Mri-based semi-quantitative evaluation benchmark guided knee image semi-quantitative evaluation method

Abstract

This invention relates to the interdisciplinary field of medical image processing, computer vision, and artificial intelligence-assisted diagnosis, and particularly to a semi-quantitative assessment method for knee joint images guided by MRI semi-quantitative assessment benchmarks. The method includes: S1. Generating a structured semi-quantitative scoring vector based on knee MRI images of multiple groups of subjects; S2. Obtaining knee X-ray or CT images of the multiple groups of subjects within a preset clinical stability window, and training a model based on the structured semi-quantitative scoring vector and the knee X-ray or CT images to construct a mapping relationship between a parameter set and the knee X-ray or CT images; S3. Obtaining only the knee X-ray or CT images of the target subject, and outputting the semi-quantitative assessment result of the knee joint based on the mapping relationship through the model; S4. A screening-level result generation step. This invention can solve the problem of rapid risk stratification and management of knee joint diseases in primary healthcare institutions lacking high-precision screening methods.

Description

Technical Field

[0001] This invention relates to the interdisciplinary fields of medical image processing, computer vision, and artificial intelligence-assisted diagnosis, and particularly to a method, system, and storage medium for semi-quantitative evaluation of knee joint images guided by MRI semi-quantitative evaluation benchmarks. Background Technology

[0002] Knee degenerative diseases have a high incidence rate among middle-aged and elderly people. In clinical practice, precise assessment of the severity of these lesions typically relies on MRI imaging, combined with semi-quantitative assessment criteria such as WORMS, MOAKS, BLOKS, and KOSS. These assessment criteria can systematically grade changes in cartilage, bone marrow, edema, meniscus, and joint structure. However, MRI examinations are costly and time-consuming, making it difficult to implement on a large scale as a routine screening method in primary healthcare institutions and health check-up centers.

[0003] In comparison, knee X-ray and low-dose CT imaging have advantages such as high equipment availability, simple examination procedures, and low cost, making them suitable for initial screening. Current techniques typically attempt to find image features in X-ray and CT images that resemble MRI images. However, due to the limited dimensionality of information in X-ray or CT images, a significant amount of soft tissue information is lost. This method of finding similarities is extremely unstable when dealing with tissues such as cartilage and synovium that are not visible on X-rays. Moreover, due to the limited dimensionality of information in X-ray or CT images, current techniques struggle to provide a reliable and consistent assessment of the degree of knee joint degeneration directly based on X-ray or CT images. Summary of the Invention

[0004] In view of the above-mentioned prior art, the first aspect of this application proposes a method for semi-quantitative evaluation of knee joint images guided by MRI semi-quantitative evaluation benchmarks, which includes the following steps: S1. The assessment benchmark establishment step includes: generating a structured semi-quantitative scoring vector based on knee MRI images of multiple groups of subjects according to a preset semi-quantitative assessment standard for knee joints; the structured semi-quantitative scoring vector includes multiple assessment dimensions determined based on the semi-quantitative assessment standard and the output scale and result grade corresponding to each assessment dimension. S2. The engineering mapping relationship construction step includes: acquiring knee X-ray images or CT images of the multiple groups of subjects within a preset clinical stable window period; and training the model based on the structured semi-quantitative scoring vector and the knee X-ray images or CT images to construct an image assessment mapping relationship between the structured semi-quantitative scoring vector and the knee X-ray images or CT images, so that the multiple assessment dimensions, output scales, and result grades extracted by the model from the knee X-ray images or CT images based on the image assessment mapping relationship are consistent with the assessment dimensions, output scales, and result grades in the structured semi-quantitative scoring vector generated based on MRI images; S3. Single-modal reasoning evaluation step, which includes: acquiring only the knee joint X-ray image or CT image of the target object, and based on the image evaluation mapping relationship, outputting the semi-quantitative evaluation result of the knee joint of the target object containing the evaluation dimension, output scale and result classification through the model; S4. Screening-level result generation step, which includes: using the semi-quantitative knee joint assessment results from step S3 as clinical decision support parameters to generate a knee joint status assessment report.

[0005] According to some embodiments of the present invention, the semi-quantitative assessment criteria include one or more of WORMS, MOAKS, BLOKS, and KOSS; the preset clinical stability window period is 0-30 days.

[0006] According to some embodiments of the present invention, the assessment dimensions include cartilage damage, bone marrow lesions, meniscus injury, and synovitis manifestations.

[0007] According to some embodiments of the present invention, the semi-quantitative assessment results of the knee joint further include: Global assessment score: A quantitative value reflecting the overall degree of knee joint degeneration; Localized risk warning: Alerts for suspected lesions in specific areas of cartilage, bone marrow, or meniscus.

[0008] According to some embodiments of the present invention, the method further includes preprocessing the X-ray or CT images before model training based on the structured semi-quantitative scoring vector and the knee joint X-ray or CT images, including: automatically locating and cropping the region of interest on the X-ray or CT images.

[0009] According to some embodiments of the present invention, the model training in step S2 employs cross-modal feature distillation technology, constraining the model through a cross-modal loss function, so that the multiple evaluation dimensions, output scales, and result grades extracted by the model from knee X-ray images or CT images based on the image evaluation mapping relationship are consistent with the evaluation dimensions, output scales, and result grades in the structured semi-quantitative scoring vector generated based on MRI images.

[0010] According to some embodiments of the present invention, the assessment report generated in step S4 includes a risk classification, which is automatically divided into "healthy", "medium risk" and "high risk" levels based on the assessment score, and matched with corresponding personalized treatment recommendations.

[0011] This application also proposes a semi-quantitative assessment system for knee joint images guided by MRI semi-quantitative assessment benchmarks, which includes: The assessment benchmark construction module is used to generate a structured semi-quantitative scoring vector based on the knee joint MRI images of the examinee and according to the preset semi-quantitative assessment criteria for the knee joint. The structured semi-quantitative scoring vector includes assessment dimensions, output scale and result grading. The mapping relationship construction module is used to train the model based on the structured semi-quantitative scoring vector and the knee X-ray or CT images of the subject within a preset clinical stable window period, so that the model can construct an image assessment mapping relationship between the structured semi-quantitative scoring vector and the knee X-ray or CT images for engineering applications, so that the assessment dimensions, output scales and result grades extracted by the model from the knee X-ray or CT images based on the image assessment mapping relationship are consistent with the assessment dimensions, output scales and result grades of the knee MRI images. The single-modal reasoning evaluation module is used to acquire only the knee joint X-ray or CT images of the target object, and based on the image evaluation mapping relationship, output a semi-quantitative evaluation result of the knee joint that includes evaluation dimensions, output scale and result classification through the model. The report generation module is used to generate a knee joint status assessment report by using the semi-quantitative assessment results of the knee joint as clinical decision support parameters.

[0012] According to some embodiments of the present invention, the system further includes a quality control module for automatically detecting the quality of X-ray or CT images and providing retake suggestions for unqualified images, wherein the quality of the X-ray or CT images is related to exposure, position, or motion artifacts.

[0013] This application also proposes a computer-readable storage medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the semi-quantitative evaluation method for knee joint images.

[0014] This invention introduces a high-precision MRI benchmark during the training phase, enabling near-MRI-level assessment accuracy during the inference phase using only low-cost, widely available X-ray or CT images. This effectively addresses the lack of high-precision screening methods in primary healthcare institutions, achieving rapid risk stratification and management of knee joint diseases. Specifically, in terms of quantifiable effects, the invention reduces the cost per examination by approximately 40% compared to MRI screening, making it suitable for large-scale population screening. Regarding diagnostic gains, the invention demonstrates a 20-30% increase in sensitivity for early detection of cartilage lesions compared to traditional X-ray assessment. Finally, in terms of standardization, the output results directly correspond to internationally recognized MOAKS / WORMS scores, facilitating multi-center studies and longitudinal follow-up. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the semi-quantitative evaluation method for knee joint images based on MRI semi-quantitative evaluation benchmarks provided in an embodiment of the present invention. Figure 2 A structural block diagram of a knee joint imaging semi-quantitative assessment system based on MRI semi-quantitative assessment benchmarks provided in an embodiment of the present invention; Figure 3 A schematic block diagram of a knee joint imaging semi-quantitative assessment device based on MRI semi-quantitative assessment benchmarks, according to some embodiments of the present invention, is shown. Detailed Implementation

[0016] The present invention will now be described by way of example with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. Furthermore, the described embodiments are only some, not all, of the embodiments of the present invention.

[0017] Figure 1 A flowchart illustrating a semi-quantitative assessment method for knee joint images guided by MRI semi-quantitative assessment benchmarks according to some embodiments of the present invention is shown. Figure 1 As shown, the method includes S1-S4: S1. The assessment benchmark establishment step includes: generating a structured semi-quantitative scoring vector based on knee MRI images of multiple groups of subjects according to a preset semi-quantitative assessment standard for the knee joint; the structured semi-quantitative scoring vector includes multiple assessment dimensions determined based on the semi-quantitative assessment standard, as well as the output scale and result grading corresponding to each assessment dimension.

[0018] In this step, multiple groups of subjects were selected and asked to undergo MRI (magnetic resonance imaging). MRI images have extremely high soft tissue resolution, allowing for clear observation of cartilage thickness, the degree of meniscus tear, and synovial inflammation.

[0019] According to some embodiments of the present invention, the semi-quantitative assessment criteria include one or more of the internationally recognized WORMS, MOAKS, BLOKS, and KOSS. Assessment dimensions include cartilage damage, bone marrow lesions, meniscus injuries, and synovitis manifestations. Through expert annotation or high-precision algorithms, each MRI image can be transformed into a "structured semi-quantitative scoring vector." This scoring parameter set is not a total score, but a set of finely structured data, including cartilage damage grading, bone marrow lesion scores, meniscus injury severity, and synovitis manifestations. These parameters constitute the "ground truth" for subsequent model training, i.e., the ultimate goal of AI learning. The table below, using MAOKS as an example, provides an exemplary structured semi-quantitative scoring vector.

[0020]

[0021] The meniscus injury assessment includes: geometric feature analysis based on the deviation of the joint space biomechanical line; and biomechanical modeling combined with the relative position of the femoral condyle and tibial plateau.

[0022] The assessment of synovitis manifestations includes: fuzzy quantitative analysis of the infrapatellar fat pad shadow; detection of indirect signs of joint effusion (such as suprapatellar bursa dilatation).

[0023] S2. The engineering mapping relationship construction step includes: based on the structured semi-quantitative scoring vector and the knee X-ray or CT images of the subject within a preset clinical stable window period, performing model training, so that the model constructs an image evaluation mapping relationship between the structured semi-quantitative scoring vector for engineering applications and the knee X-ray or CT images, so that the evaluation dimensions, output scales and result grades extracted by the model from the knee X-ray or CT images based on the image evaluation mapping relationship are consistent with the evaluation dimensions, output scales and result grades of the knee MRI images.

[0024] In this step, multiple groups of subjects from step S1 are required to complete X-ray imaging within a very short period, i.e., within the clinically stable window (ensuring that the pathological state of the knee joint has not undergone physiological changes). The preset clinically stable window is 0-30 days, preferably no more than 7 days, and this period is determined based on the following clinical criteria: (a) The natural progression timescale of knee cartilage degeneration; (b) The stability window of inflammatory activity in osteoarthritis; (c) The balance between patient compliance and clinical feasibility.

[0025] During the training phase, the input device receives pairs of image data. One of the pairs of image data is an MRI image with high soft tissue resolution, and the other is an X-ray or CT image taken of the same subject within a clinically stable window.

[0026] During training, by introducing a cross-modal loss function, the model is forced to output results that approximate the structured semi-quantitative scoring vector based on MRI image data of the same subject when receiving X-ray or CT images as input.

[0027] For example, when X-ray or CT images are input into the model, the model might initially predict "cartilage damage 3 points" and "synovitis 2 points," etc. Calculations show an error between the initial prediction and the true MRI values ​​(1 point and 1 point respectively). The neural network parameters are adjusted using a backpropagation algorithm until the model can learn, from subtle features such as bone edges and gaps in X-ray or CT images, the cartilage loss pattern aligned with the structured semi-quantitative scoring vector of MRI, as shown in Table 2 below. The mapping in Table 2 is merely illustrative; in practice, through continuous training, it can learn, based on more details and a superhuman understanding of image patterns, the cartilage loss pattern aligned with the structured semi-quantitative scoring vector of MRI from features beyond the conventional features of X-ray or CT images.

[0028]

[0029] The goal of this step is to train a deep neural network that can output assessment dimensions, output scales, and result grading similar to the "gold standard" MRI, even when only seeing X-ray or CT images (only bone outlines). This will greatly improve the sensitivity of screening tools and provide technical support for early warning of knee osteoarthritis.

[0030] According to some embodiments of the present invention, in the model training described in step S2, a cross-modal feature distillation technique is used to constrain the model through a cross-modal loss function, so that the multiple evaluation dimensions, output scales and result grades extracted by the model from knee X-ray images or CT images based on the image evaluation mapping relationship are consistent with the evaluation dimensions, output scales and result grades in the structured semi-quantitative scoring vector generated based on MRI images.

[0031] Due to the physical imaging principles, X-rays inherently lose soft tissue data. Simple mapping is merely rote memorization, while distillation involves logical reasoning. For example, it teaches the model that "although cartilage is not visible, when such minute sclerosis or relative displacement occurs at the bone edge, the corresponding cartilage on the MRI side is usually already thinned." Through distillation, the model outputs no longer a vague "good" or "bad," but a professional score that strictly conforms to internationally recognized semi-quantitative MRI standards such as WORMS and MOAKS. This means that X-ray imaging reports from primary hospitals have extremely high reference value in the eyes of authoritative doctors.

[0032] S3. Single-modal reasoning evaluation step, which includes: acquiring only the knee joint X-ray or CT image of the target object, and based on the image evaluation mapping relationship, outputting the semi-quantitative evaluation result of the knee joint of the target object containing the evaluation dimension, output scale and result classification through the model.

[0033] When deploying the algorithm in a real hospital environment, image preprocessing is the first step. Since X-ray or CT images taken by different machines vary in size, angle, and brightness, "standardized cropping" is necessary to ensure accurate AI interpretation. Based on MRI anatomical partitioning benchmarks (such as one or both sides of the medial and lateral tibiofemoral joint region or subchondral bone region (weight-bearing area), we automatically locate the corresponding region of interest (ROI) on the X-ray or CT images. Next, grayscale normalization is performed to adjust the image's brightness distribution to a specific range to meet the computational requirements set during training of the cross-modal loss function.

[0034] When a standardized X-ray or CT image is input into the system, the pre-trained model performs "single-modal inference." This means that the system no longer needs MRI data; based solely on the X-ray or CT image, it can directly output a semi-quantitative assessment result of the knee joint, including assessment dimensions, output scale, and result grading, through internal neuronal calculations.

[0035] This step enables independent assessment completely independent of MRI equipment, significantly reducing examination costs. The standardized preprocessing process ensures the algorithm's generalization ability and robustness across different devices, enabling large-scale deployment in primary healthcare institutions and demonstrating high engineering practical value.

[0036] S4. Screening-level result generation step, which includes: using the semi-quantitative knee joint assessment results from step S3 as clinical decision support parameters to generate a knee joint status assessment report.

[0037] After receiving the score output by the AI, the system's internal report generation module automatically compares it with the "risk level database" in the background. If the score exceeds a certain threshold, the system will automatically classify it into "high risk," "medium risk," or "healthy" levels. Simultaneously, the report can also be linked to a pre-set database of treatment recommendations. For example, for a patient diagnosed with "mild cartilage damage," the report will automatically include personalized suggestions such as "reducing weight-bearing exercises and increasing quadriceps training."

[0038] By transforming complex algorithmic outputs into easily readable status assessment reports, a closed loop for medical decision-making is achieved. The automatically matched risk levels and treatment recommendations not only improve diagnostic efficiency but also provide a standardized data interface for the tiered diagnosis and treatment of knee joint diseases and chronic disease management.

[0039] This invention provides a high-precision "digital ruler" for previously blurry X-ray or CT images. This method breaks through the limitation of traditional X-ray assessment, which can only observe bone hyperplasia, and elevates the assessment dimension to the soft tissue level, thereby improving the scientific rigor and authority of the assessment system from the source.

[0040] Figure 2 A schematic block diagram of a knee joint imaging semi-quantitative assessment system 3000 guided by MRI semi-quantitative assessment benchmarks according to some embodiments of the present invention is shown. The system includes: The evaluation benchmark construction module 3001 is used to generate a structured semi-quantitative scoring vector based on the knee joint MRI images of multiple groups of subjects, according to a preset semi-quantitative evaluation standard for the knee joint; the structured semi-quantitative scoring vector includes multiple evaluation dimensions determined based on the semi-quantitative evaluation standard, as well as the output scale and result grading corresponding to each evaluation dimension.

[0041] The mapping relationship construction module 3002 is used to acquire the knee X-ray images or CT images of the multiple groups of subjects within a preset clinical stable window period; and to train the model based on the structured semi-quantitative scoring vector and the knee X-ray images or CT images to construct the image assessment mapping relationship between the structured semi-quantitative scoring vector and the knee X-ray images or CT images, so that the multiple assessment dimensions, output scales and result grades extracted by the model from the knee X-ray images or CT images based on the image assessment mapping relationship are consistent with the assessment dimensions, output scales and result grades in the structured semi-quantitative scoring vector generated based on MRI images.

[0042] The single-modal reasoning evaluation module 3003 is used to acquire only the knee joint X-ray or CT images of the target object, and based on the image evaluation mapping relationship, output the semi-quantitative evaluation results of the knee joint of the target object, including the evaluation dimension, output scale and result classification, through the model.

[0043] The report generation module 3004 is used to generate a knee joint status assessment report by using the semi-quantitative assessment results of the knee joint as clinical decision support parameters.

[0044] According to some embodiments of the present invention, the system further includes a quality control module for automatically detecting the quality of X-ray or CT images and providing retake suggestions for unqualified images, wherein the quality of the X-ray or CT images is related to exposure, position, or motion artifacts.

[0045] According to some embodiments of the present invention, the semi-quantitative evaluation criteria include one or more of WORMS, MOAKS, BLOKS, and KOSS.

[0046] According to some embodiments of the present invention, the assessment dimensions include cartilage damage, bone marrow lesions, meniscus injury, and synovitis manifestations.

[0047] According to some embodiments of the present invention, the semi-quantitative assessment results of the knee joint further include: Global assessment score: A quantitative value reflecting the overall degree of knee joint degeneration; Localized risk warning: Alerts for suspected lesions in specific areas of cartilage, bone marrow, or meniscus.

[0048] According to some embodiments of the present invention, the mapping relationship construction module preprocesses the X-ray image or CT image before model training, including automatically locating and cropping the region of interest on the X-ray image or CT image.

[0049] According to some embodiments of the present invention, the model training employs cross-modal feature distillation technology, which constrains the model through a cross-modal loss function, so that the multiple evaluation dimensions, output scales, and result grades extracted by the model from knee X-ray images or CT images based on the image evaluation mapping relationship are consistent with the evaluation dimensions, output scales, and result grades in the structured semi-quantitative scoring vector generated based on MRI images.

[0050] According to some embodiments of the present invention, the assessment report includes a risk grading, which is automatically divided into "healthy", "medium risk" and "high risk" levels based on the assessment score, and matched with corresponding personalized treatment recommendations.

[0051] In addition, the system may include more modules or units for implementing the various steps and details of the aforementioned method respectively.

[0052] In this application, the units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0053] In addition, the previous reference Figure 1 The corresponding detailed descriptions also apply to, or, with appropriate adjustments, to, those descriptions. Figure 2 For the sake of simplicity, the specific implementation examples will not be described in detail here.

[0054] Figure 3 A schematic structural diagram of a knee joint imaging semi-quantitative assessment device 500 guided by an MRI semi-quantitative assessment benchmark, according to some embodiments of the present invention, is shown. Figure 3 As shown, the device includes a processor 51, a memory 52, and a bus 53.

[0055] In some instances, the device may further include an input device 501, an input port 502, an output port 503, and an output device 504. The input port 502, processor 51, memory 52, and output port 503 are interconnected via a bus 53. The input device 501 and output device 504 are connected to the bus 53 via the input port 502 and output port 503, respectively, and thus connected to other components of the device. It should be noted that the output and input interfaces here can also be represented by I / O interfaces. Specifically, the input device 501 receives input information from the outside, such as various image and non-image data, and transmits the input information to the processor 51 via the input port 502. The processor 51 processes the input information based on computer-executable instructions stored in the memory 52 to generate output information, temporarily or permanently stores the output information in the memory 52, and then transmits the output information to the output device 504 via the output port 503. The output device 504 outputs the output information to the outside of the device. The output information may be, for example, a semi-quantitative assessment report of knee joint images guided by MRI semi-quantitative assessment benchmarks.

[0056] The processor 51 executes the semi-quantitative knee joint imaging assessment method based on the computer program stored in the memory 52.

[0057] According to further embodiments of the present invention, the computer program can be divided into one or more units in various ways, stored in the memory, and executed by the processor to complete the present invention. The one or more units can be a series of computer program instruction segments capable of performing a specific function, the instruction segments describing the execution process of the computer program in the device. The computer program can be divided into multiple units according to the functions of the units in the various embodiments described above with reference to the figures, or may include the units in the various embodiments described above with reference to the figures. For simplicity, these will not be repeated here.

[0058] The processor may be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits (ASICs), off-the-shelf programmable gate arrays (OPGs) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the device, connecting various parts of the device via various interfaces and lines. The device may be a desktop computer, laptop, handheld computer, cloud server, or other computing devices or a part thereof. The device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the schematic diagrams are merely examples of the device and do not constitute a limitation on the device.

[0059] The detailed descriptions of the figures mentioned above are included here by reference and will not be repeated here.

[0060] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method for semi-quantitative evaluation of knee joint images guided by MRI semi-quantitative evaluation benchmarks.

[0061] The computer program includes computer program code, which may be in the form of source code, object code, executable file, or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention. It should be noted that although the structure of the device and its operation method are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be executed in a different order. Additionally or alternatively, certain steps can be omitted, multiple steps can be combined into one step, and / or one step can be broken down into multiple steps.

Claims

1. A semi-quantitative assessment method for knee joint images guided by MRI semi-quantitative assessment benchmarks, comprising the following steps: S1. The assessment benchmark establishment step includes: generating a structured semi-quantitative scoring vector based on knee MRI images of multiple groups of subjects according to a preset semi-quantitative assessment standard for knee joints; the structured semi-quantitative scoring vector includes multiple assessment dimensions determined based on the semi-quantitative assessment standard and the output scale and result grade corresponding to each assessment dimension. S2. The engineering mapping relationship construction step includes: acquiring knee X-ray images or CT images of the multiple groups of subjects within a preset clinical stable window period; and training the model based on the structured semi-quantitative scoring vector and the knee X-ray images or CT images to construct an image assessment mapping relationship between the structured semi-quantitative scoring vector and the knee X-ray images or CT images, so that the multiple assessment dimensions, output scales, and result grades extracted by the model from the knee X-ray images or CT images based on the image assessment mapping relationship are consistent with the assessment dimensions, output scales, and result grades in the structured semi-quantitative scoring vector generated based on MRI images; S3. Single-modal reasoning evaluation step, which includes: acquiring only the knee joint X-ray image or CT image of the target object, and based on the image evaluation mapping relationship, outputting the semi-quantitative evaluation result of the knee joint of the target object containing the evaluation dimension, output scale and result classification through the model; S4. Screening-level result generation step, which includes: using the semi-quantitative knee joint assessment results from step S3 as clinical decision support parameters to generate a knee joint status assessment report.

2. The method according to claim 1, wherein, The semi-quantitative assessment criteria include one or more of WORMS, MOAKS, BLOKS, and KOSS; the preset clinical stability window period is 0-30 days.

3. The method according to claim 1, wherein, The assessment dimensions include cartilage damage, bone marrow lesions, meniscus injury, and synovitis manifestations.

4. The method according to claim 1, wherein, The semi-quantitative assessment results of the knee joint also include: Global assessment score: A quantitative value reflecting the overall degree of knee joint degeneration; Localized risk warning: Alerts for suspected lesions in specific areas of cartilage, bone marrow, or meniscus.

5. The method according to claim 1, further comprising preprocessing the X-ray or CT images before model training based on the structured semi-quantitative scoring vector and the knee joint X-ray or CT images, including: Automatically locate and crop the region of interest on X-ray or CT images.

6. The method according to claim 2, wherein, The model training described in step S2 employs cross-modal feature distillation technology. By constraining the model through a cross-modal loss function, the multiple evaluation dimensions, output scales, and result grades extracted from knee X-ray or CT images based on the image evaluation mapping relationship are consistent with the evaluation dimensions, output scales, and result grades in the structured semi-quantitative scoring vector generated based on MRI images.

7. The method according to claim 1, wherein, The assessment report generated in step S4 includes a risk classification, which is automatically divided into "healthy", "medium risk" and "high risk" levels based on the assessment score, and matched with corresponding personalized treatment recommendations.

8. A semi-quantitative assessment system for knee joint images guided by MRI semi-quantitative assessment benchmarks, comprising: The assessment benchmark construction module is used to generate a structured semi-quantitative scoring vector based on the knee joint MRI images of the examinee and according to the preset semi-quantitative assessment criteria for the knee joint. The structured semi-quantitative scoring vector includes assessment dimensions, output scale and result grading. The mapping relationship construction module is used to train the model based on the structured semi-quantitative scoring vector and the knee X-ray or CT images of the subject within a preset clinical stable window period, so that the model can construct an image assessment mapping relationship between the structured semi-quantitative scoring vector and the knee X-ray or CT images for engineering applications, so that the assessment dimensions, output scales and result grades extracted by the model from the knee X-ray or CT images based on the image assessment mapping relationship are consistent with the assessment dimensions, output scales and result grades of the knee MRI images. The single-modal reasoning evaluation module is used to acquire only the knee joint X-ray or CT images of the target object, and based on the image evaluation mapping relationship, output a semi-quantitative evaluation result of the knee joint that includes evaluation dimensions, output scale and result classification through the model. The report generation module is used to generate a knee joint status assessment report by using the semi-quantitative assessment results of the knee joint as clinical decision support parameters.

9. The system according to claim 8, wherein, The system also includes a quality control module, which is used to automatically detect the quality of X-ray or CT images and provide suggestions for reshooting unqualified images. The quality of the X-ray or CT images is related to exposure, body position, or motion artifacts.

10. A computer-readable storage medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the semi-quantitative evaluation method for knee joint images as described in any one of claims 1-8.

Images