Three-dimensional reconstruction method and system for knee joint

Abstract

The invention relates to the technical field of medical processing, and discloses a three-dimensional reconstruction method and system for a knee joint, and the method comprises the steps: obtaining a knee joint image through high-resolution CT and MRI scanning, storing the knee joint image in a DICOM format, and importing Mimics software; generating a femur and tibia binary image based on the MRI image, determining a region of interest containing the cruciate ligament, and separating a ligament image; acquiring anterior and posterior cruciate ligament angle lines by using an angle line principle, extracting an accurate contour through region growth and an active contour model, and determining ligament dead points; registering the MRI image and the CT image based on an ICP algorithm, and mapping a dead point to the CT image; and performing three-dimensional surface drawing on the dead point image and the CT image to generate a knee joint three-dimensional model containing bones and soft tissues. According to the method, the knee joint three-dimensional model of natural connection of the cruciate ligament stop point and the skeleton surface is generated.

Description

Technical Field

[0001] This invention relates to the field of medical processing technology, and more specifically, to a three-dimensional reconstruction method and system for the knee joint. Background Technology

[0002] The knee joint is a vital weight-bearing joint in the human body, with a complex structure composed of the femur, tibia, patella, and multiple ligaments. Among these, the cruciate ligaments (anterior cruciate ligament and posterior cruciate ligament) play a crucial role in knee joint stability. In clinical surgeries such as knee injuries, ligament reconstruction, and joint replacement, accurately determining the insertion point of the cruciate ligaments is of great significance for preoperative planning, surgical navigation, and postoperative rehabilitation assessment.

[0003] For example, the method for extracting and reconstructing the cruciate ligament insertion point of the human knee joint disclosed in CN113936100A, although capable of preprocessing and 3D modeling MRI or CT images, still has the following problems in extracting the cruciate ligament insertion point: low ligament segmentation accuracy, making it difficult to accurately separate the cruciate ligament from the surrounding soft tissue, resulting in deviations in insertion point positioning; difficulty in angle and direction identification, as the attachment angles of the anterior cruciate ligament and posterior cruciate ligament on the femur and tibia are complex, and conventional image analysis methods cannot accurately identify the ligament direction and angle; low registration efficiency, as manual matching and registration of MRI and CT images is time-consuming, and there is a lack of automated methods to accurately calculate the position of the insertion point in the CT image; and limited 3D reconstruction accuracy, as existing methods have registration errors in the 3D modeling of bones and soft tissues, and cannot accurately show the spatial relationship between the ligament insertion point and the bone surface.

[0004] Therefore, it is necessary to design a three-dimensional reconstruction method and system for the knee joint to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a three-dimensional reconstruction method and system for the knee joint, aiming to solve the problems of low ligament segmentation accuracy, difficulty in angle and direction identification, low registration efficiency, and limited three-dimensional reconstruction accuracy.

[0006] In one aspect, the present invention proposes a three-dimensional reconstruction method for the knee joint, comprising: Medical image data of the human knee joint is acquired through high-resolution CT and MRI scans and stored in DICOM format; the DICOM format data is then imported into Mimics software. In the Mimics software, binary images of the femur and tibia are acquired based on MRI images, and a region of interest containing the cruciate ligament is determined; the region of interest is processed to separate the ligament image and obtain images of the ligament that is disconnected from other soft tissues; the angle lines of the anterior cruciate ligament and the posterior cruciate ligament are obtained based on the angle line principle. The coarse outline of the cruciate ligament is obtained using region growing technology, and the precise outline is obtained through iterative optimization using an active contour model. The intersection points of the precise outline with the femoral and tibial contours are extracted as the insertion points of the cruciate ligament. Based on the ICP algorithm, the MRI and CT images are registered, and the insertion points are mapped into the CT images to construct the insertion point images. The stop point image and CT image are used to perform surface rendering and three-dimensional reconstruction to generate a three-dimensional model of the bones and soft tissues around the knee joint. The three-dimensional model includes a bone model and a tissue and organ model.

[0007] Furthermore, when acquiring binary images of the femur and tibia based on MRI images and determining the region of interest containing the cruciate ligament, the process includes: The MRI image is binarized, and the binarized image is traversed to obtain the principal axis based on the shape features of the femur and tibia in the image. Contour detection is performed on the binarized image, and the contours containing points on the principal axis are traversed. The two contours with the largest areas are selected as the femoral and tibial contours, with the one above being the femoral contour and the one below being the tibial contour. Using the positional information of the cruciate ligament on the femur and tibia, the rightmost point of the femur, the rightmost point of the tibia, the leftmost point of the tibia, and the principal axis are used as the boundary points of the cruciate ligament of the human knee joint to determine the region of interest of the cruciate ligament.

[0008] Furthermore, processing the region of interest to separate the ligament image and obtain an image of the ligament that has broken away from other soft tissues includes: The obtained region of interest was divided into four intervals using the K-Means algorithm. Pixels in the same interval were represented by the same color to obtain a pseudo-color image of the region of interest. The gray value region of the cruciate ligament was extracted and binarized. The ligament was then separated from other soft tissues using an erosion operation. The erosion operation was performed using a 3×3 structuring element for 3 iterations to obtain an image of the ligament separated from other soft tissues.

[0009] Furthermore, when obtaining the anterior cruciate ligament (ACL) and posterior cruciate ligament (PCL) angle lines based on the angle line principle, the following steps are included: Based on the principle that the anterior cruciate ligament (ACL) and the intercondylar fossa of the femur are at a 13° angle, the rightmost point of each row is traversed downwards from the rightmost point of the femur until the principal axis is encountered, thus obtaining the points on the intercondylar fossa of the femur. A straight line is fitted to these points using the least squares method. The ACL angle line is obtained by rotating counterclockwise by 13° with the rightmost point of the femur as the center point. Straight lines at 30° and 60° to the horizontal direction are constructed respectively, and the obtained ligament images that are disconnected from other soft tissues are traversed. When the two straight lines contain the most ligament pixels respectively, the posterior cruciate ligament (PCL) angle line is obtained.

[0010] Furthermore, when using region growing techniques to obtain the coarse outline of the cruciate ligament and iteratively optimizing it through an active contour model to obtain the precise outline, the process includes: Starting from the rightmost point of the femur, the search proceeds downwards, using the first ligament point encountered as the seed point. Region growth technology is used to obtain the coarse outline of the anterior cruciate ligament (ACL). Similarly, starting from the intersection of the two angle lines of the posterior cruciate ligament (PCL), the search proceeds upwards or downwards, using the first ligament point encountered as the seed point. Region growth technology is used to obtain the coarse outline of the PCL, thus obtaining the coarse outline of the cruciate ligament (CXL). Using the coarse outline of the CXL as the initial outline, an active outline model is used for iteration. The initial outline deforms under the influence of internal and external forces. External energy attracts the active outline towards the ligament edge, while internal energy maintains the smoothness and topology of the active outline. When the energy function reaches its minimum, the active outline converges to the ligament edge, obtaining the precise outline of the CXL.

[0011] Furthermore, when extracting the intersection point of the precise contour with the contour lines of the femur and tibia as the insertion point of the cruciate ligament, the following steps are included: Curve fitting is performed on the precise contour to obtain a smooth cruciate ligament contour curve; the intersection point is determined by calculating the minimum Euclidean distance between the cruciate ligament contour curve and the femoral and tibial contour lines; the intersection point is precisely located at the subpixel level and used as the insertion point of the cruciate ligament.

[0012] Furthermore, based on the ICP algorithm, the MRI image and CT image are registered, and the endpoint is mapped onto the CT image. When constructing the endpoint image, the process includes: The CT image is binarized. Contour detection is used to obtain the two contours with the largest area in the CT image. The upper contour is taken as the femur, and the lower contour as the tibia. Based on the ICP algorithm, the distance from each contour point in the CT image to the contour in the MRI image is calculated for both the femur and tibia. The sum of the distance values ​​is then selected as the CT image corresponding to the smallest sum of distance values. The contours of the femur and tibia in the MRI image are registered with the corresponding femur and tibia obtained from the CT image using a two-dimensional affine transformation registration function, resulting in two rotation and translation matrices. The stop point information of the femur and tibia is transformed according to the rotation and translation matrices obtained from the registration, creating a pure background binary image of the same size as the CT image. The stop points are then placed at the corresponding positions in the CT image to obtain the stop point image.

[0013] Furthermore, when performing surface rendering and three-dimensional reconstruction on the stop point image and CT image to generate a three-dimensional model of the bones and soft tissues around the knee joint, the process includes: The Marching Cubes algorithm was used to render the CT images to generate a 3D model of the knee joint bones. 3D interpolation and surface fitting were performed on the insertion point images to generate a 3D model of the cruciate ligament insertion point. The bone model, the cruciate ligament insertion point model, and other soft tissue models were spatially registered and fused to generate a complete 3D model of the bones and soft tissues around the knee joint.

[0014] Furthermore, when spatially registering and fusing skeletal models, cruciate ligament insertion models, and other soft tissue models, the following steps are included: A unified world coordinate system was established to transform the skeletal model, cruciate ligament (ACL) insertion model, and soft tissue model into the same coordinate space. The lateral epicondyle of the femur, medial epicondyle, tibial plateau edge, and patellar contour were selected as registration reference points to calculate the initial correspondence between the skeletal and soft tissue models. Based on an improved non-rigid registration algorithm, the soft tissue model was locally deformed and adjusted using the skeletal model as the reference, ensuring that the average distance between key anatomical landmarks in the soft tissue model and their corresponding points in the skeletal model was less than 0.5 mm. A feature-based registration method was used to precisely align the ACL insertion model with the skeletal model. The spatial position of the ACL insertion model was adjusted by calculating the geometric relationship between the ligament insertion and the anatomical landmarks of the femoral intercondylar fossa and tibial plateau. Boolean operations and boundary smoothing were performed on the registered models to create a transition region at the bone-ligament connection, eliminating gaps and overlaps between models. A physics-based fusion algorithm was used to elastically deform the fusion region according to the biomechanical characteristics of the knee joint anatomy, ensuring a natural connection between the ACL insertion and the bone surface.

[0015] Compared with existing technologies, the advantages of this invention are as follows: By combining MRI and CT images, fully utilizing the soft tissue resolution of MRI and the bone precision of CT, a high-precision three-dimensional model of the bones and soft tissues around the knee joint can be generated. Accurate contours of the cruciate ligaments are extracted using region growing, active contour modeling, and angle line principles, and sub-pixel-level stop point localization is performed, improving the accuracy of stop point recognition. Precise registration of MRI and CT images is achieved based on the ICP algorithm and two-dimensional affine transformation; improved non-rigid registration and feature-based registration methods are used to achieve sub-millimeter-level accuracy in the correspondence between soft tissue and bone models. The bone, ligament, and soft tissue models are unified to the same coordinate system, and a natural transition connection between bones and ligaments is achieved through Boolean operations, boundary smoothing, and elastic deformation processing; this solves the problem of gaps or overlaps between bones, ligaments, and soft tissues in traditional models.

[0016] On the other hand, this application also provides a three-dimensional reconstruction system for the knee joint, for applying the aforementioned three-dimensional reconstruction method for the knee joint, comprising: The acquisition unit is configured to acquire medical image data of the human knee joint through high-resolution CT and MRI scans, and store the data in DICOM format; and import the DICOM format data into Mimics software; The extraction unit is configured in the Mimics software to acquire binary images of the femur and tibia based on MRI images and determine a region of interest containing the cruciate ligament; process the region of interest to separate the ligament image and acquire images of the ligament that is disconnected from other soft tissues; and acquire the angle lines of the anterior cruciate ligament and the posterior cruciate ligament based on the angle line principle. The processing unit is configured to obtain the coarse outline of the cruciate ligament using region growing technology, and obtain the precise outline through active contour model iterative optimization; extract the intersection point of the precise outline with the femoral and tibial contour lines as the insertion point of the cruciate ligament; register the MRI image and CT image based on the ICP algorithm, and map the insertion point into the CT image to construct the insertion point image; The construction unit is configured to perform surface rendering and three-dimensional reconstruction on the stop point image and CT image to generate a three-dimensional model of the bones and soft tissues around the knee joint, the three-dimensional model including a bone model and a tissue and organ model.

[0017] It is understandable that the above-mentioned three-dimensional reconstruction method and system for the knee joint have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A flowchart illustrating a three-dimensional reconstruction method for the knee joint provided in an embodiment of the present invention; Figure 2 This is a functional block diagram of a three-dimensional reconstruction system for the knee joint provided in an embodiment of the present invention. Detailed Implementation

[0019] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Currently, existing 3D reconstruction methods for the knee joint mainly rely on manual delineation and localization by surgeons on MRI or CT images, or the use of simple thresholding and 3D modeling tools. These methods suffer from low segmentation accuracy. The cruciate ligament and surrounding soft tissues are similar in grayscale values ​​and morphology, making it easy for conventional thresholding or region segmentation methods to misidentify soft tissue as ligaments, leading to inaccurate segmentation results. Insertion point localization relies on human experience; surgeons typically estimate the insertion point location on images based on anatomical knowledge, which is highly subjective and subject to individual differences, easily resulting in localization errors. Image registration is inaccurate. MRI images are superior in displaying soft tissues, while CT images are clearer in depicting bone structures. Existing technologies lack high-precision automated algorithms for MRI and CT image registration, often requiring manual intervention, affecting registration results and 3D reconstruction accuracy. The application of 3D reconstruction results is limited. Due to inaccurate ligament insertion point locations in the model, the generated 3D model has limited reference value in surgical navigation or personalized prosthesis design.

[0021] For example, in a historical cruciate ligament (ACL) reconstruction surgery, surgeons need to determine the ACL insertion points on the femur and tibia using MRI images and complete a three-dimensional reconstruction on CT images to develop a surgical drilling plan. Because current technology easily misidentifies some soft tissue as ligament during image segmentation, the femoral insertion point position can shift by approximately 2-3 millimeters. When this location is used for surgical navigation, the drilling position deviates, potentially causing abnormal tension in the ligament graft, thus affecting postoperative knee stability and functional recovery. This clearly demonstrates the shortcomings of current technology in terms of segmentation accuracy and insertion point localization.

[0022] In this regard, see some embodiments of this application. Figure 1 As shown, the three-dimensional reconstruction method for the knee joint includes: S100: Acquires medical image data of the human knee joint through high-resolution CT and MRI scans and stores the data in DICOM format; imports the DICOM format data into Mimics software; S200: In Mimics software, binary images of the femur and tibia are acquired based on MRI images, and the region of interest containing the cruciate ligament is determined; the region of interest is processed to separate the ligament image and obtain the ligament image that is disconnected from other soft tissues; the angle lines of the anterior cruciate ligament and the posterior cruciate ligament are obtained based on the angle line principle. S300: The coarse outline of the cruciate ligament is obtained using region growing technology, and the precise outline is obtained through iterative optimization of the active outline model; the intersection of the precise outline with the femoral and tibial outlines is extracted as the insertion point of the cruciate ligament; the MRI image and CT image are registered based on the ICP algorithm, and the insertion point is mapped into the CT image to construct the insertion point image. S400: Performs surface rendering and 3D reconstruction on the stop point image and CT image to generate a 3D model of the bones and soft tissues around the knee joint. The 3D model includes a bone model and a tissue and organ model.

[0023] Specifically, high-resolution CT and MRI scans were used to acquire bone and soft tissue images of the knee joint, which were stored in a standardized DICOM format to ensure compatibility and repeatability in subsequent medical image processing software. The DICOM data was imported into Mimics software, where binarization was performed based on the MRI images to extract the main contours of the femur and tibia, and regions of interest (ROIs) including the anterior and posterior cruciate ligaments (ACLs). By segmenting and processing these ROIs, the ligament images were separated from surrounding soft tissues, thus obtaining relatively independent ligament structures. Angle lines were constructed for the ACL and posterior cruciate ligaments using the angle line principle to clarify their orientation and anatomical position in three-dimensional space. In the initial identification of ligament contours, a rough contour was quickly generated using region growing technology. Then, through an iterative optimization process using an Active Contour Model (ACM), the rough contour boundaries were progressively refined to obtain a more accurate, continuous contour that conforms to the actual anatomical morphology. By extracting the intersection points of this precise contour with the contours of the femur and tibia, the insertion point of the cruciate ligament on the bone surface can be obtained. To ensure the matching relationship between soft tissue and bone tissue in the same coordinate system, this method further employs the ICP (Iterative Closest Point) registration algorithm to spatially align MRI and CT images, thereby accurately mapping the extracted ligament insertion point information onto the CT image to form an insertion point image. Based on this, surface rendering and 3D reconstruction operations are performed on the insertion point image and CT image data to finally generate a 3D knee joint model including the femur, tibia, and related soft tissues. This model not only includes the geometric morphology of the bones but also fully displays the precise location and structural features of tissues such as the cruciate ligament, providing a reliable 3D visualization basis for clinical diagnosis, surgical planning, and personalized treatment.

[0024] The working principle and process of this application are as follows: based on the fusion and 3D reconstruction technology of multimodal medical images, the knee joint bones and soft tissues are accurately presented in digital space. The entire process begins with obtaining clear data of the bone structure through high-resolution CT scans, while simultaneously using MRI scans to obtain high-contrast images of soft tissues such as the cruciate ligaments. All data is stored in DICOM format and imported into Mimics software, where the MRI images are binarized to clarify the boundaries of the femur and tibia, and a region of interest containing the cruciate ligaments is selected. This region is segmented to remove irrelevant soft tissues, retaining only the ligament structure to ensure the accuracy of subsequent modeling. After ligament segmentation, the orientation of the anterior and posterior cruciate ligaments is determined using the angle line principle, providing a reliable reference for anatomical localization. Next, a region growing method is used to obtain the preliminary contour of the ligaments, and iterative optimization is performed using an active contour model to ensure that the final ligament boundaries are highly consistent with the actual anatomy. The intersection points of the optimized ligament contours with the surfaces of the femur and tibia are calculated to extract the attachment (insertion) locations; this step is particularly crucial for surgical planning and kinematic studies. Precise registration of MRI and CT images across different imaging modalities requires the use of the ICP algorithm. This algorithm iteratively compares two sets of point cloud data, gradually reducing spatial errors, and ultimately maps the ligament insertion points obtained from MRI onto the CT bone image, thus achieving the fusion of soft tissue and bone in the same coordinate system. Combining the insertion point information and CT images, a 3D model is created using surface rendering methods, generating a mesh layer by layer to construct the overall model. The generated 3D model of the knee joint not only includes the skeletal structure but also fully presents key soft tissues such as ligaments, enabling doctors to intuitively analyze the anatomical relationships and pathological conditions of the knee joint.

[0025] As a preferred embodiment, the specific implementation of this application is as follows: A patient suffered a partial rupture of the anterior cruciate ligament (ACL) due to a sudden stop during basketball play. The doctor first performed a high-resolution CT scan to obtain the knee joint skeletal structure, and simultaneously performed an MRI scan to obtain ligament and soft tissue information, storing all images in DICOM format. This data was imported into Mimics software, where the MRI images were binarized to identify the femoral and tibial contours and determine the region of interest (ROI) containing the ACL. The K-Means algorithm and erosion processing were used to separate the ligament from the surrounding soft tissue, generating an image of the ligament severed from other soft tissues. Based on the principle of angle lines, the ACL was determined to have a 13° orientation, and the posterior cruciate ligament (PCL) to have 30° and 60° orientations, respectively, and a rough outline of the ligaments was obtained using region growing technology. Subsequently, an active contour model was used for iterative optimization to obtain precise ligament boundaries, and the intersections with the femoral and tibial contours were extracted to obtain the insertion positions of the anterior and posterior cruciate ligaments. To achieve spatial alignment of bones and ligaments in surgical planning, surgeons used an improved ICP algorithm to register MRI and CT images, mapping ligament insertion points extracted from MRI onto the CT images to construct insertion point images. A 3D model of the knee joint, including the femoral, tibial, and cruciate ligament insertion points, was then generated using a surface rendering 3D reconstruction method. This model clearly displays the bone structure and ligament attachment points, enabling surgeons to accurately assess the fracture site and ligament length in a virtual environment. This allows for personalized planning of anterior cruciate ligament reconstruction surgery, improving surgical precision and reducing intraoperative risks.

[0026] Through the aforementioned technical solution, this application utilizes the joint analysis of high-resolution CT and MRI images, combined with the principles of angle lines, region growing techniques, and active contour models, to accurately identify the insertion points of the anterior and posterior cruciate ligaments (ACLs), reducing manual annotation errors and improving the accuracy of insertion point extraction. By precisely registering MRI and CT images and performing surface rendering for 3D reconstruction, a complete 3D model of the knee joint, including the insertion points of the femur, tibia, and cruciate ligaments, can be generated, intuitively presenting the spatial relationship between bones and soft tissues, providing a reliable basis for clinical diagnosis and surgical planning. This method is highly automated, requiring only input of medical image data, eliminating the need for doctors to manually annotate ligament insertion points point by point, saving significant time and labor intensity, while also reducing the risk of errors caused by human operation. With the reconstructed precise 3D model, doctors can develop personalized ACL or posterior cruciate ligament reconstruction surgical plans based on the patient's knee joint anatomy and the specific locations of ligament attachment points, improving the safety and effectiveness of intraoperative procedures.

[0027] This application further proposes a method for acquiring binary images of the femur and tibia based on MRI images and determining the region of interest containing the cruciate ligament, including: The MRI images were binarized, and the binarized images were traversed to obtain the principal axis based on the shape features of the femur and tibia in the images. Contour detection was performed on the binarized images, and the contours containing points on the principal axis were traversed. The two contours with the largest areas were selected as the femoral and tibial contours, with the one above being the femoral contour and the one below being the tibial contour. Using the positional information of the cruciate ligaments on the femur and tibia, the rightmost point of the femur, the rightmost point of the tibia, the leftmost point of the tibia, and the principal axis were used as the boundary points of the cruciate ligaments of the human knee joint to determine the region of interest of the cruciate ligaments.

[0028] Specifically, in the process of acquiring binary images of the femur and tibia based on MRI images and determining the region of interest containing the cruciate ligament, the MRI images are first binarized to transform the grayscale information in the image into clear foreground and background regions. By setting an appropriate threshold, pixels above the threshold are marked as bone tissue, and pixels below the threshold are treated as background. This step significantly reduces the interference of soft tissue and noise on subsequent processing, while simplifying the image structure and facilitating accurate identification of the main bones of the knee joint. By traversing the binarized image and combining the morphological features of bone tissue in the image, the principal axis of the knee joint is extracted. This principal axis accurately reflects the alignment direction of the femur and tibia, serving as a reference benchmark for subsequent contour analysis and ligament localization. Based on the principal axis, a contour detection algorithm is used to identify all closed contours in the image, and the two largest contours are selected according to their area. Typically, the upper contour corresponds to the femur, and the lower contour corresponds to the tibia. This eliminates interference from small bone fragments or noise, ensuring that the selected contours represent the main skeletal structures and providing a stable basis for ligament localization. By combining the typical locations of the cruciate ligaments on the femur and tibia, the boundary points of the region of interest are further determined. Specifically, the rightmost point is extracted from the femoral contour as one of the boundary points, and the rightmost and leftmost points are extracted from the tibial contour as the left and right boundary points of the ligament on the tibia, respectively. These boundary points are then combined with the previously obtained principal axis to form a polygonal region surrounding the cruciate ligaments. This region not only completely covers the anterior and posterior cruciate ligaments but also effectively excludes most irrelevant soft tissue, thus providing accurate input for subsequent image segmentation, K-Means clustering, region growing, and active contour modeling.

[0029] As a preferred embodiment, the specific implementation of this application is as follows: When assessing anterior cruciate ligament (ACL) injury in the knee joint of a 30-year-old male patient, high-resolution MRI scans were first used to acquire image data of the patient's knee joint, which was saved in DICOM format. The MRI images were then imported into analysis software for processing. During processing, the MRI images were first binarized to separate bone tissue from soft tissue. The principal axis of the knee joint was extracted by traversing the binarized image and combining it with skeletal morphological features. The principal axis presents the approximate longitudinal arrangement of the femur and tibia. Closed contours in the image were identified using a contour discovery algorithm. After traversing the contours belonging to each point on the principal axis, the two contours with the largest areas were selected: the upper contour is the femoral contour, and the lower contour is the tibial contour. Based on the anatomical position of the ACL in the knee joint, the rightmost point of the femoral contour, the rightmost point of the tibial contour, and the leftmost point were determined, and combined with the principal axis to form a polygonal region surrounding the ligament. Through this region of interest, the software successfully located the area of ​​the ACL and posterior cruciate ligament, providing accurate input for subsequent region growth, active contour model optimization, and insertion point extraction.

[0030] Through the above technical solution, this application reduces the workload of the method, enabling automated and accurate extraction of the femur and tibia contours. By traversing the principal axis and judging the contour area, it accurately distinguishes the superior femur and inferior tibia, avoiding the uncertainty and subjective error of manual annotation. By determining the rightmost point of the femur, the rightmost point and the leftmost point of the tibia, and the principal axis as boundary points, the region of interest containing the cruciate ligament can be accurately located, covering the entire range of the ligament while excluding soft tissue areas unrelated to the ligament. This provides a reliable foundation for subsequent image processing steps, including region growth to obtain the coarse contour of the ligament, active contour model optimization, and insertion point extraction. Due to the improved accuracy and completeness of the region of interest, the accuracy and stability of subsequent calculations are significantly improved, making the 3D reconstruction and insertion point localization results more accurate and reliable. This reduces the workload of clinicians in knee ligament analysis. Traditional methods usually rely on manual judgment of ligament position and contour, which is easily affected by differences in operational experience and visual judgment. This method automates the processing, improves work efficiency, and reduces deviations caused by human error.

[0031] This application further proposes processing the region of interest to separate ligament images, and when obtaining images of ligaments that are disconnected from other soft tissues, including: The obtained region of interest was divided into four intervals using the K-Means algorithm. Pixels in the same interval were represented by the same color to obtain a pseudo-color image of the region of interest. The gray value region of the cruciate ligament was extracted and binarized. The ligament was then separated from other soft tissues using an erosion operation. The erosion operation was performed using a 3×3 structuring element for 3 iterations to obtain an image of the ligament separated from other soft tissues.

[0032] Specifically, the identified region of interest (ROI) of the cruciate ligament is input into the image processing module, and the K-Means clustering algorithm is applied to divide the pixel grayscale values ​​of this region. Typically, the pixels are divided into four intervals, with pixels within each interval having similar grayscale characteristics, thus enabling preliminary differentiation of different tissue types. After segmentation, for ease of visualization and subsequent processing, pixels in the same interval are represented by the same color, generating a pseudo-color image. This pseudo-color image visually displays the grayscale differences between the ligament and surrounding soft tissues, providing a basis for subsequent separation operations. Based on the grayscale characteristics of the human cruciate ligament in the pseudo-color image, pixels in the ligament region are extracted, and this region is binarized. Binarization marks the ligament region as the foreground with a value of 1, while other soft tissues and the background are marked as 0, thus forming a preliminary ligament image. To further completely separate the ligament from adjacent soft tissues, morphological erosion is required. Specifically, a 3×3 structuring element (usually a square or cross shape) is selected, and the binarized ligament image is subjected to three iterative erosion processes. The erosion operation breaks the connection between the ligament and adjacent tissues by cutting off the pixels at the foreground edge, removing small noise and redundant connection areas, thereby obtaining an image of the ligament that is completely disconnected from other soft tissues.

[0033] As a preferred embodiment, the specific implementation of this application is as follows: When processing the MRI image of the right knee joint of a 30-year-old male patient, the region of interest containing the cruciate ligament is first selected from the MRI image. The region is divided into four grayscale intervals using the K-Means algorithm, and a different color is assigned to each interval. The generated pseudo-color image clearly shows the distribution of different tissues such as ligaments, synovial fluid, cartilage, and tendons. The anterior cruciate ligament region is extracted based on the grayscale features of the ligament and binarized, marking the ligament as the foreground and other soft tissues and background as the background. To further separate the ligament from adjacent tissues, a 3×3 structuring element is used for three erosion operations, successfully severing the ligament from the surrounding articular cartilage and tendons, resulting in a separate ligament image. The processed image shows that the anterior cruciate ligament is clear and intact, without connection to surrounding tissues, providing an accurate data foundation for subsequent coarse contour extraction, active contour optimization, and insertion point localization.

[0034] Through the aforementioned technical solutions, this application effectively distinguishes ligaments from synovial fluid, tendons, cartilage, and other tissues using pseudo-color imaging and binarization processing, thereby improving the identifiability of ligament regions. Erosion operations remove interfering tissues connected to the ligaments, making subsequent coarse and fine contour extraction more accurate. Automated separation of ligaments from other soft tissues reduces the workload and subjective errors of manual annotation by physicians. Independent ligament images provide clean and accurate input for algorithms such as region growing and active contour models, improving the accuracy of endpoint localization and 3D reconstruction.

[0035] This application further proposes methods for obtaining the anterior cruciate ligament (ACL) and posterior cruciate ligament (PCL) angle lines based on the angle line principle, including: Based on the principle that the anterior cruciate ligament (ACL) and the intercondylar fossa of the femur form a 13° angle, the rightmost point of each row is traversed downwards from the rightmost point of the femur until the principal axis is encountered, thus obtaining the points on the intercondylar fossa of the femur. A straight line is fitted to these points using the least squares method. The ACL angle line is obtained by rotating counterclockwise by 13° with the rightmost point of the femur as the center point. Straight lines at 30° and 60° to the horizontal direction are constructed respectively, and the obtained ligament images that are disconnected from other soft tissues are traversed. When the two straight lines contain the most ligament pixels respectively, the posterior cruciate ligament (PCL) angle line is obtained.

[0036] Specifically, in obtaining the angle lines of the anterior cruciate ligament (ACL) and posterior cruciate ligament (PCL) based on the angle line principle, the entire process fully integrates knee joint anatomy features and image processing technology to achieve precise localization of the ligament's course. For the extraction of the ACL angle line, key points in the intercondylar fossa of the femur need to be determined. This process starts from the rightmost point of the femur and traverses the rightmost pixel of each row vertically downwards until the principal axis is encountered. In this way, the set of points along the upper edge of the intercondylar fossa of the femur can be accurately extracted, and these points can realistically reflect the geometry of the femur. Subsequently, a straight line is fitted to these points using the least squares method to obtain the direction of the center line of the intercondylar fossa of the femur. This fitted straight line provides a stable reference direction, reducing the influence of noise and image artifacts on ligament localization. Using the rightmost point of the femur as the center of rotation, the fitted straight line is rotated counterclockwise by 13°. According to anatomical studies, this angle best approximates the natural course of the ACL on the femur, thus obtaining the ACL angle line. This angle line not only provides a directional reference for the ligament in the 2D image but also provides an accurate starting point for subsequent region growing and coarse contour extraction. The extraction of the posterior cruciate ligament (PCL) angle line needs to consider the relatively complex orientation of the PCL. Two reference lines are constructed, forming angles of 30° and 60° with the horizontal direction, respectively. These angles cover the possible directional range of the PCL. In the previously separated images of ligaments disconnected from other soft tissues, the number of ligament pixels along these two lines is counted. By comparing the pixel coverage of the two lines, when the number of ligament pixels on a certain line reaches its maximum, that line is considered to best match the actual orientation of the PCL, thus obtaining the PCL angle line. This method fully utilizes pixel distribution information, avoids the uncertainty of manual annotation, and improves the automation and accuracy of the extraction process. The role of the entire angle line extraction process is to provide clear directional guidance for subsequent region growing and coarse contour construction, enabling the precise positioning and separation of the ligament's 2D contour. Simultaneously, this method also lays the foundation for 3D reconstruction, as the accurate orientation lines of the anterior and posterior cruciate ligaments ensure the reliability of subsequent fine contour extraction and endpoint positioning. For example, when processing a patient's knee MRI image, the combined use of the 13° angle line of the anterior cruciate ligament and the 30° and 60° reference lines of the posterior cruciate ligament not only successfully determined the initial course of the ligament, but also significantly reduced the errors caused by image noise or soft tissue interference, providing high-precision data support for 3D modeling and clinical surgical planning.

[0037] As a preferred embodiment, the specific implementation of this application is as follows: When analyzing the knee MRI image of a 30-year-old male patient, the rightmost point of the femur is located, and the image is scanned vertically downwards from this point, recording the rightmost pixel of each row until the femoral axis is encountered, resulting in a series of points on the intercondylar fossa of the femur. A line is fitted to these points using the least squares method to obtain the center line of the intercondylar fossa. Using the rightmost point of the femur as the center of rotation, the fitted line is rotated counterclockwise by 13° to obtain the angle line of the anterior cruciate ligament. In the same MRI slice, to extract the angle line of the posterior cruciate ligament, two reference lines are constructed, at 30° and 60° to the horizontal direction, respectively. The number of ligament pixels contained in each line is counted in the separated ligament images. It is found that the line at 60° covers the most ligament pixels; therefore, this line is determined as the angle line of the posterior cruciate ligament.

[0038] Through the above technical solution, this application uses the rightmost pixel of each row downwards from the rightmost point of the femur, and combines the anatomical features of the intercondylar fossa of the femur to fit the baseline straight line of the anterior cruciate ligament using the least squares method. Based on this, the anterior cruciate ligament angle line is obtained by rotating it counterclockwise by 13°, which can accurately reflect the spatial orientation of the anterior cruciate ligament on the femur. By constructing two reference straight lines at 30° and 60° to the horizontal direction, and counting the number of ligament pixels covered by each line, the line covering the most ligament pixels is selected as the posterior cruciate ligament angle line, which can accurately reflect the spatial orientation of the posterior cruciate ligament.

[0039] This application further proposes a method for obtaining a coarse outline of the cruciate ligament using region growing technology, and obtaining a precise outline through iterative optimization using an active contour model, including: Starting from the rightmost point of the femur, the search proceeds downwards, using the first ligament point encountered as the seed point. Region growth technology is used to obtain the coarse outline of the anterior cruciate ligament (ACL). Similarly, starting from the intersection of the two angle lines of the posterior cruciate ligament (PCL), the search proceeds upwards or downwards, using the first ligament point encountered as the seed point. Region growth technology is used to obtain the coarse outline of the PCL, thus obtaining the coarse outline of the cruciate ligament (CXL). Using the coarse outline of the CXL as the initial outline, an active outline model is used for iteration. The initial outline deforms under the influence of internal and external forces. External energy attracts the active outline towards the ligament edge, while internal energy maintains the smoothness and topology of the active outline. When the energy function reaches its minimum, the active outline converges to the ligament edge, obtaining the precise outline of the CXL.

[0040] Specifically, obtaining a precise outline of the cruciate ligament (ACL) is a crucial step. The core of this process involves first extracting a coarse outline using region growing techniques, followed by iterative optimization using an Active Contour Model (ACM) to obtain highly accurate ligament edge information. In the coarse outline extraction stage, starting from the rightmost point of the femur, the image is traversed line by line from top to bottom. The first pixel encountered that belongs to the ligament is selected as the seed point for region growing. The region growing technique uses this seed point as the center and, based on pixel grayscale similarity, gradually adds surrounding pixels that meet the criteria to the growing region, thus forming the coarse outline of the anterior cruciate ligament (ACL). For the posterior cruciate ligament (PCL), the intersection of two angle lines obtained from the previous angle line analysis is used as the starting point. Traversing upwards or downwards, the region is expanded similarly after encountering the first ligament pixel, ultimately obtaining the coarse outline of the entire cruciate ligament. In this way, the coarse outline can largely cover the spatial distribution of the ligament, providing initial conditions for the next optimization step, while eliminating most interference from non-ligamentous tissues. In the precise contour optimization stage, the coarse contour is used as the initial contour of the active contour model for iterative optimization. During the iteration process, the contour is affected by both internal and external energy: internal energy constrains the smoothness and continuity of the contour, preventing excessive bending or breakage; external energy is driven by image gradients or grayscale changes, causing the contour to approach the actual ligament edge. As iteration progresses, the contour continuously adjusts its shape, gradually approaching the actual boundary of the ligament. When the energy function converges to its minimum value, the contour reaches a stable state, thus obtaining the precise contour of the cruciate ligament.

[0041] As a preferred embodiment, the specific implementation of this application is as follows: When performing three-dimensional reconstruction of knee joint MRI images, the rightmost point of the femur is determined as the starting point through preprocessing. Scanning the image line by line from top to bottom, the first pixel belonging to the anterior cruciate ligament (ACL) with a grayscale value of 120 is identified and used as the seed point for region growing. Using region growing technology, pixels with grayscale values ​​within ±10 are gradually added to the growing region to extract the coarse outline of the ACL, covering the general shape of the ligament. Using the intersection of two angle lines of the posterior cruciate ligament (PCL) as the starting point, scanning along the up and down directions to the first ligament pixel with a grayscale value of 115 is used as the seed point for the PCL. Similarly, region growing technology is applied to obtain the coarse outline of the PCL. Combining the coarse outlines of the anterior and posterior cruciate ligaments, a complete coarse outline of the cruciate ligament is formed. Based on the coarse outline, an active contour model (ACM) is used for iterative optimization. Internal energy weights are set to ensure smooth contours, while external energy, drawn by the image gradient, attracts the contour towards the ligament edge. After 30 iterations, the active contour gradually converged and finally precisely matched the edge of the ligament, extracting the actual contours of the ACL and PCL for subsequent 3D reconstruction and surgical planning.

[0042] Through the aforementioned technical solution, this application, by combining region growing with an active contour model, can accurately depict the morphology and edge details of the cruciate ligament, achieving higher accuracy compared to simple threshold segmentation or manual annotation. It can distinguish the ligament from adjacent soft tissues and bone structures, reducing missegmentation and noise effects, and improving the reliability of contour extraction. Relying on the principles of local seed points and angle lines, it can adapt to individual differences in different knee joint morphologies, realizing personalized ligament contour extraction.

[0043] This application further proposes methods for extracting the intersection of the precise contour with the contours of the femur and tibia as the insertion point of the cruciate ligament, including: Curve fitting is performed on the precise contour to obtain a smooth cruciate ligament contour curve; the intersection point is determined by calculating the minimum Euclidean distance between the cruciate ligament contour curve and the femoral and tibial contour lines; the intersection point is precisely located at the subpixel level and used as the insertion point of the cruciate ligament.

[0044] Specifically, in the process of extracting the intersection points of the precise cruciate ligament contour with the femoral and tibial contours as insertion points, curve fitting is performed on the obtained precise cruciate ligament contour to generate a smooth and continuous contour curve. This step effectively eliminates minor irregularities caused by noise or image resolution limitations during contour extraction, ensuring that the contour curve reflects the true geometric shape of the ligament. Using geometric analysis methods, the minimum Euclidean distance between the fitted ligament contour curve and the femoral and tibial contours is calculated to identify the intersection point on the contour curve that is closest to the bone contour. This intersection point is considered the anatomical insertion point of the ligament and bone, reflecting the attachment position of the ligament. To improve positioning accuracy, sub-pixel-level precise positioning processing is performed on the determined intersection point. This step refines the point coordinates on the contour curve through interpolation methods, allowing the intersection point to exceed the original image resolution limitations and achieve higher-precision spatial positioning. The obtained intersection point serves as the insertion points of the anterior cruciate ligament and posterior cruciate ligament on the femur and tibia, providing reliable anatomical basis data for subsequent knee joint 3D reconstruction, biomechanical analysis, and surgical planning.

[0045] As a preferred embodiment, the specific implementation of this application is as follows: After performing MRI and CT scans on the knee joint of an adult patient, the precise contour curve of the cruciate ligament is obtained. The contour is smoothed using a curve fitting method to obtain a continuous and smooth ligament contour curve. Subsequently, the minimum Euclidean distance between this contour curve and the femoral and tibial contour lines is calculated. It is found that the closest point of the anterior cruciate ligament to the femoral contour is located approximately 12.3 mm above the medial condyle of the femur, and the closest point to the tibia is located approximately 8.7 mm from the center of the anterior tibial crest; the posterior cruciate ligament intersects at approximately 14.1 mm below the lateral condyle of the femur and approximately 9.5 mm from the posterior tibial crest. To further improve accuracy, subpixel-level interpolation is used to fine-tune the intersection points, refining the coordinates to three decimal places. Finally, these four determined intersection points are used as the insertion points of the anterior and posterior cruciate ligaments on the femur and tibia, respectively, for subsequent 3D modeling and surgical navigation.

[0046] Through the above technical solution, this application obtains a continuous and smooth ligament contour curve by curve fitting the precise contour, avoiding contour irregularities caused by image noise or discrete points. The intersection points of the ligament contour with the femoral and tibial contours are precisely located using minimum Euclidean distance calculation, making the insertion point position more consistent with the actual anatomical structure. Sub-pixel-level precise positioning can accurately locate the intersection point coordinates to three decimal places, improving the spatial positioning accuracy of the insertion point.

[0047] This application further proposes a method for registering MRI and CT images based on the ICP algorithm, mapping the endpoint to the CT image, and constructing the endpoint image, including: The CT images are binarized. Contour detection is used to identify the two contours with the largest area in the CT image. The upper contour is designated as the femur, and the lower contour as the tibia. Based on the ICP algorithm, the distance from each contour point in the CT image to the contour in the MRI image is calculated for both the femur and tibia. The sum of these distances is then selected as the CT image corresponding to the smallest sum of distances. The contours of the femur and tibia in the MRI image are registered with their corresponding CT images using a two-dimensional affine transformation registration function, resulting in two rotation and translation matrices. The insertion point information of the femur and tibia is then transformed according to the registered rotation and translation matrices to create a pure background binary image of the same size as the CT image. The insertion points are then placed at their corresponding positions in the CT image to obtain the insertion point image.

[0048] Specifically, in the process of registering MRI and CT images and mapping stop points to construct stop point images based on the ICP algorithm, the CT image is binarized to highlight the skeletal structure and remove background interference. A contour discovery algorithm identifies the two largest contours in the CT image, and based on their spatial location, the upper contour is identified as the femur and the lower contour as the tibia, providing basic geometric information for subsequent registration. The Iterative Closest Point (ICP) algorithm is used to match the corresponding points of the femur and tibia contours in the MRI image with the contours in the CT image. Specifically, this involves calculating the distance from each contour point in the CT image to the contour in the MRI image, summing these distances, and selecting the CT image with the smallest sum as the optimal registration target for the current MRI image to ensure accurate spatial alignment. For the femur and tibia contours, a two-dimensional affine transformation function is used to calculate the rotation and translation matrices respectively, achieving accurate mapping from the MRI contours to the CT contours. The stop point information of the femur and tibia extracted from the MRI image is transformed using the rotation and translation matrices obtained from registration to ensure the accurate position of the stop points in the CT coordinate system. Generate a pure background binary image with the same size as the CT image, and embed the transformed stop point into the corresponding position to form a complete stop point image.

[0049] As a preferred embodiment, the specific implementation of this application is as follows: Knee joint imaging data of a male patient is obtained through high-resolution CT and MRI scans, respectively, and stored in DICOM format. When processing the CT images, after binarization, a contour discovery algorithm identifies the two largest contours, with the upper contour labeled as the femur and the lower contour as the tibia. The femur and tibia contours in the patient's MRI image are registered with the CT image contours using the ICP algorithm. Specifically, the Euclidean distance from each point of the femur contour in the CT image to the femur contour in the MRI image is calculated and summed; similarly, the sum of distances to the tibia contour is calculated. The CT image with the smallest sum of distances is selected as the optimal registration result. A two-dimensional affine transformation function is used to map the MRI contours to the CT image coordinate system, obtaining rotation and translation matrices. The femur and tibia insertion points extracted from the MRI are then transformed according to these matrices. A pure background binary image of the same size as the CT image is generated, and the transformed insertion points are placed in their corresponding positions, forming the insertion point image of the patient's knee joint.

[0050] Through the above technical solution, this application achieves precise alignment between MRI and CT by calculating the minimum sum of distances from contour points in CT images to MRI contours, thus ensuring consistency of femoral and tibial contours in both modalities. By using a two-dimensional affine transformation matrix to map the femoral and tibial insertion information extracted from MRI to the CT image coordinate system, combined with sub-pixel-level precise positioning, high accuracy and stability of the cruciate ligament insertion position can be ensured, which is helpful for subsequent 3D model construction and surgical navigation. The generated insertion image is the same size as the CT image, and the insertion point is located at a precise position in the CT image, providing accurate anatomical markers for knee joint 3D reconstruction and achieving a unified model of bone and soft tissue.

[0051] This application further proposes a method for generating a 3D model of the bones and soft tissues surrounding the knee joint by performing surface rendering and 3D reconstruction on the endpoint image and CT image, including: The Marching Cubes algorithm was used to render the CT images to generate a 3D model of the knee joint bones. 3D interpolation and surface fitting were performed on the insertion point images to generate a 3D model of the cruciate ligament insertion point. The bone model, the cruciate ligament insertion point model, and other soft tissue models were spatially registered and fused to generate a complete 3D model of the bones and soft tissues surrounding the knee joint.

[0052] Specifically, the Marching Cubes algorithm is applied to CT images. By traversing voxel data, voxels with grayscale values ​​exceeding a set threshold in the bone structure are meshed to generate a continuous 3D bone surface model. This step accurately reconstructs the spatial morphology of the knee joint bones, including key anatomical structures such as the femoral condyle, tibial plateau, and patellar contour. The cruciate ligament insertion information extracted from the insertion images undergoes 3D interpolation, and a 3D model of the ligament insertion is generated through surface fitting, ensuring the accurate position and morphological continuity of the insertion in space. This step not only preserves the anatomical accuracy of the insertion but also provides a reliable spatial reference for ligament path modeling. After the bone model and ligament insertion model are generated, they are spatially registered and fused with other soft tissue models (such as the meniscus, joint capsule, and surrounding ligaments). A unified coordinate system and optimization algorithms eliminate deviations between different models, achieving complete 3D reconstruction of the tissues surrounding the knee joint. The final 3D model includes not only an accurate bone surface but also the spatial relationships of key ligament insertions and related soft tissues.

[0053] As a preferred embodiment, the specific implementation of this application is as follows: A 30-year-old male knee patient underwent high-resolution CT and MRI scans. The CT images were imported into 3D reconstruction software, and the Marching Cubes algorithm was applied to voxels with gray values ​​higher than the bone density threshold to generate 3D skeletal models of the femur, tibia, and patella. The anterior and posterior cruciate ligament (ACL) insertion information extracted from the MRI images was subjected to 3D interpolation and surface fitting to generate ligament insertion models, which were then marked at their corresponding positions on the femoral condyle and tibial plateau. The skeletal model and the ligament insertion model were spatially registered to ensure the precise location of the insertion points on the bones. Simultaneously, soft tissue models such as the meniscus and joint capsule were fused with the skeletal model to form a complete 3D model of the knee joint. The surgeon can visually observe the positional relationship of the femur, tibia, and patella, as well as the precise distribution of the ACL insertion points on the bones, providing a precise 3D anatomical reference for preoperative planning of ligament reconstruction surgery or knee replacement. Through the aforementioned technical solution, this application employs the Marching Cubes algorithm to render surfaces in CT images, accurately reconstructing the three-dimensional structure of the knee joint bones, including details such as the femur, tibia, patella, and articular processes, providing a clear, rotatable skeletal model. By performing three-dimensional interpolation and surface fitting on the insertion point images, the generated cruciate ligament insertion point model can display the attachment position of the ligament on the bone. After spatial registration and fusion of the skeletal model, ligament insertion point model, and other soft tissue models, the resulting complete three-dimensional knee joint model can simultaneously present the spatial relationships of bones, ligaments, and soft tissues, providing a reliable three-dimensional anatomical basis.

[0054] This application further proposes methods for spatial registration and fusion of skeletal models, cruciate ligament insertion models, and other soft tissue models, including: A unified world coordinate system was established to transform the skeletal model, cruciate ligament insertion model, and soft tissue model into the same coordinate space. The lateral epicondyle of the femur, medial epicondyle, the edge of the tibial plateau, and the contour of the patella were selected as registration reference points to calculate the initial correspondence between the skeletal and soft tissue models. Based on an improved non-rigid registration algorithm, the soft tissue model was locally deformed and adjusted using the skeletal model as the reference, ensuring that the average distance between the key anatomical landmarks of the soft tissue model and their corresponding points on the skeletal model was less than 0.5 mm. A feature-based registration method was used to precisely align the cruciate ligament insertion model with the skeletal model. The spatial position of the ligament insertion model was adjusted by calculating the geometric relationship between the ligament insertion and the anatomical landmarks of the intercondylar fossa of the femur and the tibial plateau. Boolean operations and boundary smoothing were performed on the registered models to create a transition region at the bone-ligament connection, eliminating gaps and overlaps between models. A physics-based fusion algorithm was used to elastically deform the fusion region according to the biomechanical characteristics of the knee joint anatomy, ensuring a natural connection between the cruciate ligament insertion and the bone surface.

[0055] Specifically, in the process of spatial registration and fusion of skeletal models, cruciate ligament insertion models, and other soft tissue models, a unified world coordinate system needs to be established to transform all models into the same coordinate space, ensuring spatial consistency and accuracy in subsequent operations. The initial registration between the skeletal and soft tissue models relies on the selection of key anatomical landmarks, such as the lateral epicondyle and medial epicondyle of the femur, the edge of the tibial plateau, and the patellar contour. These points are used to establish the initial correspondence between the models, ensuring that the alignment positions of each model are basically consistent during the coarse registration stage. After coarse registration, an improved non-rigid registration algorithm is used, with the skeletal model as the reference, to perform local deformation adjustments on the soft tissue model. During this process, the average distance between key anatomical landmarks is optimized to be less than 0.5 mm, thereby ensuring that the soft tissue model can closely conform to the bone surface in local areas, maintaining anatomical realism. For the cruciate ligament (ACL) insertion model, a feature-based precise registration method is used. This involves calculating the geometric relationship between the insertion point and anatomical landmarks such as the femoral intercondylar fossa and tibial plateau, adjusting the spatial position of the insertion model to achieve precise anatomical alignment with the skeletal model. After registration, the models undergo fusion processing. Boolean operations and boundary smoothing techniques are used to create a transition region at the bone-ligament junction, effectively eliminating gaps and overlaps between models and ensuring structural continuity and surface smoothness. A physics-based fusion algorithm elastically deforms the fusion area according to the biomechanical characteristics of the knee joint anatomy, allowing the ACL insertion point to naturally attach to the bone surface without causing non-physiological stretching or displacement.

[0056] As a preferred embodiment, the specific implementation of this application is as follows: During preoperative digital simulation of the knee joint in a 30-year-old male patient, high-resolution CT and MRI scan data were acquired, and three-dimensional models of the femur, tibia, cruciate ligament insertion points, and surrounding soft tissues were constructed. To ensure spatial consistency between models, all models were transformed into the same world coordinate system. The lateral epicondyle of the femur, the medial epicondyle, the edge of the tibial plateau, and the contour of the patella were selected as reference points to calculate the initial correspondence between the skeletal model and the soft tissue model, achieving rough alignment. An improved non-rigid registration algorithm was used to locally deform the soft tissue model, ensuring that the average distance between its key anatomical landmarks (such as the intercondylar fossa of the femur and the soft tissue attachment points of the tibial plateau) and the corresponding points of the skeletal model was controlled within 0.5 mm, guaranteeing that the soft tissue closely adhered to the bone surface. For the cruciate ligament (ACL) insertion model, a feature registration method was used to calculate the geometric relationship between the insertion point and key anatomical landmarks of the femoral intercondylar fossa and tibial plateau, precisely adjusting the spatial position of the insertion point to ensure its position on the bone matches the actual anatomical location. After registration, Boolean operations and boundary smoothing were performed on the model to create transition regions at the femoral-ligament and tibial-ligament junctions, effectively eliminating gaps and local overlaps between models. A physically-based elastic fusion algorithm was then used to fine-tune the fusion region, allowing the ACL insertion point to naturally conform to the bone surface, forming a complete and continuous three-dimensional model of the knee joint.

[0057] Through the above technical solution, this application establishes a unified coordinate system to transform the skeletal model, cruciate ligament insertion model, and soft tissue model into the same space; selects the femoral condyle, tibial plateau edge, and patellar contour as reference points to achieve initial alignment; uses an improved non-rigid registration algorithm to adjust the soft tissue model so that the average distance between key anatomical points and corresponding bone points is less than 0.5 mm; accurately aligns the ligament insertion model through feature registration and calculates its geometric relationship with skeletal anatomical landmarks; performs Boolean operations and boundary smoothing on the registered model to create a transition region at the bone-ligament connection; and adjusts the fusion region using a physics-based elastic fusion algorithm to achieve a natural connection between the ligament insertion and the bone surface, thereby obtaining an accurate and continuous three-dimensional model of the knee joint.

[0058] In another preferred embodiment based on the above embodiments, see [reference] Figure 2 As shown, this embodiment provides a three-dimensional reconstruction system for the knee joint, used to apply the above-described three-dimensional reconstruction method for the knee joint, including: The acquisition unit is configured to acquire medical image data of the human knee joint through high-resolution CT and MRI scans and store the data in DICOM format; and import the DICOM format data into Mimics software. The extraction unit is configured in Mimics software to acquire binary images of the femur and tibia based on MRI images and determine the region of interest containing the cruciate ligament; process the region of interest to separate the ligament image and acquire images of the ligament that is disconnected from other soft tissues; and acquire the angle lines of the anterior cruciate ligament and the posterior cruciate ligament based on the angle line principle. The processing unit is configured to obtain the coarse outline of the cruciate ligament using region growing technology and obtain the precise outline through active contour model iterative optimization; extract the intersection point of the precise outline with the femoral and tibial contour lines as the insertion point of the cruciate ligament; register the MRI image and CT image based on the ICP algorithm, map the insertion point into the CT image, and construct the insertion point image. The building unit is configured to perform surface rendering and 3D reconstruction on the stop point image and CT image to generate a 3D model of the bones and soft tissues around the knee joint. The 3D model includes a bone model and a tissue and organ model.

[0059] In summary, by combining MRI and CT images, and fully utilizing the soft tissue resolution of MRI and the bone precision of CT, a high-precision 3D model of the bones and soft tissues surrounding the knee joint can be generated. Accurate contour extraction of the cruciate ligament is achieved using region growing, active contour modeling, and the angle line principle, followed by sub-pixel-level stop point localization, improving the accuracy of stop point identification. Precise registration of MRI and CT images is achieved based on the ICP algorithm and 2D affine transformation; improved non-rigid registration and feature-based registration methods are employed to achieve sub-millimeter-level accuracy in the correspondence between soft tissue and bone models. The bone, ligament, and soft tissue models are unified to the same coordinate system, and a natural transition connection between bones and ligaments is achieved through Boolean operations, boundary smoothing, and elastic deformation processing; this solves the problem of gaps or overlaps between bones, ligaments, and soft tissues in traditional models.

[0060] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0061] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0062] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0063] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0064] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A three-dimensional reconstruction method for the knee joint, characterized in that, include: Medical image data of the human knee joint is acquired through high-resolution CT and MRI scans and stored in DICOM format. Import the DICOM format data into Mimics software; In the Mimics software, binary images of the femur and tibia are acquired based on MRI images, and a region of interest containing the cruciate ligament is determined; the region of interest is processed to separate the ligament image and obtain images of the ligament that is disconnected from other soft tissues; the angle lines of the anterior cruciate ligament and the posterior cruciate ligament are obtained based on the angle line principle. The coarse outline of the cruciate ligament is obtained using region growing technology, and the precise outline is obtained through iterative optimization using an active contour model. The intersection points of the precise outline with the femoral and tibial contours are extracted as the insertion points of the cruciate ligament. Based on the ICP algorithm, the MRI and CT images are registered, and the insertion points are mapped into the CT images to construct the insertion point images. The stop point image and CT image are used to perform surface rendering and three-dimensional reconstruction to generate a three-dimensional model of the bones and soft tissues around the knee joint. The three-dimensional model includes a bone model and a tissue and organ model.

2. The three-dimensional reconstruction method for the knee joint according to claim 1, characterized in that, When acquiring binary images of the femur and tibia based on MRI images and determining the region of interest containing the cruciate ligament, the following steps are included: The MRI image is binarized, and the binarized image is traversed to obtain the principal axis based on the shape features of the femur and tibia in the image. Contour detection is performed on the binarized image, and the contours containing points on the principal axis are traversed. The two contours with the largest areas are selected as the femoral and tibial contours, with the one above being the femoral contour and the one below being the tibial contour. Using the positional information of the cruciate ligament on the femur and tibia, the rightmost point of the femur, the rightmost point of the tibia, the leftmost point of the tibia, and the principal axis are used as the boundary points of the cruciate ligament of the human knee joint to determine the region of interest of the cruciate ligament.

3. The three-dimensional reconstruction method for the knee joint according to claim 2, characterized in that, Processing the region of interest to separate the ligament image, and obtaining an image of a ligament that has broken away from other soft tissues, includes: The obtained region of interest was divided into four intervals using the K-Means algorithm. Pixels in the same interval were represented by the same color to obtain a pseudo-color image of the region of interest. The gray value region of the cruciate ligament was extracted and binarized. The ligament was then separated from other soft tissues using an erosion operation. The erosion operation was performed using a 3×3 structuring element for 3 iterations to obtain an image of the ligament separated from other soft tissues.

4. The three-dimensional reconstruction method for the knee joint according to claim 3, characterized in that, When obtaining the anterior cruciate ligament (ACL) and posterior cruciate ligament (PCL) angle lines based on the angle line principle, the following is included: Based on the principle that the anterior cruciate ligament (ACL) and the intercondylar fossa of the femur are at a 13° angle, the rightmost point of each row is traversed downwards from the rightmost point of the femur until the principal axis is encountered, thus obtaining the points on the intercondylar fossa of the femur. A straight line is fitted to these points using the least squares method. The ACL angle line is obtained by rotating counterclockwise by 13° with the rightmost point of the femur as the center point. Straight lines at 30° and 60° to the horizontal direction are constructed respectively, and the obtained ligament images that are disconnected from other soft tissues are traversed. When the two straight lines contain the most ligament pixels respectively, the posterior cruciate ligament (PCL) angle line is obtained.

5. A three-dimensional reconstruction method for the knee joint according to claim 4, characterized in that, When using region growing techniques to obtain the coarse outline of the cruciate ligament and then iteratively optimizing it through an active contour model to obtain the precise outline, the process includes: Starting from the rightmost point of the femur, the search proceeds downwards, using the first ligament point encountered as the seed point. Region growth technology is used to obtain the coarse outline of the anterior cruciate ligament (ACL). Similarly, starting from the intersection of the two angle lines of the posterior cruciate ligament (PCL), the search proceeds upwards or downwards, using the first ligament point encountered as the seed point. Region growth technology is used to obtain the coarse outline of the PCL, thus obtaining the coarse outline of the cruciate ligament (CXL). Using the coarse outline of the CXL as the initial outline, an active outline model is used for iteration. The initial outline deforms under the influence of internal and external forces. External energy attracts the active outline towards the ligament edge, while internal energy maintains the smoothness and topology of the active outline. When the energy function reaches its minimum, the active outline converges to the ligament edge, obtaining the precise outline of the CXL.

6. A three-dimensional reconstruction method for the knee joint according to claim 5, characterized in that, When extracting the intersection point of the precise contour with the contour lines of the femur and tibia as the insertion point of the cruciate ligament, the following steps are included: Curve fitting is performed on the precise contour to obtain a smooth cruciate ligament contour curve; the intersection point is determined by calculating the minimum Euclidean distance between the cruciate ligament contour curve and the femoral and tibial contour lines; the intersection point is precisely located at the subpixel level and used as the insertion point of the cruciate ligament.

7. A three-dimensional reconstruction method for the knee joint according to claim 6, characterized in that, Based on the ICP algorithm, MRI images and CT images are registered, and the endpoints are mapped onto the CT images. The construction of the endpoint image includes: The CT image is binarized. Contour detection is used to obtain the two contours with the largest area in the CT image. The upper contour is taken as the femur, and the lower contour as the tibia. Based on the ICP algorithm, the distance from each contour point in the CT image to the contour in the MRI image is calculated for both the femur and tibia. The sum of the distance values ​​is then selected as the CT image corresponding to the smallest sum of distance values. The contours of the femur and tibia in the MRI image are registered with the corresponding femur and tibia obtained from the CT image using a two-dimensional affine transformation registration function, resulting in two rotation and translation matrices. The stop point information of the femur and tibia is transformed according to the rotation and translation matrices obtained from the registration, creating a pure background binary image of the same size as the CT image. The stop points are then placed at the corresponding positions in the CT image to obtain the stop point image.

8. A three-dimensional reconstruction method for the knee joint according to claim 7, characterized in that, When performing surface rendering and 3D reconstruction on the stop point image and CT image to generate a 3D model of the bones and soft tissues around the knee joint, the following steps are included: The Marching Cubes algorithm was used to render the CT images to generate a 3D model of the knee joint bones. 3D interpolation and surface fitting were performed on the insertion point images to generate a 3D model of the cruciate ligament insertion point. The bone model, the cruciate ligament insertion point model, and other soft tissue models were spatially registered and fused to generate a complete 3D model of the bones and soft tissues around the knee joint.

9. A three-dimensional reconstruction method for the knee joint according to claim 8, characterized in that, Spatial registration and fusion of skeletal models, cruciate ligament insertion models, and other soft tissue models includes: A unified world coordinate system was established to transform the skeletal model, cruciate ligament (ACL) insertion model, and soft tissue model into the same coordinate space. The lateral epicondyle of the femur, medial epicondyle, tibial plateau edge, and patellar contour were selected as registration reference points to calculate the initial correspondence between the skeletal and soft tissue models. Based on an improved non-rigid registration algorithm, the soft tissue model was locally deformed and adjusted using the skeletal model as the reference, ensuring that the average distance between key anatomical landmarks in the soft tissue model and their corresponding points in the skeletal model was less than 0.5 mm. A feature-based registration method was used to precisely align the ACL insertion model with the skeletal model. The spatial position of the ACL insertion model was adjusted by calculating the geometric relationship between the ligament insertion and the anatomical landmarks of the femoral intercondylar fossa and tibial plateau. Boolean operations and boundary smoothing were performed on the registered models to create a transition region at the bone-ligament connection, eliminating gaps and overlaps between models. A physics-based fusion algorithm was used to elastically deform the fusion region according to the biomechanical characteristics of the knee joint anatomy, ensuring a natural connection between the ACL insertion and the bone surface.

10. A three-dimensional reconstruction system for the knee joint, used for applying a three-dimensional reconstruction method for the knee joint as described in any one of claims 1-9, characterized in that, include: The acquisition unit is configured to acquire medical image data of the human knee joint through high-resolution CT and MRI scans and store the data in DICOM format. Import the DICOM format data into Mimics software; The extraction unit is configured in the Mimics software to acquire binary images of the femur and tibia based on MRI images and determine a region of interest containing the cruciate ligament; process the region of interest to separate the ligament image and acquire images of the ligament that is disconnected from other soft tissues; and acquire the angle lines of the anterior cruciate ligament and the posterior cruciate ligament based on the angle line principle. The processing unit is configured to obtain the coarse outline of the cruciate ligament using region growing technology, and obtain the precise outline through active contour model iterative optimization; extract the intersection point of the precise outline with the femoral and tibial contour lines as the insertion point of the cruciate ligament; register the MRI image and CT image based on the ICP algorithm, and map the insertion point into the CT image to construct the insertion point image; The construction unit is configured to perform surface rendering and three-dimensional reconstruction on the stop point image and CT image to generate a three-dimensional model of the bones and soft tissues around the knee joint, the three-dimensional model including a bone model and a tissue and organ model.

Images