Artificial-intelligence preoperative planning system for multiple orthopedic disease types

Abstract

The present application provides an artificial-intelligence preoperative planning system for multiple orthopedic disease types, the system comprising: an import module, configured to import medical image data; a three-dimensional reconstruction module, configured to establish a corresponding three-dimensional image model on the basis of the medical image data; a bone segmentation module, configured to segment bones in the medical image data; and at least three of the following modules: a hip joint simulation module, a knee joint simulation module, a spinal joint simulation module, a sports medicine simulation module, and a trauma surgery simulation module. In the present disclosure, by providing multiple surgical simulation modules respectively corresponding to different types of orthopedic surgery, the coverage scope of the preoperative planning system is significantly increased.

Description

A comprehensive orthopedic multi-disease artificial intelligence preoperative planning system Technical Field

[0001] This application relates to the field of medical device technology, and more specifically, to an artificial intelligence preoperative planning system for multiple diseases in orthopedics. Background Technology

[0002] Traditional orthopedic surgery planning is done independently by the surgeon, which relies too heavily on the surgeon's experience. This may increase surgical risks and result in low precision and stability of the surgery.

[0003] Therefore, preoperative planning systems using computer software can assist doctors in preoperative planning. However, current preoperative planning systems are only designed for one type of orthopedic surgery and have a limited scope of application. Summary of the Invention

[0004] The problem addressed by this application is that existing preoperative planning systems are only designed for one category of orthopedic surgery.

[0005] To address the aforementioned issues, this application provides a comprehensive orthopedic multi-disease artificial intelligence preoperative planning system, comprising:

[0006] The import module is configured to import medical imaging data;

[0007] The 3D reconstruction module is configured to build a corresponding 3D image model based on medical imaging data;

[0008] The skeleton segmentation module is configured to segment the skeleton in medical image data;

[0009] And at least the following three modules:

[0010] The hip joint simulation module is configured to simulate the hip joint surgery process and determine the preoperative planning scheme for hip joint surgery based on medical imaging data of the hip joint.

[0011] The knee joint simulation module is configured to simulate the knee joint surgery process and determine the preoperative planning scheme for knee joint surgery based on medical imaging data of the knee joint.

[0012] The spinal joint simulation module is configured to simulate the spinal joint surgery process and determine the preoperative planning scheme for spinal joint surgery based on medical imaging data of the spinal joint.

[0013] The sports medicine simulation module is configured to simulate the movement process of the knee joint based on medical imaging data of the knee joint;

[0014] The trauma surgery simulation module is configured to simulate the trauma surgery process and determine the preoperative planning scheme for trauma surgery based on medical imaging data of the trauma site.

[0015] Furthermore, the hip joint simulation module includes a first key point recognition submodule, a first template measurement submodule, a first display submodule, and a first guide plate submodule;

[0016] The first key point recognition submodule is configured to recognize key points of the femur and acetabulum;

[0017] The first module measurement submodule is configured to measure or calculate the anteversion angle, abduction angle, coverage of the acetabular cup prosthesis template, postoperative leg length difference, offset difference, and the amount and location of osteotomy in the proximal femur.

[0018] The first display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation;

[0019] The first guide plate submodule is configured to output the planned osteotomy amount and osteotomy location, and export the osteotomy guide plate model.

[0020] Furthermore, the key points identified by the first key point identification submodule include: the center of the left femoral head, the center of the femoral diameter, above the lesser trochanter, the lesser trochanter, the proximal end of the femur, the distal end of the femur, the medial epicondyle and the lateral epicondyle, the center of the right femoral head, above the lesser trochanter, the lesser trochanter, the proximal end of the femur, and the distal end of the femur.

[0021] Furthermore, the knee joint simulation module includes a second key point recognition submodule, a second template measurement submodule, a second display submodule, and a second guide plate submodule;

[0022] The second key point recognition submodule is configured to recognize key points on the femur and tibia;

[0023] The second module measurement submodule is configured to measure or calculate the distal femoral osteotomy amount, the posterior femoral condyle osteotomy amount, the tibial plateau osteotomy amount, and to calculate the sum of the medial distal femoral osteotomy amount and the medial tibial plateau osteotomy amount, the sum of the lateral distal femoral osteotomy amount and the lateral tibial plateau osteotomy amount, the sum of the medial posterior femoral condyle osteotomy amount and the medial tibial plateau osteotomy amount, and the sum of the lateral posterior femoral condyle osteotomy amount and the lateral tibial plateau osteotomy amount.

[0024] The second display submodule is configured to display an image of the prosthesis template fitted to the skeleton, and adjust it according to the operator's operation;

[0025] The second guide plate submodule is configured to output the planned osteotomy amount and osteotomy location, and export the osteotomy guide plate model.

[0026] Furthermore, the key points identified by the second key point identification submodule include: medial epicondyle, lateral epicondyle, distal medial, distal lateral, medial posterior condyle, lateral posterior condyle, anterior condyle, proximal femoral end and distal femoral end, medial tibial plateau point, lateral tibial plateau point, medial 1 / 3 of the tibial tuberosity, posterior cruciate ligament insertion point, proximal tibial end and distal tibial end.

[0027] Furthermore, the spinal joint simulation module includes a third key point recognition submodule, a third template measurement submodule, a third display submodule, and a third guide plate submodule;

[0028] The third key point recognition submodule is configured to recognize key points of each vertebra of the spine.

[0029] The third module measurement submodule is configured to measure or calculate the dimensions and setting positions of the screw module, fusion template, titanium rod template and cross-connecting template;

[0030] The third display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation;

[0031] The third guide plate submodule is configured to output the planned positions of the screw module, fusion template, titanium rod template, and transverse connecting template, and export the spinal guide plate model.

[0032] Furthermore, the sports medicine simulation module includes a fourth key point recognition submodule, a fourth template measurement submodule, and a fourth display submodule;

[0033] The fourth key point recognition submodule is configured to recognize key points of the femur and tibia;

[0034] The fourth module, the measurement submodule, is configured to measure or calculate the length and expansion / contraction changes of the bone tunnel at different flexion angles of the knee joint.

[0035] The fourth display submodule is configured to display the dynamic effects of the knee joint at different flexion angles and the changes in the bone tunnel.

[0036] Furthermore, the trauma surgery simulation module includes a fifth key point recognition submodule, a fifth template measurement submodule, a fifth display submodule, and a fifth guide plate submodule;

[0037] The fifth key point recognition submodule is configured to recognize key points of the femur and acetabulum;

[0038] The fifth module, the measurement submodule, is configured to measure or calculate the dimensions and placement of the main screw module, the tail cap template, the spiral blade template, and the interlocking screw template.

[0039] The fifth display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation;

[0040] The fifth guide plate submodule is configured to output the planned positions of the main nail module, tail cap template, spiral blade template, and interlocking screw template.

[0041] Furthermore, the knee joint simulation module also includes a unicompartmental submodule, which is connected to the second key point recognition submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule. When the unicompartmental submodule is active, the second key point recognition submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule only process operations corresponding to a single knee joint.

[0042] Furthermore, the bone segmentation module is equipped with a pre-trained segmentation model, which segments the input medical image data as follows:

[0043] The medical image data is input into the feature extraction structure to obtain a low-level feature map and multiple high-level feature maps;

[0044] The high-level feature map is input into the encoding structure to obtain the encoded feature map;

[0045] The encoded feature map and low-level feature map are input into the decoding structure to obtain the segmentation result.

[0046] In this application, by setting up multiple surgical simulation modules, each corresponding to different types of orthopedic surgery, the coverage of the preoperative planning system is greatly increased. Attached Figure Description

[0047] Figure 1 is a schematic diagram of the preoperative planning system for multiple diseases in orthopedics according to this application;

[0048] Figure 2 is a schematic diagram of the interface of the artificial intelligence preoperative planning system for multiple diseases in orthopedics in this application;

[0049] Figure 3 is a schematic diagram of the hip joint simulation module of the multi-disease artificial intelligence preoperative planning system for all orthopedic diseases in this application.

[0050] Figure 4 is a schematic diagram of the knee joint simulation module of the multi-disease artificial intelligence preoperative planning system for all orthopedic diseases in this application.

[0051] Figure 5 is a schematic diagram of the spinal joint simulation module of the preoperative planning system for multiple diseases in orthopedics in this application.

[0052] Figure 6 is a schematic diagram of the sports medicine simulation module of the artificial intelligence preoperative planning system for multiple diseases in orthopedics in this application;

[0053] Figure 7 is a schematic diagram of the trauma surgery simulation module of the artificial intelligence preoperative planning system for multiple diseases in orthopedics in this application.

[0054] Figure 8 is a model architecture diagram of the skeletal segmentation module of the preoperative planning system for multiple diseases in orthopedics in this application;

[0055] Figure 9 is a model architecture diagram of the feature extraction structure of the skeletal segmentation module of the preoperative planning system for multiple diseases in orthopedics in this application. Detailed Implementation

[0056] To make the above-mentioned objects, features, and advantages of this application more apparent and understandable, specific embodiments of this application will be described in detail below with reference to the accompanying drawings. Although exemplary embodiments of this application are shown in the drawings, it should be understood that this application can 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 this application and to fully convey the scope of this application to those skilled in the art.

[0057] It should be noted that, unless otherwise stated, the technical or scientific terms used in this application shall have the ordinary meaning as understood by one of ordinary skill in the art to which this application pertains.

[0058] This application provides a comprehensive orthopedic multi-disease artificial intelligence preoperative planning system. The structure of this device is shown in Figures 1-9. As shown in Figures 1 and 2, it is a schematic diagram of the structure of the comprehensive orthopedic multi-disease artificial intelligence preoperative planning system. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system includes:

[0059] The import module is configured to import medical imaging data;

[0060] The 3D reconstruction module is configured to build a corresponding 3D image model based on medical imaging data;

[0061] The skeleton segmentation module is configured to segment the skeleton in medical image data;

[0062] And at least the following three modules:

[0063] The hip joint simulation module is configured to simulate the hip joint surgery process and determine the preoperative planning scheme for hip joint surgery based on medical imaging data of the hip joint.

[0064] The knee joint simulation module is configured to simulate the knee joint surgery process and determine the preoperative planning scheme for knee joint surgery based on medical imaging data of the knee joint.

[0065] The spinal joint simulation module is configured to simulate the spinal joint surgery process and determine the preoperative planning scheme for spinal joint surgery based on medical imaging data of the spinal joint.

[0066] The sports medicine simulation module is configured to simulate the movement process of the knee joint based on medical imaging data of the knee joint;

[0067] The trauma surgery simulation module is configured to simulate the trauma surgery process and determine the preoperative planning scheme for trauma surgery based on medical imaging data of the trauma site.

[0068] In this application, by setting up multiple surgical simulation modules, each corresponding to different types of orthopedic surgery, the coverage of the preoperative planning system is greatly increased.

[0069] In one embodiment, as shown in FIG3, the hip joint simulation module includes a first key point recognition submodule, a first template measurement submodule, a first display submodule, and a first guide plate submodule.

[0070] The first key point recognition submodule is configured to recognize key points of the femur and acetabulum;

[0071] The first module measurement submodule is configured to measure or calculate the anteversion angle, abduction angle, coverage of the acetabular cup prosthesis template, postoperative leg length difference, offset difference, and the amount and location of osteotomy in the proximal femur.

[0072] The first display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation;

[0073] The first guide plate submodule is configured to output the planned osteotomy amount and osteotomy location, and export the osteotomy guide plate model.

[0074] In one implementation, the key points identified by the first key point identification submodule include: the center of the left femoral head, the center of the femoral diameter, above the lesser trochanter, the lesser trochanter, the proximal end of the femur, the distal end of the femur, the medial epicondyle and the lateral epicondyle, the center of the right femoral head, above the lesser trochanter, the lesser trochanter, the proximal end of the femur, and the distal end of the femur.

[0075] The process by which the first key point recognition submodule identifies the femoral head center is as follows:

[0076] Based on the centroid formula of the planar image corresponding to the 3D model, the pixel coordinates of the femoral head center point in the femoral region of the 3D image are obtained; the pixel coordinates are converted into image coordinates; and the position of the femoral head rotation center is determined.

[0077] In this application, the identification of other key points can be carried out according to the actual situation, or through a pre-trained identification model. The specific process will not be described in this application.

[0078] In this application, the first module measurement submodule is configured to calculate and determine the replacement acetabular cup prosthesis:

[0079] Based on the position of the femoral head rotation center, first dimension information is obtained, and second dimension information is determined based on the first dimension. The first dimension information is the dimension information corresponding to the femoral head, and the second dimension information is the dimension information corresponding to the acetabular cup prosthesis model.

[0080] The first module's measurement submodule is also configured to determine the type and placement of the femoral stem prosthesis model:

[0081] Based on the identified femoral and cortical regions, the medullary canal region is determined; the midpoint coordinates of each medullary canal layer within the medullary canal region are calculated, and based on these midpoint coordinates, all center points are fitted with straight lines to determine the anatomical axis of the medullary canal; based on the anatomical axis of the medullary canal and the femoral neck axis, the angle value of the femoral neck-shaft angle is calculated; based on the angle value, the medullary canal region, and the position of the femoral head rotation center, the type and placement position of the femoral stem prosthesis model are determined.

[0082] In this application, after determining the type and placement of the prosthesis, the osteotomy location and amount are directly determined by performing Boolean operations on the prosthesis in the corresponding position.

[0083] The first display submodule is configured to display an image of the prosthesis template and the bone fitting, showing the femoral stem prosthesis and the acetabular prosthesis, and covering the corresponding position of the hip joint with the prosthesis; if the placement position of the acetabular cup prosthesis and the placement position of the femoral stem prosthesis both meet the preset position requirements, they can be directly output through the first guide plate submodule, or adjusted and output according to the operator's operation.

[0084] Among them, the preset position requirements for the acetabular cup prosthesis can be, for example, that the coverage rate of the acetabular fossa is greater than 75% after the acetabular cup prosthesis is placed in the acetabular fossa, and the preset position requirements for the femoral stem prosthesis can be, for example, that the angle between the long axis of the femoral stem prosthesis and the long axis of the femur is less than or equal to 3° after the femoral stem prosthesis is placed in the medullary canal.

[0085] In one embodiment, as shown in FIG4, the knee joint simulation module includes a second key point recognition submodule, a second template measurement submodule, a second display submodule, and a second guide plate submodule.

[0086] The second key point recognition submodule is configured to recognize key points on the femur and tibia;

[0087] The second module measurement submodule is configured to measure or calculate the distal femoral osteotomy amount, the posterior femoral condyle osteotomy amount, the tibial plateau osteotomy amount, and to calculate the sum of the medial distal femoral osteotomy amount and the medial tibial plateau osteotomy amount, the sum of the lateral distal femoral osteotomy amount and the lateral tibial plateau osteotomy amount, the sum of the medial posterior femoral condyle osteotomy amount and the medial tibial plateau osteotomy amount, and the sum of the lateral posterior femoral condyle osteotomy amount and the lateral tibial plateau osteotomy amount.

[0088] The second display submodule is configured to display an image of the prosthesis template fitted to the skeleton, and adjust it according to the operator's operation;

[0089] The second guide plate submodule is configured to output the planned osteotomy amount and osteotomy location, and export the osteotomy guide plate model.

[0090] In one embodiment, the key points identified by the second key point identification submodule include: medial epicondyle, lateral epicondyle, distal medial, distal lateral, medial posterior condyle, lateral posterior condyle, anterior condyle, proximal femoral end point and distal femoral end point, medial tibial plateau point, lateral tibial plateau point, medial third of the tibial tuberosity, posterior cruciate ligament insertion point, proximal tibial end point and distal tibial end point.

[0091] In this application, the key points are identified using any one or more neural network models selected from MTCNN, locnet, Pyramid Residual Module, Densenet, hourglass, resnet, SegNet, Unet, R-CNN, Fast R-CNN, Faster R-CNN, R-FCN, and SSD.

[0092] In this application, the second module measurement submodule is further configured to calculate key axes, which include the femoral anatomical axis, the femoral mechanical axis, the tibial anatomical axis and the tibial mechanical axis, as well as the tibiofemoral angle and the distal femoral angle.

[0093] Preferably, the key axis further includes one or more of the following: the transcondylar line, the posterior condylar line, the tibio-patellar joint line, the femoral sagittal axis, the femoral-patellar joint line, and the posterior condylar angle of the femur.

[0094] In this application, the second display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation; wherein, the displayed image of the prosthesis template and the bone fitting can be scaled, rotated along any axis, and moved.

[0095] This provides a more intuitive view of the three-dimensional structure of the skeleton, allowing doctors (or other medical personnel) to observe the images of the skeleton from multiple angles and levels.

[0096] In one embodiment, as shown in FIG5, the spinal joint simulation module includes a third key point recognition submodule, a third template measurement submodule, a third display submodule, and a third guide plate submodule.

[0097] The third key point recognition submodule is configured to recognize key points of each vertebra of the spine.

[0098] The third module measurement submodule is configured to measure or calculate the dimensions and setting positions of the screw module, fusion template, titanium rod template and cross-connecting template;

[0099] The third display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation;

[0100] The third guide plate submodule is configured to output the planned positions of the screw module, fusion template, titanium rod template, and transverse connecting template, and export the spinal guide plate model.

[0101] The size and placement of the screws can be generated using a pre-trained neural network model.

[0102] For example, the preprocessed image data is input into the neural network model in the measurement submodule of the third module to obtain the screw insertion point and insertion direction. Based on the insertion point and insertion direction, the screw diameter and screw length are calculated. Based on the screw diameter and screw length, the appropriate screw model is determined in the prosthesis database.

[0103] The size and location of the screws can also be determined by the setting method:

[0104] Left entry point: Identify the midpoint of the inferior articular process and draw a vertical line 3 mm lateral to this point. Identify the base of the transverse process and draw a horizontal line at the upper 1 / 3 of the base. Take the intersection of the two lines. Right entry point: Identify the midpoint of the transverse process and draw a horizontal line through this point. Identify the outer edge of the superior articular process and draw a vertical line through the outer edge of the superior articular process. Take the intersection of the two lines. Screw insertion direction: In the transverse section, the left and right screws are each inclined inward at 35°. In the sagittal plane, the left and right screws are parallel to the superior and inferior endplates.

[0105] Screw diameter: Perpendicular to the insertion direction, divide the pedicle into n faces. Calculate the pedicle diameter for each face. When the minimum pedicle diameter in all faces is ≤4.0mm, select a screw diameter of 3.5mm. When the minimum pedicle diameter in all faces is greater than 4.0mm, select a screw diameter of 4.0mm.

[0106] Screw length: Starting from the insertion point, draw a straight line along the insertion direction until it breaks through the anterior cortical bone of the vertebral body. Calculate the distance from the insertion point to the breakthrough point; 80% of this distance is the screw insertion depth. Calculate the value a = screw length - insertion depth * 80%. If a > 0 and the value of a is minimized, then the optimal screw length is found.

[0107] The generation of the titanium rod's size and placement involves: calculating the width of the U-shaped structure at the screw's tail and fitting the titanium rod curve; calculating the optimal titanium rod width based on the width of the U-shaped structure at the screw's tail; and determining the appropriate titanium rod model from the prosthesis database based on the optimal titanium rod width.

[0108] The generation of the fusion device's size and placement involves calculating the length and width of the intervertebral space between two adjacent vertebrae, and then determining the appropriate fusion device model from the prosthesis database based on these dimensions.

[0109] The principle behind calculating the intervertebral disc length and width to match the optimal model is as follows:

[0110] Length: Identify the anterior and posterior edges of the vertebral body; the distance between these two points is the intervertebral space length.

[0111] Width: Identify the anterior and posterior edges of the vertebral body, and draw a straight line connecting the two points. At the midpoint of this line, draw a perpendicular line. Find the two points where this perpendicular line intersects the vertebral body. The distance between these two points is the width of the intervertebral space.

[0112] Optimal matching: Calculate the value a = intervertebral space length - fusion cage length. The optimal fusion cage length is when a > 0 and the value of a is the smallest. Calculate the value b = intervertebral space width / 2 - fusion cage width. The optimal fusion cage width is when b > 0 and the value of b is the smallest.

[0113] Among them, the generation of the size and setting position of the transverse connector is as follows: the required transverse connector length is calculated based on the fitting curve of the titanium rod, and the appropriate transverse connector model is determined from the prosthesis database based on the required transverse connector length.

[0114] Mark the first and second punctuation points at certain lengths on the left and right titanium rods, and connect the first and second punctuation points with a straight line.

[0115] Within the length of the titanium rod, the straight line is translated vertically, and during the translation, the straight line always connects the left and right titanium rods;

[0116] The length range of the straight line is the length range of the horizontal connection. The maximum length is taken as m. The calculated value a = horizontal connection model (i.e., the length of the horizontal connection) - m. When a ≥ 0 and the value of a is the smallest, it is the optimal horizontal connection model.

[0117] In one embodiment, as shown in FIG6, the sports medicine simulation module includes a fourth key point recognition submodule, a fourth template measurement submodule, and a fourth display submodule.

[0118] The fourth key point recognition submodule is configured to recognize key points of the femur and tibia;

[0119] The fourth module, the measurement submodule, is configured to measure or calculate the length and expansion / contraction changes of the bone tunnel at different flexion angles of the knee joint.

[0120] The fourth display submodule is configured to display the dynamic effects of the knee joint at different flexion angles and the changes in the bone tunnel.

[0121] The key points identified by the fourth key point identification submodule are two points connecting to the condyle (medial condyle point E and lateral condyle point F), point A on the posterosuperior medial side of the lateral femoral condyle, point B on the lateral border of the anterior femoral condyle, point C at the anteromedial 1 / 3 of the intercondylar eminence of the tibia, and point D on the medial border of the tibial tuberosity, for a total of 6 points.

[0122] The fourth module, the measurement submodule, is further configured to: calculate the length variation difference between point A, located posterosuperior to the medial side of the lateral femoral condyle, and point C, located anteromedial 1 / 3 of the intercondylar eminence of the tibia, at different angles; and determine the optimized positions of point A and point C, where the length variation difference is smallest. This achieves optimization of key points.

[0123] Specifically, optimization of key point locations:

[0124] Calculate the shortest AC length stretching positions of the femur and tibia at 0°, 90°, and 120°. Points A and C are the initial identification points for point recognition.

[0125] Keeping the tibia still, rotate the femur 90° and 120° around the condylar line, and calculate the distance between points A and C.

[0126] Optimization methods for points A and C:

[0127] Using the initial points A and C identified by the point recognition network as the center of the sphere, find a set of points on the bone surface with a radius of 5mm, and take one point from each of the A and C point sets to form N AC point combinations.

[0128] Calculate the difference in AC length expansion and contraction at each AC point at 0°, 90°, and 120°.

[0129] Repeat step two, calculating the difference in AC length expansion / contraction for each group of AC points at 0°, 90°, and 120°. Find the pair of AC points corresponding to the minimum expansion / contraction change.

[0130] Output the optimized coordinates of points A and C.

[0131] The fourth module, the measurement submodule, is also configured to: perform bone tunnel planning based on knee joint CT images to obtain bone tunnel planning results for the tibial side, ligament side, and femoral side.

[0132] Specifically, bone tunnel planning:

[0133] The knee joint bone tunnel is divided into three segments: the tibial side, the ligament side, and the femoral side. The femoral side tunnel is formed by points A and B, the tibial side tunnel is formed by points C and D, and the ligament tunnel is formed by points A and C.

[0134] The knee joint bone tunnel is formed based on the tibial side, ligaments, and femoral side.

[0135] The fourth display submodule is configured to display the changes in AC length of the femur and tibia at 0°, 90°, and 120°.

[0136] In one embodiment, as shown in Figure 7,

[0137] The trauma surgery simulation module includes a fifth key point recognition submodule, a fifth template measurement submodule, a fifth display submodule, and a fifth guide plate submodule;

[0138] The fifth key point recognition submodule is configured to recognize key points of the femur and acetabulum;

[0139] The fifth module, the measurement submodule, is configured to measure or calculate the dimensions and placement of the main screw module, the tail cap template, the spiral blade template, and the interlocking screw template.

[0140] The fifth display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation;

[0141] The fifth guide plate submodule is configured to output the planned positions of the main nail module, tail cap template, spiral blade template, and interlocking screw template.

[0142] The trauma is an intertrochanteric fracture of the femur.

[0143] The fifth module, the measurement submodule, is also configured to: repair the broken bone at the injury site and determine the rotation matrix of the broken bone; and determine the size and setting position of the instrument template based on the repaired injury site.

[0144] The repair of the fractured bone at the site of injury, and the determination of the rotation matrix of the fractured bone, includes:

[0145] The proximal fracture bone is fitted to determine the center point of the femoral head and the center point of the femoral neck, and the axis of the femoral head and neck is determined; the proximal fracture bone is the femoral head, and the distal fracture bone is the femoral shaft.

[0146] The distal fracture bone was sampled and fitted to obtain the femoral shaft axis.

[0147] Extract the fracture surface point set of the proximal fracture bone and the fracture surface point set of the distal fracture bone;

[0148] The rotation matrix of the femoral head is determined based on the set of fracture surface points, the axis of the femoral shaft, and the axis of the femoral head and neck.

[0149] The determination of the rotation matrix of the femoral head based on the fracture surface point set, the femoral shaft axis, and the femoral head-neck axis includes:

[0150] Generate the initial rotation matrix of the femoral head;

[0151] Calculate the neck-shaft angle based on the femoral shaft axis and the adjusted femoral head-neck axis.

[0152] Calculate the anteversion angle based on the adjusted femoral head-neck axis and the reference plane;

[0153] Iterate the rotation matrix until the neck angle and the forward tilt angle meet the preset angle range;

[0154] Calculate the point set distance based on the fracture surface point set of the femoral shaft and the iterated fracture surface point set of the femoral head;

[0155] Repeat the rotation matrix until the distance between the points is minimized.

[0156] In this application, the neck-shaft angle and anteversion angle are first constrained within a preset angle range by adjusting the rotation matrix; then, under the premise of constraining the neck-shaft angle and anteversion angle, the rotation matrix is ​​finely adjusted to minimize the distance between the point sets. At this point, the fracture surface of the femoral shaft and the fracture surface of the femoral head should fit exactly, thereby achieving the positional repair of the femoral fracture.

[0157] In this application, the distance between the fracture surfaces of the femoral head and femoral shaft is calculated to verify whether the splicing result meets the alignment splicing standard; the neck-shaft angle and anteversion angle are calculated to verify whether the splicing result meets the alignment splicing standard; thereby achieving accurate alignment splicing and repositioning of the femur, and achieving accurate repair.

[0158] In this application, the point set distance is the Euclidean distance between two fracture surface point sets; the ICP (Iterative Closest Point) algorithm can be used to find the set of point pairs with the minimum distance matching, and the rotation matrix can be further optimized.

[0159] In this application, the process for obtaining the initial planned positions of the main screw module, tail cap template, spiral blade template, and interlocking screw template in the image showing the fitting of the instrument template and the skeleton is as follows:

[0160] Master nail planning: Determine the optimal insertion path for the master nail between proximal and distal femoral fractures. Typically, an opening at the greater trochanter or the lower rim of the acetabulum is chosen. Determine the length and diameter of the master nail to allow it to traverse the entire fracture area and provide adequate support. Alignment with the femoral axis: Align the centerline of the master nail with the femoral axis to ensure the nail can pass through the medullary canal without disrupting the skeletal anatomy.

[0161] Tail cap design: The tail cap is typically installed at the end of the master screw and should be perfectly aligned with it. Ensure the tail cap's position will not cause damage to soft tissues (such as tendons or nerves). Select a tail cap of appropriate diameter and length based on the skeletal anatomy, avoiding the tail cap protruding outside the bone or entering the joint cavity.

[0162] Planning the helical blade insertion: Select an appropriate insertion path in the femoral head and neck region to ensure the helical blade can fully penetrate the femoral head and cover a large area of ​​bone. The insertion angle should be aligned with the femoral neck axis to avoid damaging the articular surface of the femoral head. Select the length of the helical blade based on the size of the femoral head and the length of the femoral neck. Generally, ensure that the blade tip is at least 5 mm away from the articular surface of the femoral head to avoid cutting the articular surface.

[0163] Positioning of the interlocking screws: Install one or more interlocking screws at the distal end of the master nail. The screw positions should be precisely aligned with the lock holes on the master nail.

[0164] In one embodiment, as shown in Figure 4, the knee joint simulation module further includes a unicompartmental submodule. The unicompartmental submodule is connected to the second key point recognition submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule. When the unicompartmental submodule is active, the second key point recognition submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule only process operations corresponding to a single knee joint.

[0165] In this application, the preoperative planning of the unicompartment is similar to that of the preoperative planning of the knee joint, except that in the preoperative planning of the knee joint, both knee joints are identified and planned; while in the preoperative planning of the unicompartment, only the knee joint on the side to be treated is identified and planned.

[0166] In this application, by setting up a unicompartment submodule connected to each submodule in the knee joint simulation module, the second key point identification submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule are controlled to process only one knee joint during unicompartmental preoperative planning.

[0167] In this application, the specific preoperative planning for unicompartmental surgery can be referenced from the preoperative planning for knee surgery, and will not be elaborated further.

[0168] In one implementation, as shown in Figure 8, the bone segmentation module includes a pre-trained segmentation model for segmenting the input medical image data, including:

[0169] The medical image data is input into the feature extraction structure to obtain a low-level feature map and multiple high-level feature maps;

[0170] Multiple high-level feature maps are input into the encoding structure to obtain the encoded feature map;

[0171] The encoded feature map and low-level feature map are input into the decoding structure to obtain the segmentation result.

[0172] In one implementation, as shown in Figure 8, the step of inputting multiple high-level feature maps into the coding structure to obtain coded feature maps includes:

[0173] Multiple high-level feature maps are input into multiple branches set in parallel to obtain corresponding branch feature maps;

[0174] The global feature map is obtained by performing convolution processing on the feature maps of multiple branches separately.

[0175] The global feature maps of some branches are concatenated to obtain a concatenated feature map;

[0176] The spliced ​​feature map is processed by channel attention and convolution to obtain an attention feature map.

[0177] The attention feature map is concatenated with the global feature map of the remaining branches, and spatial attention and 1×1 convolution are performed to obtain the encoded feature map.

[0178] In one implementation, as shown in Figure 8, the step of inputting the encoded feature map and the low-level feature map into the decoding structure to obtain the segmentation result includes:

[0179] Multiple branches are set in parallel to input the low-level feature map, resulting in multiple corresponding decoding branch feature maps;

[0180] The feature maps of multiple branches are convolved separately to obtain the decoded global feature map;

[0181] The global feature maps of some branches are concatenated and convolved to obtain convolutional feature maps;

[0182] The convolutional feature map and the decoded global feature map of the remaining branch are concatenated and processed by a convolutional layer to obtain the decoded convolutional map;

[0183] The decoded convolutional map and the upsampled encoded feature map are fused, convolved, and upsampled to obtain the segmentation result.

[0184] As shown in Figure 8, there are five high-level feature maps. The encoding structure of these five feature maps has five parallel branches, each corresponding to one of the five extracted feature maps from the input. Specifically, the first branch is concatenated after 1×1 convolution; the second branch is concatenated after 3×3 convolution and 1×1 convolution; the third branch is concatenated after 3×3 convolution and 1×1 convolution; the fourth branch is concatenated after 3×3 convolution and 1×1 convolution; and the fifth branch is concatenated after 1×1 convolution.

[0185] In the specific splicing process, the first, second, third and fourth branches are spliced ​​directly, and after channel attention processing, they are spliced ​​with the fifth branch, and then spatial attention processing is performed.

[0186] In this application, channel attention and spatial attention are extracted separately, which can make full use of the characteristics of each and capture information in different dimensions. This avoids interference from information in different dimensions, allows the model to focus more on information extraction in specific dimensions, and improves the model's performance.

[0187] In this application, the 1×1 convolution in the first branch and the 3×3 convolutions in the second, third, and fourth branches are dilated convolutions with different dilation rates. This enhances the extracted feature maps through different dilated convolutions, then they are concatenated and dynamically enhanced or weakened through a channel attention mechanism to distinguish the different features of the bone to be segmented from the surrounding tissue. This feature is then extracted again and concatenated with the remaining branches. Subsequently, spatial attention is utilized—that is, based on the attention intensity of different regions—to strengthen the features of the region of interest, improving the contrast between the bone region and the surrounding tissue, making the features more easily distinguishable. This highlights the shape and texture of the bone, resulting in a clearer and more complete segmentation of the boundary between the bone and the surrounding tissue.

[0188] In this application, by setting dilated convolution, the convolution kernel has a larger receptive field to cover a wider range of input image regions, which helps to capture a wider range of contextual information and makes the extracted features more semantic.

[0189] It should be noted that the 3×3 convolutions in different branches have the same structure but different parameters.

[0190] As shown in Figure 8, there is one low-level feature map, which is the feature map obtained first by the feature extraction structure, and it is also one of the high-level feature maps.

[0191] As shown in Figure 8, the decoding structure has five parallel branches, with low-level feature maps input to all five branches simultaneously. The first branch is concatenated after 1×1 dilated convolution; the second branch is concatenated after 3×3 dilated convolution; the third branch is concatenated after 3×3 dilated convolution; the fourth branch is concatenated after 3×3 dilated convolution; and the fifth branch is concatenated after 1×1 dilated convolution. The 1×1 and 3×3 dilated convolutions in different branches have the same structure, but different parameters and dilation rates.

[0192] In the specific splicing process, the first branch, the second branch, the third branch, and the fourth branch are spliced ​​directly, and after 3×3 convolution processing, they are spliced ​​with the fifth branch, and then 1×1 convolution processing is performed to obtain the decoded convolutional image.

[0193] In this process, the encoded feature map is upsampled and then fused with the decoded convolutional map. This fusion process can be a splicing process or other processing.

[0194] In this application, convolution and upsampling are performed after concatenation to obtain the segmentation result. After upsampling, the corresponding segmentation result image can be obtained through a 1x1 convolutional layer, or a Softmax or Sigmoid activation function.

[0195] It should be noted that in this application, the feature maps that are spliced ​​and fused are of the same size; if the sizes are different, they can be corrected by adding an additional 1×1 convolution or other size adjustment processes. This will not be elaborated further in this application.

[0196] In one implementation, as shown in Figure 9, the medical image data is input into a feature extraction structure to obtain a low-level feature map and multiple high-level feature maps, including:

[0197] Medical image data is processed by 7×7 convolution to obtain large convolution feature maps;

[0198] Global extraction processing is performed on the large convolutional feature map to obtain the global extracted feature map;

[0199] The first high-level feature map is obtained by concatenating the global extracted feature map and the large convolutional feature map; the first high-level feature map is also a low-level feature map.

[0200] The first high-level feature map is pooled to obtain a pooled feature map.

[0201] The pooling feature map is processed by a combination of 1×1 convolution, 3×3 convolution, and 1×1 convolution to obtain a combined feature map.

[0202] The pooled feature map is concatenated with the combined feature map to obtain the second high-level feature map;

[0203] The second high-level feature map is processed by a combination of 1×1 convolution, 3×3 convolution, and 1×1 convolution to obtain a high-level combined feature map.

[0204] The high-level combined feature map and the combined feature map are concatenated to obtain the third high-level feature map;

[0205] The third high-level feature map is processed by a combination of 1×1 convolution, 3×3 convolution, and 1×1 convolution to obtain a high-level three-combined feature map.

[0206] The high-level combined feature map and the high-level three-combined feature map are concatenated to obtain the fourth high-level feature map;

[0207] The fourth high-level feature map is processed by pooling and fully connected layers to obtain the fifth high-level feature map.

[0208] In this application, by directly performing 7×7 convolution processing on medical image data, richer contextual information and more large targets or large-scale features in the medical image data can be directly obtained, thereby capturing more global information at an early level, so that the generated low-level feature map contains more global information and improves global characteristics.

[0209] In this application, the first high-level feature map is simultaneously used as a low-level feature map input into the decoding structure, and the first high-level feature map is reused. While maintaining the multi-view and multi-size of the encoding structure, the global characteristics of the low-level feature map are improved.

[0210] In this application, by combining convolution and residual concatenation, key features are continuously extracted on the one hand, and global characteristics and quasi-low-level features in subsequent high-level feature maps are maintained through continuous residuals, thereby greatly improving the corresponding characteristics in the output high-level feature maps.

[0211] In one implementation, a global extraction process is performed on the large convolutional feature map to obtain a globally extracted feature map, including:

[0212] The large convolutional feature map is input into a parallel 1×1 convolutional layer to obtain the first extraction map, the second extraction map, and the third extraction map, respectively; wherein the first extraction map and the third extraction map are in HW×N format, and the second extraction map is in N×HW format;

[0213] Multiply the first extracted image and the second extracted image to obtain the multiplied feature map;

[0214] The multiplication feature maps are processed using a softmax-like method to obtain the multiplication coefficients;

[0215] Multiply the multiplication coefficients by the third extraction graph to obtain the coefficient multiplication graph;

[0216] The coefficient multiplication map and the large convolution feature map are added together to obtain the global extracted feature map.

[0217] In this application, the large convolutional feature map is divided into three branches and processed by 1×1 convolution. The dimensions of the first and third branches are adjusted to HW×N format through this convolution, and the dimension of the second branch is N×HW format. Then, the first branch and the third branch are multiplied to obtain a feature map with dimensions of HW×HW, and the coefficients are obtained by processing it with a softmax-like method. The coefficients are multiplied with the third branch, and the multiplied feature map is added to the large convolutional feature map to obtain the global extracted feature map.

[0218] In this application, the dimensions of the first and second branches are reversed by 1×1 convolution, thereby realizing the multiplication between feature maps; through this multiplication, long-range dependencies are captured; by setting coefficients, the dependencies are embedded in the feature maps; and on this basis, the original input is added and residual connections are used to embed the global extraction into the model without destroying the parameters.

[0219] In this application, the long-range dependencies between captured features are extracted globally, thereby enabling the model to extract global features and overcoming the deficiency of current models that mainly extract local features.

[0220] In one implementation, the activation function of the SoftMax-like operation is:

[0221] Where SoftMax1(x) is the activation function, x i x j Let i and j be the elements in the input vector, where i and j are the element indices.

[0222] In this application, SoftMax is a mathematical function typically used to convert a set of arbitrary real numbers into real numbers representing a probability distribution. Essentially, it is a normalization function that transforms a set of arbitrary real values ​​into probability values ​​between [0,1]. Because SoftMax converts them to values ​​between 0 and 1, they can be interpreted as probabilities. If one of the inputs is small or negative, SoftMax transforms it into a low probability; if the input is large, it transforms it into a high probability, but it will always remain between 0 and 1.

[0223] However, for the standard SoftMax function, since the inputs are mapped to the range of 0 to 1 and the sum of all output values ​​is 1, this means that even some very small input values ​​will have a non-zero output value after processing by the SoftMax function. This amplifies the noise, resulting in the final output being more affected by noise.

[0224] In this application, a 1 is added to the denominator of the SoftMax-like functions; this change means that when the input values ​​are very small, their output values ​​can be closer to zero. This allows the corresponding output to tend to zero when there is no valuable information to add, thus greatly reducing unnecessary noise.

[0225] It should be noted that the segmentation model in this application (conventional models are only used for segmenting partial skeletons and are too simplistic in structure) is intended for segmenting the entire orthopedic AI, including not only the knee joint but also the hip joint, spine, and trauma areas. This application achieves accurate segmentation of complex human skeletons by repeatedly preserving global features and setting low-level and high-level feature maps.

[0226] The specific training process of the segmentation model will not be described in detail in this application.

[0227] The terms used in this application are all standard orthopedic terms, and their explanations are as follows:

[0228] Femoral anatomical axis: the center line of the femoral shaft.

[0229] The femoral mechanical axis has one end located at the center of the hip joint and the other end located at the center of the knee joint of the femur (the apex of the intercondylar fossa of the femur).

[0230] Tibial anatomical axis: the central line of the tibial shaft.

[0231] Tibial mechanical axis: one end is located at the center of the tibial knee joint (the center of the intercondylar spine), and the other end is located at the center of the tibial ankle joint (the midpoint of the line connecting the lateral cortical bone of the medial and lateral malleoli).

[0232] Condylar line: The line connecting the highest point of the medial condyle and the lateral condyle of the femur.

[0233] Posterior condyle line: The line connecting the lowest points of the medial and lateral posterior condyles of the femur.

[0234] Femoral-knee joint line: The line connecting the lowest points of the distal femur.

[0235] Tibial-knee joint line: The line connecting the lowest point on the medial side and the highest point on the lateral side of the tibial plateau.

[0236] The femoral sagittal axis is the line connecting the center of the posterior cruciate ligament insertion point and the inner edge of the tibial tuberosity.

[0237] Tibiofemoral angle (mTFA): The angle formed by the mechanical axis of the femur and the mechanical axis of the tibia.

[0238] Distal femoral angle: The angle between the mechanical axis and the anatomical axis of the femur.

[0239] Posterior condyle angle (PCA): The angle between the projection lines of the femoral condyle line and the posterior condyle line on the cross section.

[0240] Furthermore, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include one or more of that feature. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0241] In this application, unless otherwise expressly specified and limited, the terms "set up," "connected," "linked," "connected," etc., should be interpreted broadly. For example, they can refer to electrical connection / communication connection, electrical coupling / communication coupling, or integration; they can refer to direct connection or indirect connection through an intermediate medium; they can refer to the internal connection of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this application according to the specific circumstances.

[0242] In the description of this application, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0243] It should be noted that numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some embodiments, well-known structures and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0244] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A comprehensive orthopedic multi-disease artificial intelligence preoperative planning system, comprising: The import module is configured to import medical imaging data; The 3D reconstruction module is configured to build a corresponding 3D image model based on medical imaging data; The skeleton segmentation module is configured to segment the skeleton in medical image data; And at least the following three modules: The hip joint simulation module is configured to simulate the hip joint surgery process and determine the preoperative planning scheme for hip joint surgery based on medical imaging data of the hip joint. The knee joint simulation module is configured to simulate the knee joint surgery process and determine the preoperative planning scheme for knee joint surgery based on medical imaging data of the knee joint. The spinal joint simulation module is configured to simulate the spinal joint surgery process and determine the preoperative planning scheme for spinal joint surgery based on medical imaging data of the spinal joint. The sports medicine simulation module is configured to simulate the movement process of the knee joint based on medical imaging data of the knee joint; The trauma surgery simulation module is configured to simulate the trauma surgery process and determine the preoperative planning scheme for trauma surgery based on medical imaging data of the trauma site.

2. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to claim 1, wherein, The hip joint simulation module includes a first key point recognition submodule, a first template measurement submodule, a first display submodule, and a first guide plate submodule; The first key point recognition submodule is configured to recognize key points of the femur and acetabulum; The first module measurement submodule is configured to measure or calculate the anteversion angle, abduction angle, coverage of the acetabular cup prosthesis template, postoperative leg length difference, offset difference, and the amount and location of osteotomy in the proximal femur. The first display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation; The first guide plate submodule is configured to output the planned osteotomy amount and osteotomy location, and export the osteotomy guide plate model.

3. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to claim 2, wherein, The key points identified by the first key point identification submodule include: the center of the left femoral head, the center of the femoral diameter, above the lesser trochanter, the lesser trochanter, the proximal end of the femur, the distal end of the femur, the medial epicondyle and the lateral epicondyle, the center of the right femoral head, above the lesser trochanter, the lesser trochanter, the proximal end of the femur, and the distal end of the femur.

4. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to any one of claims 1-3, wherein, The knee joint simulation module includes a second key point recognition submodule, a second template measurement submodule, a second display submodule, and a second guide plate submodule; The second key point recognition submodule is configured to recognize key points on the femur and tibia; The second module measurement submodule is configured to measure or calculate the distal femoral osteotomy amount, the posterior femoral condyle osteotomy amount, the tibial plateau osteotomy amount, and to calculate the sum of the medial distal femoral osteotomy amount and the medial tibial plateau osteotomy amount, the sum of the lateral distal femoral osteotomy amount and the lateral tibial plateau osteotomy amount, the sum of the medial posterior femoral condyle osteotomy amount and the medial tibial plateau osteotomy amount, and the sum of the lateral posterior femoral condyle osteotomy amount and the lateral tibial plateau osteotomy amount. The second display submodule is configured to display an image of the prosthesis template fitted to the skeleton, and adjust it according to the operator's operation; The second guide plate submodule is configured to output the planned osteotomy amount and osteotomy location, and export the osteotomy guide plate model.

5. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to claim 4, wherein, The key points identified by the second key point identification submodule include: medial epicondyle, lateral epicondyle, distal medial, distal lateral, medial posterior condyle, lateral posterior condyle, anterior condyle, proximal femoral end and distal femoral end, medial tibial plateau point, lateral tibial plateau point, medial 1 / 3 of the tibial tuberosity, posterior cruciate ligament insertion point, proximal tibial end and distal tibial end.

6. The preoperative planning system for multiple diseases in orthopedics according to any one of claims 1-3, wherein, The spinal joint simulation module includes a third key point recognition submodule, a third template measurement submodule, a third display submodule, and a third guide plate submodule; The third key point recognition submodule is configured to recognize key points of each vertebra of the spine. The third module measurement submodule is configured to measure or calculate the dimensions and setting positions of the screw module, fusion template, titanium rod template and cross-connecting template; The third display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation; The third guide plate submodule is configured to output the planned positions of the screw module, fusion template, titanium rod template, and transverse connecting template, and export the spinal guide plate model.

7. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to any one of claims 1-3, wherein, The sports medicine simulation module includes a fourth key point recognition submodule, a fourth template measurement submodule, and a fourth display submodule; The fourth key point recognition submodule is configured to recognize key points of the femur and tibia; The fourth module, the measurement submodule, is configured to measure or calculate the length and expansion / contraction changes of the bone tunnel at different flexion angles of the knee joint. The fourth display submodule is configured to display the dynamic effects of the knee joint at different flexion angles and the changes in the bone tunnel.

8. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to any one of claims 1-3, wherein, The trauma surgery simulation module includes a fifth key point recognition submodule, a fifth template measurement submodule, a fifth display submodule, and a fifth guide plate submodule; The fifth key point recognition submodule is configured to recognize key points of the femur and acetabulum; The fifth module, the measurement submodule, is configured to measure or calculate the dimensions and placement of the main screw module, the tail cap template, the spiral blade template, and the interlocking screw template. The fifth display submodule is configured to display an image of the prosthesis template and the bone fitting, and to adjust it according to the operator's operation; The fifth guide plate submodule is configured to output the planned positions of the main nail module, tail cap template, spiral blade template, and interlocking screw template.

9. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to claim 4, wherein, The knee joint simulation module also includes a unicompartment submodule, which is connected to the second key point recognition submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule. When the unicompartment submodule is active, the second key point recognition submodule, the second template measurement submodule, the second display submodule, and the second guide plate submodule only process operations corresponding to a single knee joint.

10. The comprehensive orthopedic multi-disease artificial intelligence preoperative planning system according to any one of claims 1-3, wherein, The skeleton segmentation module is equipped with a pre-trained segmentation model, which segments the input medical image data as follows: The medical image data is input into the feature extraction structure to obtain a low-level feature map and multiple high-level feature maps; The high-level feature map is input into the encoding structure to obtain the encoded feature map; The encoded feature map and low-level feature map are input into the decoding structure to obtain the segmentation result.

Images