Method and system for measuring spine parameters and automatically typing spine

Abstract

The invention discloses a spine parameter measurement and spine automatic typing method comprising the following steps: obtaining four spine original images of a to-be-detected target in different directions, preprocessing the spine original images, and sending the preprocessed spine original images into an image neural network to obtain key information in the images; and performing parameter measurement and analysis according to the key information to obtain various parameters. And calculating according to the lateral bending Cobb angle, the segment and the sagittal posterior convex angle to obtain a lateral bending type. And judging the relative position relationship between the sacrum perpendicular bisector CSVL and the vertebral pedicle mask to obtain the corrected type of the lumbar curvature. And judging according to the thoracic vertebra posterior lobe of the sagittal plane to obtain a thoracic vertebra sagittal plane correction type. And combining the lateral bending type, the waist bending correction type parting and the thoracic vertebra sagittal plane correction type to obtain a complete Lenke parting. Through automatic identification and automatic calculation, the popularization rate of spine radiation parameter measurement can be effectively improved without professional requirements and thresholds, and the method has very high clinical practical significance and popularization significance.

Description

Technical Field

[0001] The present invention belongs to the field of medical image processing, and in particular relates to a method and system for measuring spinal parameters and automatically classifying the spinal column. Background Art

[0002] The coronal, sagittal or axial positions of the spine deviate from the normal position, resulting in morphological abnormalities, which are called spinal deformities. Spinal deformities can be divided into many types, each of which is suitable for different causes. For the classification of spinal deformities, the King classification was initially the most commonly used scoliosis classification. However, since the King classification is based on a single-plane classification method, and scoliosis is a complex three-dimensional composite deformity, the King classification is far from meeting actual clinical needs. Subsequently, in 2001, Lenke et al. proposed the Lenke classification, which is used to classify complex three-dimensional spinal morphologies. It is also an internationally recognized classification method. Currently, the Lenke classification is the most commonly used scoliosis classification method in clinical practice and is a relatively mature classification.

[0003] The Lenke classification is a very detailed classification method. First of all, the spine must be divided into zones. The classification consists of three parts: curvature type, lumbar curvature correction type, and thoracic sagittal plane correction type. In the curvature type, the spine is divided into upper thoracic curvature, main thoracic curvature, thoracolumbar curvature, and lumbar curvature. The nature of the curvature is divided into primary curvature (structural curvature) and secondary curvature (structural and non-structural). It is thus subdivided into 42 types. According to the concept of Lenke classification, different types of surgical methods are very different. Some require long-segment fixation, and some require short segments. In general, this classification gives clinicians a yardstick for treating idiopathic scoliosis, which has important reference and guiding value for the selection of surgical procedures.

[0004] However, the complicated classification process of Lenke classification also brings a lot of troubles to clinical practice. To complete a Lenke classification, four spinal X-rays are required: anteroposterior, lateral, and anteroposterior left and right dynamic positions. In addition, the four X-rays need to be measured for many parameters, such as Cobb angle, CSVL, T5-12 kyphosis angle, and pedicle. Clinicians often need to spend a lot of time and energy to complete the classification results and decide on the specific surgical method based on the classification. Summary of the invention

[0005] The technical purpose of the present invention is to provide a method and system for measuring spinal parameters and automatically classifying the spine, so as to solve the problem of low efficiency and large workload of manual Lenke classification of the spine.

[0006] To solve the above problems, the technical solution of the present invention is: A method for measuring spinal parameters and automatically classifying the spine comprises the following steps: Obtain 4 original spinal images of the target to be detected in different directions, namely, the anteroposterior direction image, the lateral direction image, and the left / right dynamic direction image. Preprocess the original spinal images and then send them into the pre-trained image neural network to obtain key information in the images, including vertebral order, vertebral angle points, and pedicle mask coordinate information; Parameters are measured and analyzed based on key information, including the angle between adjacent vertebrae, the overall shape of the spine in the positive position, the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle, the apical vertebra or apical intervertebral disc between the end vertebrae, the sagittal kyphosis angle and the CSVL coordinate of the sacral median vertical line; According to the size of the Cobb angle of scoliosis, the size of the segmental and sagittal kyphosis angles, it can be calculated whether the scoliosis of each segment is the main curve and whether it is a structural curve, and the type of scoliosis can be obtained; The relative position relationship is determined based on the CSVL coordinates of the sacral median vertical line and the pedicle mask coordinates, and the modified lumbar curvature classification is obtained. According to the size of the thoracic kyphosis angle in the sagittal plane, the sagittal plane correction type of the thoracic spine is obtained; The complete Lenke classification is obtained by combining the scoliosis curve type, the modified lumbar curve classification and the modified thoracic sagittal plane classification.

[0007] Among them, the pre-training of the image neural network is specifically as follows: collecting a number of spinal images as a training image data set, pre-processing the spinal images in the training image data set, marking the pre-processed spine at the vertebral position, vertebral sequence, vertebral corner points and pedicle mask, and constructing a spinal corner point recognition model based on the marks in the image; inputting the standardized spinal model into the spinal corner point recognition model for training, and constructing a vertebral recognition and segmentation model as a pre-trained image neural network.

[0008] The image preprocessing specifically includes: adjusting the window width and window position, contrast enhancement, and image size standardization of the image data in sequence.

[0009] The angle calculation between adjacent vertebrae is specifically as follows: for all vertebrae in the anteroposterior thoracic spine and the anteroposterior lumbar spine, the vertebrae angles between two adjacent vertebrae are calculated, the center points on the left and right sides of the vertebrae are calculated to form a center point vector, and then the angles between the center point vectors of all vertebrae in the anteroposterior thoracic spine and the anteroposterior lumbar spine are calculated respectively; Then, the overall shape of the anteroposterior spinal part is judged, specifically: the slopes of the center point vectors of the thoracic vertebra T1 and the lumbar vertebra L5 are calculated respectively; when the slopes of the center point vectors of the thoracic vertebra T1 and the lumbar vertebra L5 are one positive and one negative, the overall shape of the anteroposterior spinal part is not S-shaped, otherwise the overall shape is S-shaped.

[0010] Among them, the size of the scoliosis Cobb angle and the end vertebra corresponding to the Cobb angle are calculated based on the overall shape of the spine in the positive position: The four terminal vertebrae from top to bottom of the spine are named the first terminal vertebra V1, the second terminal vertebra V2, the third terminal vertebra V3 and the fourth terminal vertebra V4; When the overall shape of the spine is not S-shaped: take the thoracic vertebra T1 as the first-order end vertebra V1, take the upper end vertebra of the first Cobb angle as the second-order end vertebra V2, and calculate the angle between the first-order end vertebra V1 and the second-order end vertebra V2 as the second Cobb angle; take the lumbar vertebra L5 as the fourth-order end vertebra V4, take the lower end vertebra of the first Cobb angle as the third-order end vertebra V3, and calculate the angle between the fourth-order end vertebra V4 and the third-order end vertebra V3 as the third Cobb angle; When the overall shape of the spine is S-shaped: determine the positions of the upper end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle in the image data. If the upper end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle are located in the upper half of the image data, calculate the vertebral angles between the thoracic vertebra T1 and the upper end vertebra of the first Cobb angle and the upper end vertebra of the first Cobb angle, and use the maximum value of the calculated vertebral angles as the second Cobb angle; and make the upper end vertebra of the second Cobb angle the first-order end vertebra V1, and the upper end vertebra of the first Cobb angle the second-order end vertebra V2; then calculate the vertebral angles between the lumbar vertebra L5 and the lower end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle, and use the maximum value of the calculated vertebral angles as the third Cobb angle; and make the lower end vertebra of the first Cobb angle the third-order end vertebra V3, and the lower end vertebra corresponding to the third Cobb angle the fourth-order end vertebra V4; If the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the lower half of the image data, the vertebral angles are calculated for all vertebrae between the thoracic vertebra T1 and the upper vertebra of the first Cobb angle, and the maximum value of the calculated vertebral angles is used as the second Cobb angle; the upper vertebra corresponding to the second Cobb angle is used as the second-order end vertebra V2, and the upper vertebra of the first Cobb angle is used as the third-order end vertebra V3; then the vertebral angles are calculated for all vertebrae between the thoracic vertebra T1 and the second-order end vertebra V2, and the second-order end vertebra V2 is used as the third Cobb angle; the upper vertebra of the third Cobb angle is used as the first-order end vertebra V1, and the lower vertebra of the first Cobb angle is used as the fourth-order end vertebra V4.

[0011] The positions of the upper vertebrae of the first Cobb angle and the lower vertebrae of the first Cobb angle in the image data are determined as follows: the center coordinates of the two corner points below the upper vertebrae of the first Cobb angle and the lower vertebrae of the first Cobb angle are recorded as P1 and P2, respectively, and the center coordinates of the two corner points below the thoracic vertebra T1 and the lumbar vertebra L5 are recorded as P3 and P4, respectively; When (P1+P2)<1.2*(P3+P4) is satisfied, the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the upper half of the image data; When (P1+P2)>=1.2*(P3+P4) is satisfied, the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the lower half of the image data.

[0012] Among them, the top vertebra or top intervertebral disc between the end vertebrae is calculated in turn, specifically: Connect the midpoints of the first-order end vertebra V1, the second-order end vertebra V2, the third-order end vertebra V3 and the fourth-order end vertebra V4 from top to bottom to form a straight line connecting two adjacent end vertebrae; then calculate the straight-line distances between the vertebral bodies and intervertebral discs between the first-order end vertebra V1 and the second-order end vertebra V2, the second-order end vertebra V2 and the third-order end vertebra V3, and the third-order end vertebra V3 and the fourth-order end vertebra V4; between the two end vertebrae, the vertebral body or intervertebral disc with the longest straight-line distance is the top vertebra or top intervertebral disc corresponding to the current two end vertebrae.

[0013] Among them, the calculation of the sagittal kyphosis angle is as follows: According to the center coordinates of the two upper corner points of thoracic vertebra T2 and the center coordinates of the two lower corner points of thoracic vertebra T5 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T2 and the line connecting the coordinates of the lower corner point of T5 is calculated to obtain the T2-T5 kyphosis angle; According to the center coordinates of the two upper corner points of thoracic vertebra T5 and the center coordinates of the two lower corner points of thoracic vertebra T12 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T5 and the line connecting the coordinates of the lower corner point of T12 is calculated to obtain the T5-T12 kyphosis angle; According to the center coordinates of the two upper corner points of thoracic vertebra T10 and the center coordinates of the two lower corner points of thoracic vertebra L2 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T10 and the line connecting the coordinates of the lower corner point of L2 is calculated to obtain the size of the T10-L2 kyphosis angle.

[0014] The specific method for obtaining the CSVL coordinates of the sacral perpendicular line is: The center point coordinates of the left and right corner points of the sacrum S1 in the anteroposterior image data are calculated to obtain the midpoint coordinates of the sacrum S1, and the horizontal coordinates of the midpoint coordinates are fixed to obtain the coordinates corresponding to the sacral vertical line CSVL.

[0015] A system for measuring spine parameters and automatically classifying the spine, which implements the above-mentioned method for measuring spine parameters and automatically classifying the spine, comprises: The image neural network recognition module is configured to obtain four original spinal images of the target to be detected in different directions, namely, the anteroposterior direction image, the lateral direction image, and the left / right dynamic direction image, and then pre-process the original spinal images and perform neural network training to obtain key information in the images, including vertebral order, vertebral angle points, and pedicle mask coordinate information; A parameter measurement module is configured to perform parameter measurement and analysis based on key information, including angles between adjacent vertebrae, the overall shape of the spinal column in the anteroposterior position, the size of the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle, the apical vertebra or apical intervertebral disc between the end vertebrae, the size of the sagittal kyphosis angle, and the CSVL coordinates of the sacral median vertical line; The Lenke classification parameter analysis module is configured to calculate based on the Cobb angle of scoliosis, the segmental and sagittal kyphosis angles, and can determine whether the scoliosis of each segment is a major curve and whether it is a structural curve, and determine the type of scoliosis; The relative position relationship is determined based on the CSVL coordinates of the sacral median vertical line and the pedicle mask coordinates, and the modified lumbar curvature classification is obtained. According to the size of the thoracic kyphosis angle in the sagittal plane, the sagittal plane correction type of the thoracic spine is obtained; The complete Lenke classification is obtained by combining the scoliosis curve type, the modified lumbar curve classification and the modified thoracic sagittal plane classification.

[0016] Due to the adoption of the above technical solution, the present invention has the following advantages and positive effects compared with the prior art: The present invention uses an artificial intelligence neural network algorithm to process medical images and accurately identify various anatomical landmarks of the spine and the contours of the vertebrae. Compared with manual measurement, it has high measurement accuracy, lower error rate, shorter time consumption, and can process data in large quantities, which can effectively reduce medical costs.

[0017] In addition, the present invention can effectively improve the popularity of spinal radiographic parameter measurement through automatic identification and automatic calculation without professional requirements and thresholds, and has strong clinical practical significance and promotion significance. Through parameter measurement, the Lenke classification is automatically obtained to assist clinical evaluation of spinal conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the following detailed description of the preferred embodiment.The drawings are only for the purpose of illustrating the preferred embodiments and are not to be construed as limiting the invention.

[0019] Figure 1 The figure is a flow chart of a method for measuring spine parameters and automatically classifying the spine according to the present invention. DETAILED DESCRIPTION

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the specific implementation methods of the present invention will be described below with reference to the accompanying drawings. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings and other implementation methods can be obtained based on these drawings without creative work.

[0021] In order to simplify the drawings, only the parts related to the present invention are schematically shown in each figure, and they do not represent the actual structure of the product. In addition, in order to simplify the drawings and facilitate understanding, in some figures, only one of the parts with the same structure or function is schematically drawn or marked. In this article, "one" not only means "only one", but also means "more than one".

[0022] The following is a further detailed description of a method and system for measuring spinal parameters and automatically classifying the spine proposed by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description and claims.

[0023] See also Figure 1 This embodiment provides a method for measuring spinal parameters and automatically classifying the spine. Before executing this embodiment, the image neural network needs to be trained, specifically: a large number of spinal images are collected as a training image data set, and the spinal images in the training image data set are preprocessed, including but not limited to window width and window level adjustment, contrast enhancement, and image size standardization. The preprocessed spinal images are then marked at the vertebral position, vertebral sequence, vertebral corner points, and pedicle masks, and a spinal corner point recognition model is constructed based on the marks in the image. The function of this model is to automatically mark the images that are subsequently input. Next, the standardized spinal model is input into the spinal corner point recognition model for training, and a vertebral recognition and division model is constructed as a pre-trained image neural network.

[0024] Next, execute this example, and obtain 4 original spinal images of the target to be detected in different directions through the corresponding image acquisition equipment, which are the anteroposterior image, the lateral image, and the left / right dynamic image. Among them, the anteroposterior direction refers to the patient lying on his back, and the X-ray is projected from the front of the patient to the detector behind, so that the front view of the bones and joints can be observed. The lateral direction refers to the patient lying on his side, and the X-ray is projected from one side of the patient to the detector on the other side, so that the side view of the bones and joints can be observed. The left / right dynamic position refers to the X-ray film taken when the patient performs certain actions. The dynamic film can show the stability and flexibility of the spine during movement. Then, the original spinal image is also preprocessed, including but not limited to window width and window position adjustment, contrast enhancement, and image size standardization. Then the preprocessed image is sent to the previously established vertebral recognition and segmentation model to obtain key information in the image, including vertebral order, vertebral angle point, and pedicle mask coordinate information.

[0025] Next, the key information obtained in the previous step is measured and analyzed, including the angle between adjacent vertebrae, the overall shape of the spine in the positive position, the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle, the apical vertebra or apical intervertebral disc between the end vertebrae, the sagittal kyphosis angle and the CSVL coordinate of the sacral perpendicular line.

[0026] The acquisition of each parameter and the corresponding analysis are explained below: First, the angle calculation between adjacent vertebrae is as follows: for all vertebrae in the anteroposterior thoracic spine and anteroposterior lumbar spine, the vertebrae angles between two adjacent vertebrae are calculated, the center points on the left and right sides of the vertebrae are calculated to form a center point vector, and then the angles between the center point vectors of all vertebrae in the anteroposterior thoracic spine and anteroposterior lumbar spine are calculated respectively.

[0027] Then, the overall shape of the spine in the anteroposterior position is judged, specifically: the slopes of the center point vectors of thoracic vertebra T1 and lumbar vertebra L5 (thoracic vertebra T1 refers to the first thoracic vertebra, and lumbar vertebra L5 refers to the fifth lumbar vertebra) are calculated respectively; when the slopes of the center point vectors of thoracic vertebra T1 and lumbar vertebra L5 are one positive and one negative, it means that the overall shape of the spine in the anteroposterior position is not S-shaped, otherwise the overall shape is S-shaped.

[0028] Then, the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle are calculated according to the overall shape of the spine in the positive position: the four end vertebrae from top to bottom of the spine are named the first end vertebra V1, the second end vertebra V2, the third end vertebra V3 and the fourth end vertebra V4.

[0029] When the overall shape of the spine is not S-shaped: take the thoracic vertebra T1 as the first-order end vertebra V1, and the upper end vertebra of the first Cobb angle as the second-order end vertebra V2, and calculate the angle between the first-order end vertebra V1 and the second-order end vertebra V2 as the second Cobb angle; take the lumbar vertebra L5 as the fourth-order end vertebra V4, and the lower end vertebra of the first Cobb angle as the third-order end vertebra V3, and calculate the angle between the fourth-order end vertebra V4 and the third-order end vertebra V3 as the third Cobb angle.

[0030] When the overall shape of the spine is S-shaped: determine the positions of the upper end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle in the image data. If the upper end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle are located in the upper half of the image data, calculate the vertebral angles between the thoracic vertebra T1 and the upper end vertebra of the first Cobb angle and the upper end vertebra of the first Cobb angle, and use the maximum value of the calculated vertebral angles as the second Cobb angle; and make the upper end vertebra of the second Cobb angle the first-order end vertebra V1, and the upper end vertebra of the first Cobb angle the second-order end vertebra V2; then calculate the vertebral angles between the lumbar vertebra L5 and the lower end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle, and use the maximum value of the calculated vertebral angles as the third Cobb angle; and make the lower end vertebra of the first Cobb angle the third-order end vertebra V3, and the lower end vertebra corresponding to the third Cobb angle the fourth-order end vertebra V4.

[0031] If the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the lower half of the image data, the vertebral angles are calculated for all vertebrae between the thoracic vertebra T1 and the upper vertebra of the first Cobb angle, and the maximum value of the calculated vertebral angles is used as the second Cobb angle; the upper vertebra corresponding to the second Cobb angle is used as the second-order end vertebra V2, and the upper vertebra of the first Cobb angle is used as the third-order end vertebra V3; then the vertebral angles are calculated for all vertebrae between the thoracic vertebra T1 and the second-order end vertebra V2, and the second-order end vertebra V2 is used as the third Cobb angle; the upper vertebra of the third Cobb angle is used as the first-order end vertebra V1, and the lower vertebra of the first Cobb angle is used as the fourth-order end vertebra V4.

[0032] Furthermore, the positions of the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle in the image data are determined specifically as follows: the center coordinates of the two corner points below the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are recorded as P1 and P2, respectively, and the center coordinates of the two corner points below the thoracic vertebra T1 and the lumbar vertebra L5 are recorded as P3 and P4, respectively. When (P1+P2)<1.2*(P3+P4) is satisfied, the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the upper half of the image data. When (P1+P2)>=1.2*(P3+P4) is satisfied, the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the lower half of the image data.

[0033] Then, the top vertebra or top intervertebral disc between the end vertebrae is calculated in turn, specifically: the midpoint of the first-order end vertebra V1, the midpoint of the second-order end vertebra V2, the midpoint of the third-order end vertebra V3 and the midpoint of the fourth-order end vertebra V4 are connected in order from top to bottom to form a straight line connecting two adjacent end vertebrae; then the straight-line distances between the vertebral bodies and intervertebral discs between the first-order end vertebra V1 and the second-order end vertebra V2, the second-order end vertebra V2 and the third-order end vertebra V3, and the third-order end vertebra V3 and the fourth-order end vertebra V4 are calculated respectively; between the two end vertebrae, the vertebral body or intervertebral disc with the longest straight-line distance is the top vertebra or top intervertebral disc corresponding to the current two end vertebrae.

[0034] Among them, the calculation of the sagittal kyphosis angle is as follows: According to the center coordinates of the two upper corner points of thoracic vertebra T2 and the center coordinates of the two lower corner points of thoracic vertebra T5 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T2 and the line connecting the coordinates of the lower corner point of T5 is calculated to obtain the size of the T2-T5 kyphosis angle.

[0035] According to the center coordinates of the two upper corner points of thoracic vertebra T5 and the center coordinates of the two lower corner points of thoracic vertebra T12 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T5 and the line connecting the coordinates of the lower corner point of T12 is calculated to obtain the size of the T5-T12 kyphosis angle.

[0036] According to the center coordinates of the two upper corner points of thoracic vertebra T10 and the center coordinates of the two lower corner points of thoracic vertebra L2 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T10 and the line connecting the coordinates of the lower corner point of L2 is calculated to obtain the size of the T10-L2 kyphosis angle.

[0037] The specific steps of obtaining the coordinates of the sacral median perpendicular line CSVL are as follows: the center point coordinates of the left and right corner points of the sacrum S1 in the anteroposterior image data are calculated to obtain the midpoint coordinates of the sacrum S1, and the horizontal coordinates of the midpoint coordinates are fixed to obtain the coordinates corresponding to the sacral median perpendicular line CSVL.

[0038] After the above parameters are obtained, the Cobb angle of scoliosis, the segmental and sagittal kyphosis angles are calculated to determine whether the scoliosis of each segment is a major curve and whether it is a structural curve, and the scoliosis type is obtained. The relative position relationship is determined based on the CSVL coordinates of the sacral perpendicular line and the pedicle mask coordinate information, and the lumbar correction type is obtained. The sagittal thoracic kyphosis angle is determined to obtain the thoracic sagittal correction type. Finally, the scoliosis type, lumbar correction type, and thoracic sagittal correction type are combined to obtain a complete Lenke classification.

[0039] Example 2, see Figure 1 This embodiment provides a system for measuring spinal parameters and automatically classifying the spine, which implements the method for measuring spinal parameters and automatically classifying the spine as in Example 1, and mainly includes: an image neural network recognition module, a parameter measurement module and a Lenke classification parameter analysis module.

[0040] First, the image neural network recognition module can receive original image data, such as X-ray images. Specifically, the images are 4 original images of the spine in different directions, namely, the anteroposterior image, the lateral image, and the left / right dynamic image. After the image is sent to the image neural network recognition module, it is first pre-processed, that is, the original spinal image is adjusted in window width and window position, contrast enhancement, and image size standardization. Then the pre-processed image is sent to the neural network for training to dissect the key point coordinate data in the image, including vertebral order, vertebral angle point, and pedicle mask coordinate information.

[0041] The key point coordinate data is then sent to the parameter measurement module, which will perform parameter measurement and analysis based on the key information to obtain spine-related parameters, including the angle between adjacent vertebrae, the overall shape of the spine in the positive position, the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle, the apical vertebra or apical intervertebral disc between the end vertebrae, the sagittal kyphosis angle and the CSVL coordinates of the sacral perpendicular line.

[0042] Finally, the Lenke classification parameter analysis module receives the various parameters output by the parameter measurement module, analyzes them, and then classifies them. Specifically: According to the size of the scoliosis Cobb angle, the size of the segmental and sagittal kyphosis angles, it is calculated whether the scoliosis of each segment is the main curve and whether it is a structural curve, and then the scoliosis curvature type is obtained. According to the CSVL coordinates of the sacral perpendicular line and the pedicle mask coordinate information, the relative position relationship is determined to obtain the lumbar curvature correction classification. According to the size of the sagittal thoracic kyphosis angle, the thoracic sagittal plane correction type is obtained. Furthermore, the above-mentioned scoliosis curvature type, lumbar curvature correction type and thoracic sagittal plane correction type are combined to obtain a complete Lenke classification.

[0043] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Even if various changes are made to the present invention, if these changes fall within the scope of the claims of the present invention and their equivalents, they still fall within the protection scope of the present invention.

Claims

1. A method for measuring spinal parameters and automatically classifying the spine, characterized in that: The steps include: Obtain 4 original spinal images of the target to be detected in different directions, namely, the anteroposterior direction image, the lateral direction image, and the left / right dynamic direction image. Preprocess the original spinal images and then send them into the pre-trained image neural network to obtain key information in the images, including vertebral order, vertebral angle points, and pedicle mask coordinate information; Parameters are measured and analyzed based on the key information, including the angle between adjacent vertebrae, the overall shape of the spine in the positive position, the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle, the apical vertebra or apical intervertebral disc between the end vertebrae, the sagittal kyphosis angle and the CSVL coordinate of the sacral median vertical line; According to the size of the Cobb angle of scoliosis, the size of the segmental and sagittal kyphosis angles, it can be calculated whether the scoliosis of each segment is the main curve and whether it is a structural curve, and the type of scoliosis can be obtained; The relative position relationship is determined based on the CSVL coordinates of the sacral median vertical line and the pedicle mask coordinates, and the modified lumbar curvature classification is obtained. According to the size of the thoracic kyphosis angle in the sagittal plane, the sagittal plane correction type of the thoracic spine is obtained; The scoliosis curve type, the modified lumbar curve classification and the modified thoracic sagittal plane classification are combined to obtain a complete Lenke classification.

2. The method for measuring spine parameters and automatically classifying the spine according to claim 1, characterized in that: The pre-training of the image neural network is specifically as follows: collecting a number of spinal images as a training image data set, pre-processing the spinal images in the training image data set, marking the pre-processed spinal images at the vertebral position, vertebral sequence, vertebral corners and pedicle masks, and constructing a spinal corner recognition model based on the marks in the image; inputting the standardized spinal model into the spinal corner recognition model for training, and constructing a vertebral recognition and segmentation model as a pre-trained image neural network.

3. The method for measuring spine parameters and automatically classifying the spine according to claim 2, characterized in that: The image preprocessing is specifically as follows: adjusting the window width and window position, contrast enhancement, and image size standardization of the image data in sequence.

4. The method for measuring spine parameters and automatically classifying the spine according to claim 1, characterized in that: The angle calculation between adjacent vertebrae is specifically as follows: for all vertebrae in the anteroposterior thoracic spine and the anteroposterior lumbar spine, the vertebrae angles between two adjacent vertebrae are calculated, the center points on the left and right sides of the vertebrae are calculated to form a center point vector, and then the angles between the center point vectors of all vertebrae in the anteroposterior thoracic spine and the anteroposterior lumbar spine are calculated respectively; Then, the overall shape of the anteroposterior spinal part is judged, specifically: the slopes of the center point vectors of the thoracic vertebra T1 and the lumbar vertebra L5 are calculated respectively; when the slopes of the center point vectors of the thoracic vertebra T1 and the lumbar vertebra L5 are one positive and one negative, the overall shape of the anteroposterior spinal part is not S-shaped, otherwise the overall shape is S-shaped.

5. The method for measuring spine parameters and automatically classifying the spine according to claim 4, characterized in that: According to the overall shape of the spine in the positive position, the Cobb angle of scoliosis and the end vertebra corresponding to the Cobb angle are calculated as follows: The four terminal vertebrae from top to bottom of the spine are named the first terminal vertebra V1, the second terminal vertebra V2, the third terminal vertebra V3 and the fourth terminal vertebra V4; When the overall shape of the spine is not S-shaped: the thoracic vertebra T1 is taken as the first-order end vertebra V1, the upper end vertebra of the first Cobb angle is taken as the second-order end vertebra V2, and the angle between the first-order end vertebra V1 and the second-order end vertebra V2 is calculated as the second Cobb angle; the lumbar vertebra L5 is taken as the fourth-order end vertebra V4, the lower end vertebra of the first Cobb angle is taken as the third-order end vertebra V3, and the angle between the fourth-order end vertebra V4 and the third-order end vertebra V3 is calculated as the third Cobb angle; When the overall shape of the spine is S-shaped: determine the positions of the upper end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle in the image data. If the upper end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle are located in the upper half of the image data, calculate the vertebral angles between the thoracic vertebra T1 and the upper end vertebra of the first Cobb angle and the upper end vertebra of the first Cobb angle, and use the maximum value of the calculated vertebral angles as the second Cobb angle; and make the upper end vertebra of the second Cobb angle the first-order end vertebra V1, and the upper end vertebra of the first Cobb angle the second-order end vertebra V2; then calculate the vertebral angles between the lumbar vertebra L5 and the lower end vertebra of the first Cobb angle and the lower end vertebra of the first Cobb angle, and use the maximum value of the calculated vertebral angles as the third Cobb angle; and make the lower end vertebra of the first Cobb angle the third-order end vertebra V3, and the lower end vertebra corresponding to the third Cobb angle the fourth-order end vertebra V4; If the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the lower half of the image data, the vertebral angles are calculated for all vertebrae between the thoracic vertebra T1 and the upper vertebra of the first Cobb angle, respectively, and the upper vertebra of the first Cobb angle, and the maximum value of the calculated vertebral angles is used as the second Cobb angle; and the upper vertebra corresponding to the second Cobb angle is made as the second-order end vertebra V2, and the upper vertebra of the first Cobb angle is made as the third-order end vertebra V3; and then the vertebral angles are calculated for all vertebrae between the thoracic vertebra T1 and the second-order end vertebra V2, respectively, and the second-order end vertebra V2, and the maximum value of the calculated vertebral angles is used as the third Cobb angle; and the upper vertebra of the third Cobb angle is made as the first-order end vertebra V1, and the lower vertebra of the first Cobb angle is made as the fourth-order end vertebra V4.

6. The method for measuring spinal parameters and automatically classifying the spine according to claim 5, characterized in that: The determination of the positions of the upper vertebrae of the first Cobb angle and the lower vertebrae of the first Cobb angle in the image data is specifically as follows: the center coordinates of the two corner points below the upper vertebrae of the first Cobb angle and the lower vertebrae of the first Cobb angle are recorded as P1 and P2, respectively, and the center coordinates of the two corner points below the thoracic vertebra T1 and the lumbar vertebra L5 are recorded as P3 and P4, respectively; When (P1+P2)<1.2*(P3+P4) is satisfied, the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the upper half of the image data; When (P1+P2)>=1.2*(P3+P4) is satisfied, the upper vertebra of the first Cobb angle and the lower vertebra of the first Cobb angle are located in the lower half of the image data.

7. The method for measuring spine parameters and automatically classifying the spine according to claim 6, characterized in that: The top vertebra or top intervertebral disc between the end vertebrae is calculated in sequence, specifically: Connect the midpoints of the first-order end vertebra V1, the second-order end vertebra V2, the third-order end vertebra V3 and the fourth-order end vertebra V4 from top to bottom to form a straight line connecting two adjacent end vertebrae; then calculate the straight-line distances between the vertebral bodies and intervertebral discs between the first-order end vertebra V1 and the second-order end vertebra V2, the second-order end vertebra V2 and the third-order end vertebra V3, and the third-order end vertebra V3 and the fourth-order end vertebra V4; between the two end vertebrae, the vertebral body or intervertebral disc with the longest straight-line distance is the top vertebra or top intervertebral disc corresponding to the current two end vertebrae.

8. The method for measuring spine parameters and automatically classifying the spine according to claim 7, characterized in that: The calculation of the sagittal kyphosis angle is as follows: According to the center coordinates of the two upper corner points of thoracic vertebra T2 and the center coordinates of the two lower corner points of thoracic vertebra T5 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T2 and the line connecting the coordinates of the lower corner point of T5 is calculated to obtain the T2-T5 kyphosis angle; According to the center coordinates of the two upper corner points of thoracic vertebra T5 and the center coordinates of the two lower corner points of thoracic vertebra T12 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T5 and the line connecting the coordinates of the lower corner point of T12 is calculated to obtain the T5-T12 kyphosis angle; According to the center coordinates of the two upper corner points of thoracic vertebra T10 and the center coordinates of the two lower corner points of thoracic vertebra L2 in the lateral image, the angle formed by the line connecting the coordinates of the upper corner point of T10 and the line connecting the coordinates of the lower corner point of L2 is calculated to obtain the size of the T10-L2 kyphosis angle.

9. The method for measuring spine parameters and automatically classifying the spine according to claim 6, characterized in that: The specific coordinates of the CSVL of the sacral perpendicular line are: The center point coordinates of the left and right corner points of the sacrum S1 in the anteroposterior image data are calculated to obtain the midpoint coordinates of the sacrum S1, and the horizontal coordinates of the midpoint coordinates are fixed to obtain the coordinates corresponding to the sacral vertical line CSVL.

10. A system for measuring spine parameters and automatically classifying the spine, which implements the method for measuring spine parameters and automatically classifying the spine as claimed in any one of claims 1 to 9, characterized in that: include: The image neural network recognition module is configured to obtain four original spinal images of the target to be detected in different directions, namely, the anteroposterior direction image, the lateral direction image, and the left / right dynamic direction image, and then pre-process the original spinal images and perform neural network training to obtain key information in the images, including vertebral order, vertebral angle points, and pedicle mask coordinate information; A parameter measurement module is configured to perform parameter measurement and analysis according to the key information, including the angle between adjacent vertebrae, the overall shape of the spinal column in the positive position, the size of the scoliosis Cobb angle and the end vertebra corresponding to the Cobb angle, the apical vertebra or apical intervertebral disc between the end vertebrae, the size of the sagittal kyphosis angle and the CSVL coordinate of the sacral median vertical line; The Lenke classification parameter analysis module is configured to calculate based on the Cobb angle of scoliosis, the segmental and sagittal kyphosis angles, and can determine whether the scoliosis of each segment is a major curve and whether it is a structural curve, and determine the type of scoliosis; The relative position relationship is determined based on the CSVL coordinates of the sacral median vertical line and the pedicle mask coordinates, and the modified lumbar curvature classification is obtained. According to the size of the thoracic kyphosis angle in the sagittal plane, the sagittal plane correction type of the thoracic spine is obtained; The scoliosis curve type, the modified lumbar curve classification and the modified thoracic sagittal plane classification are combined to obtain a complete Lenke classification.

Images