A medical interactive system for vertebra

Abstract

The present application relates to the technical field of spine modeling, in particular to a medical technology interactive system suitable for vertebrae, comprising: collecting three-dimensional CT images of the vertebrae region and establishing a three-dimensional vertebrae model; calculating bone parameters to realize the key point representation of the vertebrae region; identifying the key points of the three-dimensional vertebrae model and adding corresponding bone parameters, then generating local vectors; generating a restoration prediction model and labeling the key point corresponding relationship. In view of the problem that the three-dimensional reconstruction process of the spine in the prior art is not accurate enough, on the basis of three-dimensional reconstruction of the vertebrae region, a number of specific key points are selected and bone parameters of the key points are calculated, effective measurement of related data of the key point parts is realized, then the local image features and bone parameters of the three-dimensional reconstruction are vectorized by the vectorization module and a vector matrix is established, more accurate representation of the vertebrae region is realized, and the actual restoration result is conveniently predicted.

Description

Technical Field

[0001] This invention relates to the field of spinal modeling technology, and more specifically to a medical technology interaction system applicable to the vertebrae. Background Technology

[0002] The skeletal system is an important component of the human body, comprising 26 vertebrae, 1 sacrum, 1 coccyx, 1 sternum, and 12 pairs of ribs, forming the spine and rib cage. Spinal deformities are a common condition. Scoliosis is defined as lateral displacement of the spine in the sagittal plane, where the angle of displacement exceeds 10 degrees. This curvature can be unidirectional, bidirectional, or asymmetrical, resulting in a C- or S-shaped body shape. It is often accompanied by spinal rotation and sagittal lordosis or kyphosis, as well as rib and shoulder asymmetry. A normal spine, viewed from the side, shows two opposite, normal physiological curves. Kyphosis occurs when multiple segments of the spine bend backward in the coronal plane, forming an arc-shaped spinal deformity, also accompanied by spinal rotation and sagittal kyphosis or lordosis. Normally, the thoracic vertebrae, from T1 to T12, have a kyphosis of 20 to 45 degrees; a kyphosis exceeding 45 degrees is considered "excessive kyphosis." Adolescent Scheuermann syndrome kyphosis is "excessive kyphosis".

[0003] There are already many technical solutions for assisting in the diagnosis of spinal deformities.

[0004] For example, patent application CN202410205610.9 discloses a spinal deformity analysis system based on key point detection, relating to the medical field. This spinal deformity analysis system based on key point detection includes: a spinal surface morphology acquisition subsystem, a key point analysis subsystem, an analysis and calculation subsystem, an evaluation index calculation subsystem, a data analysis subsystem, and a central control platform. The spinal surface morphology acquisition subsystem is used to collect morphological data of the spine on the human body surface. The key point analysis subsystem is used to process and analyze the information acquired by the spinal surface morphology acquisition subsystem to obtain the coordinates and posture information of each vertebral region of the spine to be analyzed. Through the data analysis subsystem, the health status of the spine can be understood more intuitively. The evaluation indicators and early warning prompts generated by the system help improve users' awareness of spinal health, prompting people to pay attention to and adopt positive lifestyles and postures to maintain the normal physiological state of the spine.

[0005] For example, patent application CN202411536659.9 discloses a method, device, equipment, and computer-readable storage medium for preoperative planning of spinal deformity correction based on deep learning. The method includes: acquiring patient identity information, spinal CT images, spinal Cobb angle, spinal bone density, and spinal offset distance; wherein the patient identity information includes at least gender and age; based on the patient's identity information, spinal Cobb angle, spinal bone density, and spinal offset distance, big data screening identifies the most suitable high-quality cases for the patient; inputting the most suitable high-quality cases, the patient's spinal CT images, and spinal bone density into a preset preoperative planning model for spinal deformity correction, and outputting the preoperative planning results. This allows for more accurate preoperative planning for spinal deformity correction.

[0006] However, in actual implementation, the inventors found that the above technical solution mainly focuses on the three-dimensional reconstruction process of the patient's vertebrae and simply predicts the general direction of correction based on the reconstruction model, which is not accurate enough in terms of data. Summary of the Invention

[0007] In view of the above-mentioned problems in the existing technology, a medical technology interaction system suitable for the spine is provided.

[0008] The specific technical solution is as follows: A medical technology interaction system applicable to the spine, comprising: The image acquisition module acquires three-dimensional CT images of the spinal region and establishes a three-dimensional spinal model. A modeling module, which is connected to the image acquisition module; The modeling module calculates skeletal parameters from the three-dimensional CT images to characterize key points in the vertebral region. The skeletal parameters consist of C7PL-CSVL parameters, clavicle angle, shoulder height, T1 vertebral tilt angle, pelvic level, and local coronal parameters of scoliosis. A vectorization module, which is connected to the modeling module; The vectorization module identifies key points in the three-dimensional vertebral model and adds corresponding skeletal parameters as annotations for the key points. Then, it generates local vectors based on the key points and the annotations. A restoration prediction module, which is connected to the vectorization module; The restoration prediction module generates a restoration prediction model for the vertebral region based on the local vectors and sample cases, and marks the correspondence between key points before and after adjustment.

[0009] On the other hand, the modeling module includes a C7PL-CSVL parameter generation module, which includes: The first vertebral recognition module identifies the seventh vertebra and sacrum from the cross section of the three-dimensional vertebral model, and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A normal mapping module, which is connected to the first spine recognition module; The normal establishment module establishes a first normal line for the seventh vertebra relative to the horizontal reference line, and establishes a second normal line for the sacrum relative to the horizontal reference line; A first parameter calculation module is connected to the normal establishment module; The first parameter calculation module calculates the distance between the first normal and the second normal to obtain the C7PL-CSVL parameter.

[0010] On the other hand, the modeling module includes a clavicle angle generation module, which includes: The second vertebral recognition module identifies the left and right clavicles from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A clavicle reference line generation module, wherein the clavicle reference line generation module is connected to the second spine recognition module; The clavicle reference line generation module measures the highest points of the left and right clavicles relative to the horizontal reference line, establishes a clavicle reference line based on the highest points of the clavicles, and generates a line connecting the vertices of the left and right clavicles. The second parameter calculation module is connected to the clavicle reference line generation module. The second parameter calculation module calculates the angle between the clavicle reference line and the line connecting the vertices to obtain the clavicle angle.

[0011] On the other hand, the modeling module includes a shoulder height generation module, which includes: The third vertebral recognition module identifies the left and right clavicles from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A shoulder height reference line generation module, which is connected to the third vertebral recognition module; The shoulder height reference line generation module establishes a left shoulder horizontal reference line parallel to the horizontal reference line for the highest point of the left clavicle, and establishes a right shoulder horizontal reference line parallel to the horizontal reference line for the highest point of the right clavicle. The third parameter calculation module is connected to the shoulder height reference line generation module; The third parameter calculation module calculates the distance between the horizontal reference line of the left shoulder and the horizontal reference line of the right shoulder to obtain the height of both shoulders.

[0012] On the other hand, the modeling module includes a T1 vertebral body tilt angle generation module, which includes: The fourth vertebral recognition module identifies the first thoracic vertebra from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A thoracic vertebra reference line generation module, which is connected to the fourth vertebral recognition module; The thoracic vertebra reference line generation module extracts the left and right endpoints of the first thoracic vertebra to establish a line connecting the first thoracic vertebrae. The fourth parameter calculation module is connected to the thoracic vertebra reference line generation module; The fourth parameter calculation module calculates the angle between the line connecting the first thoracic vertebrae and the horizontal reference line to obtain the clavicle angle.

[0013] On the other hand, the modeling module includes a pelvic level generation module, which includes: The fifth vertebral recognition module identifies the coronal plane of the sacrum from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A sacral coronal plane reference line generation module, which is connected to the fifth vertebral recognition module; The sacral coronal plane reference line generation module extracts the left and right endpoints of the sacral coronal plane to establish a sacral coronal plane connection line; The fifth parameter calculation module is connected to the sacral coronal plane reference line generation module; The fifth parameter calculation module calculates the angle between the line connecting the coronal plane of the sacrum and the horizontal reference line to obtain the pelvic level.

[0014] On the other hand, the vectorization module includes: The marker point acquisition module extracts the key points identified during the calculation of the skeletal parameters from the modeling module. A local image extraction module, which is connected to the marker point acquisition module; The local image extraction module extracts multiple related local images from the 3D CT image for the key points, and extracts image features from the related local images respectively. A feature stitching module, which is connected to the local image extraction module; The feature stitching module stitches the image features and the skeletal parameters as vectors along multiple dimensions to obtain the local vector.

[0015] On the other hand, the restoration prediction module includes: The vector fusion module fuses the local vectors corresponding to the clavicle region, spine region and pelvic region respectively to obtain partition vectors, and then concatenates the partition vectors to obtain a fused vector; A vector matching module, which is connected to the vector fusion module; The vector matching module obtains sample vectors with similar directions according to the fusion vector, and generates a scaling ratio between the fusion vector and the sample vector; A vector scaling module, which is connected to the vector matching module; The vector scaling module scales the sample vector according to the scaling ratio to obtain a scaled vector and establishes the restoration prediction model; A prediction module, which is connected to the vector scaling module; The prediction module performs vector difference calculations based on the scaling vector and the local vector to obtain the correspondence of the key points.

[0016] The above technical solution has the following advantages or beneficial effects: To address the issue of insufficient accuracy in the 3D reconstruction of the spine in existing technologies, this solution selects several specific key points in the vertebral region based on 3D reconstruction and calculates skeletal parameters for these key points. This enables effective measurement of relevant data at the key points. Subsequently, a vectorization module is used to vectorize the local image features and skeletal parameters from the 3D reconstruction and establish a vector matrix. This achieves a more accurate representation of the vertebral region, facilitating the prediction of the actual restoration results. Attached Figure Description

[0017] Embodiments of the invention will be described more fully with reference to the accompanying drawings. However, the drawings are for illustration and explanation only and do not constitute a limitation on the scope of the invention.

[0018] Figure 1 This is an overall schematic diagram of an embodiment of the present invention; Figure 2 This is a schematic diagram of the C7PL-CSVL parameter generation module in an embodiment of the present invention; Figure 3 This is a schematic diagram of the clavicle angle generation module in an embodiment of the present invention; Figure 4 This is a schematic diagram of the shoulder height generation module in an embodiment of the present invention; Figure 5 This is a schematic diagram of the T1 vertebral tilt angle generation module in an embodiment of the present invention; Figure 6 This is a schematic diagram of the pelvic level generation module in an embodiment of the present invention; Figure 7 This is a schematic diagram of the vectorization module in an embodiment of the present invention; Figure 8 This is a schematic diagram of the restoration prediction module in an embodiment of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0021] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, but this is not intended to limit the scope of the invention.

[0022] This invention includes: A medical technology interaction system applicable to the spine, such as Figure 1 As shown, it includes: Image acquisition module 1: Image acquisition module 1 acquires three-dimensional CT images of the spinal region and establishes a three-dimensional spinal model; Modeling module 2 is connected to image acquisition module 1; Modeling module 2 calculates skeletal parameters from 3D CT images to characterize key points in the vertebral region; Skeletal parameters consist of C7PL-CSVL parameters, clavicle angle, shoulder height, T1 vertebral tilt angle, pelvic level, and local coronal parameters of scoliosis. Vectorization module 3, which connects to modeling module 2; Vectorization module 3 identifies key points in the 3D vertebral model and adds corresponding skeletal parameters as annotations for the key points. Then, it generates local vectors based on the key points and annotations. Restoration prediction module 4, which is connected to vectorization module 3; The restoration prediction module 4 generates a restoration prediction model for the vertebral region based on local vectors and sample cases, and marks the correspondence between key points before and after adjustment.

[0023] Specifically, addressing the issue of insufficient accuracy in the 3D reconstruction of the spine in existing technologies, this solution selects several key points in the vertebral region based on 3D reconstruction and calculates skeletal parameters for these key points, enabling effective measurement of relevant data at these key points. Subsequently, a vectorization module is used to vectorize the local image features and skeletal parameters from the 3D reconstruction and establish a vector matrix, achieving a more accurate representation of the vertebral region and facilitating the prediction of the actual restoration results.

[0024] During implementation, the aforementioned medical technology interaction system is mainly configured in computer equipment as a software implementation to collect medical images of patients and assist doctors in designing orthopedic plans and orthopedic appliances and providing corresponding design parameters to manufacturers.

[0025] The image acquisition module 1 acquires a sequence of three-dimensional CT images of the patient according to the traditional process, and performs three-dimensional reconstruction based on the three-dimensional CT images to obtain a three-dimensional vertebral model. The reconstruction process is carried out on the entire area where the vertebrae are located, including the ribs, scapula, and pelvis, but the subsequent identification process mainly focuses on the spinal region.

[0026] Subsequently, modeling module 2 calculates skeletal parameters from the 3D CT images to characterize key points in the vertebral region. These key points are selected based on several key points of interest in spinal orthopedics and are used to provide additional data annotations on the basis of the 3D reconstruction model, thereby enhancing the interpretability of the subsequent restoration prediction model.

[0027] Based on this, the vectorization module 3 identifies key points in the three-dimensional vertebral model and adds corresponding skeletal parameters as annotations for the key points. Then, it splices the key points and annotations to form local vectors and organizes them into a local vector matrix. Each vector in this local vector matrix can be used to achieve effective representation of different parts of the patient's spine.

[0028] Finally, the restoration prediction module 4 generates a restoration prediction model for the spinal region based on local vectors and sample cases, and marks the correspondence between key points before and after adjustment for doctors to use.

[0029] As an optional implementation, the aforementioned restoration prediction model can be used to assist doctors in the design process of bone implants, providing the dimensions (graphics), materials, and drawings of the external guide plate for the corrective implant. It can also select suitable standard bone implants from a product library, and then create corresponding printing plans for 3D-printed bone implants. Based on different bone locations and treatment plans, the 3D-printed implants are meshed and layered, forming different layer heights. The effect of the printed product is simulated and viewed through a browser, compared and optimized, and then converted into a printable file format. Simultaneously, bone structures are generated; AIGC designs the 3D bone structure of the cortex and cancellous bone near the original bone injury site, as well as the bone implant material formulation, contact points, and the determination of growth-promoting biomaterials and regenerative cells for bone regeneration. A standard 3D BOM table is generated. It can also provide 3D image analysis of bone injuries and assist in surgical simulation. By using AI recognition and analysis of image data, it can help doctors locate and mark the bone structure connection of important surgical sites, improve surgical outcomes, surgical accuracy, enhance safety, and reduce surgical risks.

[0030] In one embodiment, such as Figure 2 As shown, modeling module 2 includes C7PL-CSVL parameter generation module 21, which includes: The first vertebral recognition module 211 identifies the seventh vertebra and sacrum from the cross section of the three-dimensional vertebral model, and extracts the line connecting the upper end of the pelvis from the three-dimensional vertebral model as a horizontal reference line. Normal line establishment module 212, which is connected to the first spine recognition module 211; The normal line establishment module 212 establishes a first normal line relative to the horizontal reference line for the seventh vertebra, and establishes a second normal line relative to the horizontal reference line for the sacrum; First parameter calculation module 213, the first parameter calculation module 213 is connected to normal establishment module 212; The first parameter calculation module 213 calculates the distance between the first normal and the second normal to obtain the C7PL-CSVL parameter.

[0031] Specifically, to achieve a better representation of the spine, in this embodiment, the C7PL-CSVL parameter is first calculated using the C7PL-CSVL parameter generation module 21. The C7PL-CSVL parameter refers to the horizontal distance between the C7 plumb line (C7PL) and the center sacral vertical line (CSVL).

[0032] Because patients often have symptoms of spinal deformity, it cannot be guaranteed that users will accurately adopt the correct lying position during CT scans. Therefore, in this application, the line connecting the upper ends of the pelvis is uniformly used as the horizontal reference line corresponding to the physical horizontal plane.

[0033] The first vertebral recognition module 211 uses a pre-trained convolutional neural network to identify the seventh vertebra, sacrum, and pelvis from the cross section of the three-dimensional vertebral model, and maps and labels the corresponding bones in the three-dimensional vertebral model according to their corresponding pixels.

[0034] Subsequently, the first spine recognition module 211 extracts the two highest grid face vertices on the left and right sides of the bone region corresponding to the pelvis along the Z-axis in the coordinate system of the three-dimensional spine model, and connects them to form a horizontal reference line.

[0035] Then, the normal establishment module 212 establishes a first normal relative to the plane of the horizontal reference line for the center point of the modeling area corresponding to the seventh vertebra in the coordinate system of the three-dimensional vertebral model, and a second normal relative to the plane of the horizontal reference line for the center point of the modeling area corresponding to the sacrum.

[0036] Finally, the first parameter calculation module 213 projects both the first normal and the second normal onto the tangential plane direction and calculates the distance between the first normal and the second normal to obtain the C7PL-CSVL parameter.

[0037] In one embodiment, such as Figure 3 As shown, modeling module 2 includes clavicle angle generation module 22, which includes: The second vertebral recognition module 221 identifies the left and right clavicles from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. Clavicle reference line generation module 222, which is connected to the second spine recognition module 221; The clavicle reference line generation module 222 measures the highest points of the left and right clavicles relative to the horizontal reference line, establishes the clavicle reference line based on the highest points of the clavicles, and generates the vertex connection line between the left and right clavicles. The second parameter calculation module 223 is connected to the clavicle reference line generation module. The second parameter calculation module 223 calculates the angle between the clavicle reference line and the line connecting the vertices to obtain the clavicle angle.

[0038] Specifically, to achieve a better representation of the spine, in this embodiment, the clavicle angle is first calculated using the clavicle angle generation module 22. Generally, the clavicle reference line (CRL) is the line connecting the highest points of the left and right clavicles along the clavicle horizontal reference line (CHRL). The angle between this line and the clavicle horizontal reference line is the clavicle angle. This angle is positive when the left shoulder is higher and negative when the right shoulder is higher. A clavicle angle greater than 2° indicates shoulder imbalance.

[0039] Because patients often have symptoms of spinal deformity, it cannot be guaranteed that users will accurately adopt the correct lying position during CT scans. Therefore, in this application, the line connecting the upper ends of the pelvis is uniformly used as the horizontal reference line corresponding to the physical horizontal plane.

[0040] The second vertebral recognition module 221 uses a pre-trained convolutional neural network to identify the left and right clavicles and pelvis from the cross section of the three-dimensional vertebral model, and maps and labels the corresponding skeletal regions in the three-dimensional vertebral model according to their corresponding pixels.

[0041] Subsequently, the second vertebral recognition module 221 extracts the two highest grid face vertebrae on the left and right sides of the skeletal region corresponding to the pelvis along the Z-axis in the coordinate system of the three-dimensional vertebral model, and connects them to form a horizontal reference line.

[0042] Then, the clavicle reference line generation module 222 establishes a reference plane on the tangential plane with the horizontal reference line as the reference. Subsequently, it establishes the normals of the left and right clavicle vertices relative to the reference plane on the tangential plane and measures the length of the normals. The vertex with the longer normal length is selected as the highest point of the clavicle, and a clavicle reference line parallel to the reference plane is established around the highest point of the clavicle.

[0043] Additionally, the clavicle reference line generation module 222 generates a clavicle connection line based on the left clavicle vertex and the right clavicle vertex.

[0044] Finally, the second parameter calculation module 223 generates the clavicle angle based on the angle between the clavicle reference line and the clavicle line on the tangential plane.

[0045] In one embodiment, such as Figure 4 As shown, modeling module 2 includes a shoulder height generation module 23, which includes: The third vertebral recognition module 231 identifies the left and right clavicles from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. Shoulder height reference line generation module 232, which is connected to the third vertebral recognition module; The shoulder height reference line generation module 232 establishes a left shoulder horizontal reference line parallel to the horizontal reference line for the highest point of the left clavicle, and establishes a right shoulder horizontal reference line parallel to the horizontal reference line for the highest point of the right clavicle. The third parameter calculation module 233 is connected to the shoulder height reference line generation module 232. The third parameter calculation module 233 calculates the distance between the horizontal reference lines of the left and right shoulders to obtain the height of both shoulders.

[0046] Specifically, to achieve a better representation of the spine, in this embodiment, the shoulder height is first calculated using the shoulder height generation module 23. Generally, a vertical line is drawn through each acromioclavicular joint, intersecting the shoulder soft tissue at a single point. A horizontal line is then drawn through this point; the higher side is defined as the superior horizontal reference line (SHRL), and the lower side as the inferior horizontal reference line (IHRL). The vertical distance between the two lines is the radiographic shoulder height. A vertical distance of less than 1 cm between the SHRL and IHRL indicates shoulder balance; 1-2 cm indicates mild shoulder imbalance; 2-3 cm indicates moderate shoulder imbalance; and greater than 3 cm indicates severe shoulder imbalance.

[0047] Because patients often have symptoms of spinal deformity, it cannot be guaranteed that users will accurately adopt the correct lying position during CT scans. Therefore, in this application, the line connecting the upper ends of the pelvis is uniformly used as the horizontal reference line corresponding to the physical horizontal plane.

[0048] The third vertebral recognition module 231 uses a pre-trained convolutional neural network to identify the left and right scapulae and pelvis from the back section of the three-dimensional vertebral model, and maps and labels the corresponding skeletal regions in the three-dimensional vertebral model according to their corresponding pixels.

[0049] Subsequently, the third vertebral recognition module 231 extracts the two highest grid face vertebrae on the left and right sides of the skeletal region corresponding to the pelvis along the Z-axis in the coordinate system of the three-dimensional vertebral model, and connects them to form a horizontal reference line.

[0050] Then, the shoulder height reference line generation module 232 establishes a reference plane on the back section with the horizontal reference line as the reference. Subsequently, it maps the left clavicle apex and the right clavicle apex on the back section and establishes the left shoulder horizontal reference line and the right shoulder horizontal reference line parallel to the reference plane.

[0051] Finally, the third parameter calculation module 233 generates the shoulder height based on the distance between the left shoulder horizontal reference line and the right shoulder horizontal reference line on the back section.

[0052] In one embodiment, such as Figure 5 As shown, modeling module 2 includes a T1 vertebral body tilt angle generation module 24, which includes: The fourth vertebral recognition module 241 identifies the first thoracic vertebra from the cross section of the three-dimensional vertebral model and extracts the upper end line of the pelvis from the three-dimensional vertebral model as a horizontal reference line. Thoracic vertebra reference line generation module 242, which is connected to the fourth vertebra recognition module 241; The thoracic vertebra reference line generation module 242 extracts the left and right endpoints of the first thoracic vertebra to establish a line connecting the first thoracic vertebrae. The fourth parameter calculation module 243 is connected to the thoracic spine reference line generation module; The fourth parameter calculation module 243 calculates the angle between the line connecting the first thoracic vertebra and the horizontal reference line to obtain the clavicle angle.

[0053] Specifically, to achieve a better representation of the spine, in this embodiment, the T1 vertebral body tilt angle is first calculated using the T1 vertebral body tilt angle generation module 24. In traditional measurement processes, a straight line is usually drawn around the superior endplate of the first thoracic vertebra, and the angle between this line and the horizontal line is the T1 vertebral body tilt angle.

[0054] Because patients often have symptoms of spinal deformity, it cannot be guaranteed that users will accurately adopt the correct lying position during CT scans. Therefore, in this application, the line connecting the upper ends of the pelvis is uniformly used as the horizontal reference line corresponding to the physical horizontal plane.

[0055] The fourth vertebral recognition module 241 uses a pre-trained convolutional neural network to identify multiple thoracic vertebrae and pelvis from the back section of the 3D vertebral model, and maps and labels the corresponding skeletal regions in the 3D vertebral model according to their corresponding pixels. Then, in the coordinate system of the 3D vertebral model, the one with the highest height in the Z-axis direction is extracted as the first thoracic vertebrae identified.

[0056] Subsequently, the fourth vertebral recognition module 241 extracts the two highest grid face vertebrae on the left and right sides of the skeletal region corresponding to the pelvis along the Z-axis in the coordinate system of the three-dimensional vertebral model, and connects them to form a horizontal reference line.

[0057] Then, the thoracic vertebra reference line generation module 242 first rotates the image on the tangential plane, and uses the normal direction of the horizontal reference line as the new Z-axis. Then, it extracts the left and right endpoints of the first thoracic vertebral body on the tangential plane to establish the connection line of the first thoracic vertebral body.

[0058] Finally, the fourth parameter calculation module 243 calculates the tilt angle of the T1 vertebra based on the line connecting the first thoracic vertebrae and the horizontal reference line.

[0059] In one embodiment, such as Figure 6 As shown, modeling module 2 includes a pelvic level generation module 25, which includes: The fifth vertebral recognition module 251 identifies the coronal plane of the sacrum from the tangential plane of the three-dimensional vertebral model and extracts the line connecting the upper end of the pelvis from the three-dimensional vertebral model as a horizontal reference line. Sacral coronal plane reference line generation module 252, which is connected to the fifth vertebra recognition module; The sacral coronal plane reference line generation module 252 extracts the left and right endpoints of the sacral coronal plane to establish the sacral coronal plane connection line; The fifth parameter calculation module 253 is connected to the sacral coronal plane reference line generation module 252; The fifth parameter calculation module 253 calculates the angle between the line connecting the coronal plane of the sacrum and the horizontal reference line to obtain the pelvic level.

[0060] Specifically, in order to achieve a better representation of the spine, in this embodiment, the pelvic level is first calculated by the pelvic level generation module 25.

[0061] Among them, the fifth vertebral recognition module 251 uses a pre-trained convolutional neural network to identify the sacrum and pelvis from the back section of the three-dimensional vertebral model, and maps and marks the corresponding skeletal regions in the three-dimensional vertebral model according to the corresponding pixels.

[0062] Subsequently, the fifth vertebral recognition module 251 extracts the two highest grid face vertebrae on the left and right sides of the skeletal region corresponding to the pelvis along the Z-axis in the coordinate system of the three-dimensional vertebral model, and connects them to form a horizontal reference line.

[0063] Then, the sacral coronal plane reference line generation module 252 extracts the left and right endpoints of the sacral coronal plane on the three-dimensional reconstruction model to establish the sacral coronal plane connection line.

[0064] Finally, the fourth parameter calculation module 243 calculates the pelvic level based on the angle between the line connecting the coronal plane of the sacrum and the horizontal reference line.

[0065] Based on the identification of the coronal plane, the coronal scoliosis angle parameter, i.e., COBB intersection, can be further calculated. Also, the spinal master curve is fitted in the model around the previously extracted thoracic vertebrae. The master curve angle, compensation curve angle and rotation degree are calculated based on the spinal master curve as local coronal plane parameters of scoliosis.

[0066] In one embodiment, such as Figure 7 As shown, vectorization module 3 includes: The marker point acquisition module 31 extracts key points identified during the calculation of skeletal parameters from the self-modeling module. Local image extraction module 32, which is connected to marker point acquisition module 31; The local image extraction module 32 extracts multiple related local images from the 3D CT image for key points, and extracts image features from the related local images respectively; Feature stitching module 33, which is connected to local image extraction module 32; The feature stitching module 33 stitches together multiple dimensions of image features and skeletal parameters as vectors to obtain local vectors.

[0067] Specifically, to achieve better annotation of the patient's spine, in this embodiment, while the modeling module 2 annotates key areas such as the thoracic vertebrae, scapula, pelvis, and sacrum, the marker point acquisition module 31 extracts key points identified during the calculation of skeletal parameters from the modeling module. Subsequently, the local image extraction module 32 extracts multiple related local images from the 3D CT images for each key point and feeds them into a 3D U-net network for feature extraction, thereby describing the features of each key point of the spine at the image level. Finally, the feature stitching module stitches together the image features and skeletal parameters of each key point as multiple dimensions of a vector to obtain a local vector corresponding to that location. The extracted local vectors are organized as vector matrices and have coordinate ranges corresponding to the 3D reconstruction model added. This vector matrix approach achieves a better representation of the patient's 3D reconstruction model.

[0068] In one embodiment, such as Figure 8 As shown, the restoration prediction module 4 includes: Vector fusion module 41 fuses the local vectors corresponding to the clavicle region, spine region and pelvic region respectively to obtain partition vectors, and then concatenates the partition vectors to obtain the fused vector; Vector matching module 42, which is connected to vector fusion module 41; The vector matching module 42 obtains sample vectors with similar directions based on the fusion vector matching, and generates a scaling ratio between the fusion vector and the sample vector; Vector scaling module 43, which is connected to vector matching module 42; The vector scaling module 43 scales the sample vector according to the scaling ratio to obtain the scaled vector and establishes the restoration prediction model; Prediction module 44, which is connected to vector scaling module 43; The prediction module 44 calculates the vector difference based on the scaling vector and the local vector to obtain the correspondence of key points.

[0069] Specifically, in order to achieve better restoration and prediction results, in this embodiment, the vector fusion module 41 first performs further partitioning and splicing processing on the above-mentioned vector matrix based on the three-dimensional reconstruction model. By splicing the local vectors corresponding to the clavicle region, spine region and pelvic region respectively, partition vectors are obtained, thereby reducing the vector dimension that needs to be matched.

[0070] Subsequently, the partition vectors are concatenated as multiple dimensions of the fusion vector. The vector matching module 42 is used to match sample vectors with similar directions according to the fusion vector. The scaling ratio between the fusion vector and the sample vector is generated according to the relative spacing between the three partition vectors contained in the fusion vector and the relative spacing in the sample vector.

[0071] Finally, the vector scaling module 43 scales the sample vector according to the scaling ratio to obtain the scaled vector and establishes the restoration prediction model. The prediction module 44 calculates the vector difference based on the scaled vector and the local vector to obtain the correspondence of key points.

[0072] The above are merely preferred embodiments of the present invention and are not intended to limit the implementation methods and protection scope of the present invention. Those skilled in the art should recognize that any equivalent substitutions and obvious changes made based on the description and illustrations of the present invention should be included within the protection scope of the present invention.

Claims

1. A medical technology interaction system suitable for the spine, characterized in that, include: The image acquisition module acquires three-dimensional CT images of the spinal region and establishes a three-dimensional spinal model. A modeling module, which is connected to the image acquisition module; The modeling module calculates skeletal parameters from the three-dimensional CT images to characterize key points in the vertebral region. The skeletal parameters consist of C7PL-CSVL parameters, clavicle angle, shoulder height, T1 vertebral tilt angle, pelvic level, and local coronal parameters of scoliosis. A vectorization module, which is connected to the modeling module; The vectorization module identifies key points in the three-dimensional vertebral model and adds corresponding skeletal parameters as annotations for the key points. Then, it generates local vectors based on the key points and the annotations. A restoration prediction module, which is connected to the vectorization module; The restoration prediction module generates a restoration prediction model for the vertebral region based on the local vectors and sample cases, and marks the correspondence between key points before and after adjustment.

2. The medical technology interaction system according to claim 1, characterized in that, The modeling module includes a C7PL-CSVL parameter generation module, which includes: The first vertebral recognition module identifies the seventh vertebra and sacrum from the cross section of the three-dimensional vertebral model, and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A normal mapping module, which is connected to the first spine recognition module; The normal establishment module establishes a first normal line for the seventh vertebra relative to the horizontal reference line, and establishes a second normal line for the sacrum relative to the horizontal reference line; A first parameter calculation module is connected to the normal establishment module; The first parameter calculation module calculates the distance between the first normal and the second normal to obtain the C7PL-CSVL parameter.

3. The medical technology interaction system according to claim 1, characterized in that, The modeling module includes a clavicle angle generation module, which includes: The second vertebral recognition module identifies the left and right clavicles from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A clavicle reference line generation module, wherein the clavicle reference line generation module is connected to the second spine recognition module; The clavicle reference line generation module measures the highest points of the left and right clavicles relative to the horizontal reference line, establishes a clavicle reference line based on the highest points of the clavicles, and generates a line connecting the vertices of the left and right clavicles. The second parameter calculation module is connected to the clavicle reference line generation module. The second parameter calculation module calculates the angle between the clavicle reference line and the line connecting the vertices to obtain the clavicle angle.

4. The medical technology interaction system according to claim 1, characterized in that, The modeling module includes a shoulder height generation module, which includes: The third vertebral recognition module identifies the left and right clavicles from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A shoulder height reference line generation module, which is connected to the third vertebral recognition module; The shoulder height reference line generation module establishes a left shoulder horizontal reference line parallel to the horizontal reference line for the highest point of the left clavicle, and establishes a right shoulder horizontal reference line parallel to the horizontal reference line for the highest point of the right clavicle. The third parameter calculation module is connected to the shoulder height reference line generation module; The third parameter calculation module calculates the distance between the horizontal reference line of the left shoulder and the horizontal reference line of the right shoulder to obtain the height of both shoulders.

5. The medical technology interaction system according to claim 1, characterized in that, The modeling module includes a T1 vertebral tilt angle generation module, which includes: The fourth vertebral recognition module identifies the first thoracic vertebra from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A thoracic vertebra reference line generation module, which is connected to the fourth vertebral recognition module; The thoracic vertebra reference line generation module extracts the left and right endpoints of the first thoracic vertebra to establish a line connecting the first thoracic vertebrae. The fourth parameter calculation module is connected to the thoracic vertebra reference line generation module; The fourth parameter calculation module calculates the angle between the line connecting the first thoracic vertebrae and the horizontal reference line to obtain the clavicle angle.

6. The medical technology interaction system according to claim 1, characterized in that, The modeling module includes a pelvic level generation module, which includes: The fifth vertebral recognition module identifies the coronal plane of the sacrum from the cross section of the three-dimensional vertebral model and extracts the line connecting the upper ends of the pelvis from the three-dimensional vertebral model as a horizontal reference line. A sacral coronal plane reference line generation module, which is connected to the fifth vertebral recognition module; The sacral coronal plane reference line generation module extracts the left and right endpoints of the sacral coronal plane to establish a sacral coronal plane connection line; The fifth parameter calculation module is connected to the sacral coronal plane reference line generation module; The fifth parameter calculation module calculates the angle between the line connecting the coronal plane of the sacrum and the horizontal reference line to obtain the pelvic level.

7. The medical technology interaction system according to claim 1, characterized in that, The vectorization module includes: The marker point acquisition module extracts the key points identified during the calculation of the skeletal parameters from the modeling module. A local image extraction module, which is connected to the marker point acquisition module; The local image extraction module extracts multiple related local images from the 3D CT image for the key points, and extracts image features from the related local images respectively. A feature stitching module, which is connected to the local image extraction module; The feature stitching module stitches the image features and the skeletal parameters as vectors along multiple dimensions to obtain the local vector.

8. The medical technology interaction system according to claim 1, characterized in that, The restoration prediction module includes: The vector fusion module fuses the local vectors corresponding to the clavicle region, spine region and pelvic region respectively to obtain partition vectors, and then concatenates the partition vectors to obtain a fused vector; A vector matching module, which is connected to the vector fusion module; The vector matching module obtains sample vectors with similar directions according to the fusion vector, and generates a scaling ratio between the fusion vector and the sample vector; A vector scaling module, which is connected to the vector matching module; The vector scaling module scales the sample vector according to the scaling ratio to obtain a scaled vector and establishes the restoration prediction model; A prediction module, which is connected to the vector scaling module; The prediction module performs vector difference calculations based on the scaling vector and the local vector to obtain the correspondence of the key points.

Images