the optimal hysteresis step size corresponding to the maximum correlation is determined by traversing different time-shifting step sizes. The hysteresis effect coefficient is calculated by dividing the product of the optimal hysteresis step size and the magnetic resonance sampling interval by the arithmetic mean of the absolute values ​​of the joint angle change sequence, thus quantifying the degree of lag of fluid volume change relative to angle change.

[0077] Specifically, fluid volume sequences with different time offsets are generated through time-shifting operations, and their correlation coefficients with the angle sequence are calculated. The step size corresponding to the largest absolute value of the correlation coefficient reflects the degree to which fluid volume changes lag behind angle changes. The time lag is calculated based on the optimal hysteresis step size and sampling interval, and the mechanical anomaly of the hysteresis effect is assessed in conjunction with the average amplitude of the angle change. When the predicted signal is mild and the hysteresis effect coefficient exceeds the slow response threshold, it indicates a delayed fluid volume response caused by abnormal synovial viscosity; when the predicted signal is severe and the hysteresis effect coefficient is below the fast response threshold, it indicates an advanced fluid volume response caused by joint capsule laxity. For example, with an MRI sampling interval of 0.5 seconds, an optimal hysteresis step size of 3, and an average angle change amplitude of 5 degrees / second, the hysteresis effect coefficient is 3 × 0.5 / 5 = 0.3 seconds / degree. If the preset slow response threshold is 0.25 seconds / degree, then 0.3 seconds / degree exceeding the threshold triggers an abnormal synovial viscosity indication.

[0078] As a preferred embodiment, the solution of this application is specifically implemented as follows:

[0079] A time-shift operation is performed on the synchronized synovial fluid volume quantification value sequence to generate a time-shifted fluid volume sequence. Specifically, the i-th element in the synovial fluid volume quantification value sequence is replaced with the (i+d)-th element, where d is an integer and d≥0, to generate a time-shifted fluid volume sequence with a time-shift step size of d.

[0080] For example, for the original sequence [100,120,140,160,180], when d=2, the generated time-shifted sequence is [140,160,180,null,null].

[0081] Furthermore, the correlation coefficient between the joint angle value sequence and the time-transfer volume sequence at each time-transfer step length is calculated. Thus, a series of correlation coefficient values ​​can be obtained, such as [-0.2, 0.5, 0.8, 0.6, 0.3].

[0082] Determine the optimal hysteresis step size that maximizes the absolute value of the correlation coefficient. In the example above, the optimal hysteresis step size is 2, corresponding to a maximum absolute value of 0.8 for the correlation coefficient.

[0083] The hysteresis effect coefficient is calculated based on the optimal hysteresis step size, magnetic resonance sampling interval, and the average amplitude of the joint angle change sequence.

[0084] The hysteresis effect coefficient is compared with a preset response threshold, and a functional abnormality indication is added to the predicted signal: if the predicted signal is mild and the hysteresis effect coefficient exceeds the slow response threshold, it indicates abnormal synovial viscosity; if the predicted signal is severe and the hysteresis effect coefficient is below the fast response threshold, it indicates abnormal joint capsule laxity.

[0085] Through the aforementioned technical solution, this application can quantitatively assess the dynamic response characteristics of the knee joint by analyzing the time delay relationship between changes in synovial fluid volume and changes in joint angle. This allows for more accurate identification of the pathological state of knee osteoarthritis and provides targeted indications of functional abnormalities. This dynamic analysis method compensates for the shortcomings of static imaging assessment, providing more comprehensive information support for clinical diagnosis and treatment decisions.

[0086] This application further proposes that the hysteresis coefficient be calculated using the following formula:

[0087]

[0088] in, The arithmetic mean of the absolute values ​​of the joint angle change sequence. To achieve the optimal hysteresis step size, The sampling time interval for magnetic resonance imaging. This is the hysteresis coefficient.

[0089] The optimal hysteresis step size is obtained by iterating through the correlation coefficients under different time shift step sizes and selecting the step size corresponding to the maximum value; the magnetic resonance sampling interval is determined by the parameters of the image acquisition equipment; the arithmetic mean of the joint angle change sequence is achieved by calculating the mean of the absolute values ​​of the angle changes. The numerator in the formula represents the delay time of fluid volume change relative to joint movement, and the denominator represents the average amplitude of joint movement. The influence of differences in movement amplitude on hysteresis assessment is eliminated by using a ratio.

[0090] Specifically, when MRI images are acquired at fixed time intervals, the optimal hysteresis step size multiplied by the sampling interval can be converted into the actual delay time of fluid volume change relative to joint movement. This delay time is normalized with the average amplitude of joint movement to obtain the fluid volume response delay per unit angle change. For example, when the average joint angle change is 5 degrees / frame, if the optimal hysteresis step size is 2 steps and the sampling interval is 0.5 seconds, the calculated HEC is (2 × 0.5) / 5 = 0.2 seconds / degree. This value is compared with a preset slow response threshold or fast response threshold to determine whether the fluid volume change lags behind joint movement. When the HEC exceeds the slow response threshold, it indicates a significant delay in fluid volume change, suggesting abnormal synovial viscosity; when the HEC is below the fast response threshold, it indicates that fluid volume change is almost synchronous with movement, suggesting abnormal joint capsule laxity. By quantifying the hysteresis effect, the type of joint mechanical dysfunction can be identified more accurately.

[0091] As a preferred embodiment, the specific implementation of this application is as follows: A time-shift operation is performed on the synchronized synovial fluid volume quantification sequence to generate fluid volume sequences with time-shift step sizes of 0, 1, and 2. Each time-shifted fluid volume sequence is standardized with the original joint angle value sequence to calculate a correlation coefficient matrix. By comparing the absolute values ​​of the correlation coefficients, the optimal hysteresis step size is determined to be 2 sampling intervals. Based on the magnetic resonance imaging sampling time interval of 0.5 seconds and the arithmetic mean of the absolute values ​​of the joint angle change sequence of 8.6 degrees, the hysteresis effect coefficient is calculated to be 0.116 using the formula. This coefficient is compared with a preset slow response threshold of 0.1, and an abnormal synovial viscosity indication is added to the predicted signal.

[0092] Through the above technical solution, this application can accurately identify mechanical functional abnormalities such as synovial tissue adhesion or joint capsule laxity by quantifying the hysteresis effect of dynamic response of synovial fluid and changes in motion angle. This solves the problem that the existing technology cannot assess the hysteresis of dynamic response of synovial fluid, and provides objective quantitative indicators for clinical judgment of synovial inflammation activity and joint capsule mechanical property degeneration.

[0093] Example 2:

[0094] like Figure 3 As shown, the dynamic grading prediction and intervention system for knee osteoarthritis based on a large model includes:

[0095] The image acquisition module is used to acquire MRI image sequences of the knee joint and time-series data of the knee joint angle during the patient's standard flexion and extension movements.

[0096] The joint cavity segmentation module is used to process the knee joint MRI image sequence based on a pre-trained joint cavity segmentation model to obtain the quantitative value of the joint fluid volume corresponding to each image frame.

[0097] The data synchronization module is used to align the knee joint angle time series data to the image frame acquisition time point to generate a synchronized joint angle sequence.

[0098] The fluid volume change rate calculation module is used to calculate the fluid volume change rate sequence at adjacent time points based on the joint fluid volume quantification value sequence;

[0099] An angle change calculation module is used to calculate the angle change sequence at adjacent time points based on the synchronous joint angle sequence.

[0100] The first judgment module is used to determine whether the average value of the absolute value of the fluid volume change rate sequence is lower than a preset stability threshold: if yes, output a low-risk prediction signal for knee osteoarthritis; otherwise, calculate the correlation coefficient between the fluid volume change rate sequence and the angle change sequence.

[0101] The second judgment module is used to determine whether the absolute value of the correlation coefficient is lower than a preset correlation threshold: if yes, it outputs a moderate knee osteoarthritis prediction signal; otherwise, it further determines whether the correlation coefficient is positive. If yes, it outputs a severe knee osteoarthritis prediction signal; otherwise, it outputs a mild or moderate knee osteoarthritis prediction signal based on the amplitude characteristics of the fluid volume change rate sequence.

[0102] The above description is merely an example and illustration of the structure of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the structure of the invention or exceed the scope defined in the claims, all of which should fall within the protection scope of the present invention.

[0103] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0104] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A large model-based dynamic grading prediction method for knee osteoarthritis, characterized by, The method comprises the following steps: Obtaining knee joint MRI image sequences and knee joint angle time series data of a patient during a standard flexion-extension movement; Processing the knee joint MRI image sequences by a pre-trained joint cavity segmentation model to obtain joint fluid volume quantification values corresponding to each image frame; Aligning the knee joint angle time series data to the image frame acquisition time points to generate a synchronized joint angle sequence; Calculating a joint fluid volume change rate sequence of adjacent time points according to the joint fluid volume quantification value sequence, and calculating an angle change amount sequence of adjacent time points according to the synchronized joint angle sequence; Judging whether the average value of the absolute value of the joint fluid volume change rate sequence is lower than a preset stability threshold value: if yes, outputting a knee osteoarthritis low-risk prediction signal, otherwise calculating a correlation coefficient of the joint fluid volume change rate sequence and the angle change amount sequence; Judging whether the absolute value of the correlation coefficient is lower than a preset correlation threshold value: if yes, outputting a knee osteoarthritis moderate prediction signal, otherwise further judging whether the correlation coefficient is a positive value: if yes, outputting a knee osteoarthritis severe prediction signal, otherwise judging whether an amplitude feature of the joint fluid volume change rate sequence is lower than a preset moderate fluctuation threshold value: if yes, outputting a knee osteoarthritis mild prediction signal, otherwise outputting a knee osteoarthritis moderate prediction signal; wherein the amplitude feature is the maximum value of the absolute value of the joint fluid volume change rate sequence.

2. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 1, characterized in that: The process of obtaining the joint fluid volume quantification value comprises: Performing joint cavity region segmentation on each MRI image, and the segmentation processing is realized by using a pre-trained deep learning model, wherein the deep learning model is generated by training based on a knee osteoarthritis magnetic resonance image data set with labeled joint cavity regions; Calculating the total intensity of the pixels in the segmented region that meet the liquid signal characteristics as the joint fluid volume quantification value of the current frame.

3. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 1, characterized in that: The process of outputting a knee osteoarthritis mild or moderate prediction signal according to the amplitude feature of the joint fluid volume change rate sequence comprises: When the correlation coefficient is negative, it indicates that the joint fluid volume decreases during joint extension or increases during joint flexion; judging whether the amplitude feature is lower than a preset moderate fluctuation threshold value: if yes, outputting a knee osteoarthritis mild prediction signal, otherwise outputting a knee osteoarthritis moderate prediction signal; The amplitude feature of the joint fluid volume change rate sequence is the maximum value of the absolute value of the joint fluid volume change rate sequence.

4. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 1, characterized in that: The correlation coefficient adopts a Pearson correlation coefficient, and the calculation process comprises: Standardizing the joint fluid volume change rate sequence and the angle change amount sequence; Calculating the sum of the products of the corresponding elements of the two sequences; Normalizing the product sum according to the sequence length.

5. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 1, characterized in that: Further comprising: Calculating an abnormal correlation strength index by nonlinear combination according to the correlation coefficient and the joint fluid volume change rate sequence; When the abnormal correlation strength index exceeds a preset pathological threshold value, modifying the prediction signal: If the original prediction signal is mild, modifying it to a prediction signal containing a progression warning; If the original prediction signal is moderate, modifying it to a prediction signal containing an activity warning; The abnormal correlation strength index is calculated according to the following logic: calculating the arithmetic mean of the absolute values of the joint fluid volume change rate sequence, multiplying the absolute value of the correlation coefficient by the logarithmic enhancement value of the arithmetic mean to generate the abnormal correlation strength index, and the calculation formula is: ； wherein is the correlation coefficient, is the arithmetic mean of the absolute values of the synovial fluid volume change rate sequence, is the abnormal correlation strength index.

6. The large model-based dynamic grading prediction method for knee osteoarthritis according to claim 1, characterized in that: When a medium or moderate grading prediction signal is output, the joint motion range is equally divided into several intervals by angle, the arithmetic mean of the synovial fluid volume quantization value in each interval is calculated, the target interval with the maximum arithmetic mean of the synovial fluid volume is identified, and the angle range corresponding to the target interval is added to the prediction signal.

7. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 1, characterized in that: Further comprising: Performing time shift operation on the synchronized synovial fluid volume quantization value sequence to generate a time-shifted fluid volume sequence; Calculating the correlation coefficient of the joint angle value sequence and the time-shifted fluid volume sequence at each time shift step; Determine the optimal hysteresis step when the absolute value of the correlation coefficient is maximum; According to the optimal hysteresis step, the magnetic resonance sampling interval and the average amplitude of the joint angle change sequence, calculate the hysteresis effect coefficient; Compare the hysteresis effect coefficient with the preset response threshold, and add the functional abnormality prompt to the prediction signal: If the prediction signal is mild and the hysteresis effect coefficient exceeds the slow response threshold, prompt the synovial membrane stickiness abnormality; If the prediction signal is severe and the hysteresis effect coefficient is lower than the fast response threshold, prompt the joint capsule relaxation abnormality.

8. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 7, characterized in that: The hysteresis effect coefficient is calculated according to the following formula: ； wherein is the arithmetic mean of the absolute values of the sequence of joint angle changes, is the optimal hysteresis step size, is the magnetic resonance image sampling time interval, is the hysteresis effect coefficient.

9. The large model-based knee osteoarthritis dynamic grading prediction method according to claim 7, characterized in that: The time shift operation specifically includes: Replace the i-th element in the synovial fluid volume quantization value sequence with the i+d-th element to generate a time-shifted fluid volume sequence with a time shift step of d, where d is an integer and d≥0.

10. A large model-based dynamic grading prediction system for knee osteoarthritis, characterized by: Use the large model-based dynamic grading prediction method for knee osteoarthritis as claimed in any one of claims 1-9, comprising: An image acquisition module for acquiring knee joint MRI image sequence and knee joint angle time sequence data of a patient during a standard flexion-extension motion; A joint cavity segmentation module for processing the knee joint MRI image sequence based on a pre-trained joint cavity segmentation model to obtain the synovial fluid volume quantization value corresponding to each image frame; A data synchronization module for aligning the knee joint angle time sequence data to the image frame acquisition time point to generate a synchronized joint angle sequence; A fluid volume change rate calculation module for calculating a fluid volume change rate sequence at adjacent time points based on the synovial fluid volume quantization value sequence; An angle change amount calculation module for calculating an angle change amount sequence at adjacent time points based on the synchronized joint angle sequence; A first judgment module for judging whether the average value of the absolute value of the fluid volume change rate sequence is lower than a preset stability threshold: if yes, output a knee osteoarthritis low-risk prediction signal, otherwise calculate the correlation coefficient of the fluid volume change rate sequence and the angle change amount sequence; A second judgment module for judging whether the absolute value of the correlation coefficient is lower than a preset correlation threshold: if yes, output a moderate prediction signal for knee osteoarthritis, otherwise further judge whether the correlation coefficient is positive: if yes, output a severe prediction signal for knee osteoarthritis, otherwise output a mild or moderate prediction signal for knee osteoarthritis according to the amplitude feature of the fluid volume change rate sequence.

Images