CN-115797320-B - Scoliosis end vertebra identification method and system

CN115797320BCN 115797320 BCN115797320 BCN 115797320BCN-115797320-B

Abstract

The invention relates to the technical field of vertebral body recognition, and provides a scoliosis end vertebral recognition method which comprises the following steps of S1, inputting original image data into a vertebral body recognition division model to obtain labeled image data after vertebral body recognition, S2, calculating the angles of vertebral bodies between the thoracic vertebrae and all vertebral bodies of the lumbar vertebrae in the labeled image data, taking the maximum value of the angles of the vertebral bodies as a first Cobb angle, taking two vertebral bodies corresponding to the first Cobb angle as an upper end vertebral body and a lower end vertebral body of the first Cobb angle, S3, judging the overall shape of a vertebral column part, calculating a second Cobb angle and a third Cobb angle according to the overall shape, and obtaining end vertebral bodies corresponding to the second Cobb angle and the third Cobb angle. Compared with the traditional manual labeling method, the method has the advantages of high measurement speed and high identification precision, is more suitable for large-batch data processing, and simplifies the operation flow of a user.

Inventors

WANG CONG
XIANG QINGZHI
LI XIAOMING
YU HAIXIN

Assignees

思浦奈医疗科技(上海)有限公司

Dates

Publication Date: 20260512
Application Date: 20221221

Claims (8)

1. The scoliosis end vertebra identification method is characterized by comprising the following steps of: s1, inputting original image data of a vertebra part into a vertebra identification dividing model to obtain labeled image data after vertebra identification; S2, calculating cone angles between every two of all cones of the thoracic vertebrae and the lumbar vertebrae in the marked image data, wherein the maximum value of the cone angles is used as a first Cobb angle, and two cones corresponding to the first Cobb angle are used as a first Cobb angle upper end cone and a first Cobb angle lower end cone; in the step S2, calculating the cone angles between every two of the thoracic vertebrae and all the cones of the lumbar vertebrae, namely calculating center points of left and right sides of all the cones of the thoracic vertebrae and the lumbar vertebrae to form a center point vector; s3, judging the overall shape of the spine part, calculating a second Cobb angle and a third Cobb angle according to the overall shape, and acquiring end vertebrae corresponding to the second Cobb angle and the third Cobb angle; In step S3, the overall shape of the spine part is judged, specifically, the slope of the central point vector of the thoracic vertebra T1 and the lumbar vertebra L5 is calculated respectively, when the slope of the central point vector of the thoracic vertebra T1 and the lumbar vertebra L5 is positive and negative, the overall shape is not S-shaped, otherwise, the overall shape is S-shaped.
2. The scoliosis end vertebral identification method according to claim 1, wherein in step S1, the vertebral body identification division model is constructed by the steps of: Collecting a plurality of image data of the spinal location as a training image data set; preprocessing the image data in the training image data set, wherein the preprocessing comprises window width, window level, contrast enhancement and picture size standardization; marking the image data in the training image data set, including cone positions, cone sequences and cone corner points, and constructing a spine corner point recognition model; and inputting the standardized spine model into the spine angular point recognition model for training, and constructing the vertebral body recognition division model.
3. The scoliosis end vertebra identification method according to claim 1, characterized in that in step S3, the second Cobb angle and the third Cobb angle are calculated according to the overall shape, and end vertebrae corresponding to the second Cobb angle and the third Cobb angle are obtained, specifically: four end vertebrae of the vertebra part from top to bottom are arranged on the first order end vertebrae V1, the second order end vertebrae V2, the third order end vertebrae V3 and the fourth order end vertebrae V4; when the overall shape is not S-shaped: Taking the thoracic vertebra T1 as a first sequence end vertebra V1, taking the upper end vertebra of the first Cobb angle as a second sequence end vertebra V2, calculating an angle between the first sequence end vertebra V1 and the second sequence end vertebra V2 as a second Cobb angle, taking the lumbar vertebra L5 as a fourth sequence end vertebra V4, taking the lower end vertebra of the first Cobb angle as a third sequence end vertebra V3, and calculating an angle between the fourth sequence end vertebra V4 and the third sequence end vertebra V3 as a third Cobb angle; when the overall shape is S-shaped: Judging positions of the upper end vertebrae of the first Cobb angle and the lower end vertebrae of the first Cobb angle in the marked image data; If the upper end cone of the first Cobb angle and the lower end cone of the first Cobb angle are positioned at the upper half part of the marked image data, calculating the cone angles between the upper end cone of the first Cobb angle and all the cones between the thoracic vertebrae T1 and the upper end cone of the first Cobb angle respectively, taking the maximum value of the cone angles as the second Cobb angle, taking the upper end cone corresponding to the second Cobb angle as the first sequence end cone V1, taking the upper end cone of the first Cobb angle as the second sequence end cone V2, and taking the maximum value of the cone angles as the third Cobb angle, and taking the lower end of the first Cobb angle as the third sequence end cone V3 and the lower end cone of the first Cobb angle as the fourth sequence end cone V4; If the upper end cone of the first Cobb angle and the lower end cone of the first Cobb angle are positioned at the lower half part of the marked image data, calculating the cone angles between the upper end cones of the first Cobb angle and the first Cobb angle respectively for all the cones between the thoracic vertebrae T1 and the upper end cone of the first Cobb angle, taking the maximum value of the cone angles as the second Cobb angle, taking the upper end cone corresponding to the second Cobb angle as the second sequential end cone V2, taking the upper end cone corresponding to the first Cobb angle as the third sequential end cone V3, and taking the upper end corresponding to the third Cobb angle as the first sequential end cone V1 and the lower end cone V4 simultaneously for all the cones between the thoracic vertebrae T1 and the second sequential end cone V2.
4. The method for identifying a scoliosis end vertebra according to claim 3, wherein determining the positions of the first Cobb angle upper end vertebra and the first Cobb angle lower end vertebra in the labeled image data specifically includes: Recording the central coordinates of two corner points of the upper end cone of the first Cobb angle and the lower side of the lower end cone of the first Cobb angle as P1 and P2 respectively; the central coordinates of two corner points at the lower sides of the thoracic vertebra T1 and the lumbar vertebra L5 are respectively denoted as P3 and P4; When (p1+p2) <1.2 (p3+p4), the first Cobb angle upper end cone and the first Cobb angle lower end cone are on the upper half of the marked image data; When (p1+p2) > =1.2 (p3+p4), the first Cobb angle upper end vertebra and the first Cobb angle lower end vertebra are in the lower half of the labeled image data.
5. The scoliosis end vertebra identification method of claim 3, further comprising: s4, calculating the top vertebrae or the top intervertebral discs among the end vertebrae in sequence, wherein the method specifically comprises the following steps: The midpoints of the first sequential end vertebra V1, the second sequential end vertebra V2, the third sequential end vertebra V3 and the fourth sequential end vertebra V4 which are connected from top to bottom form a straight line connecting the two end vertebrae; Sequentially calculating the distances between the vertebral bodies and the 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 from the straight line; and between the two end vertebrae, the vertebral body or the intervertebral disc with the longest distance from the straight line is the top vertebrae or the top intervertebral disc corresponding to the two end vertebrae at present.
6. The scoliosis end vertebra identification method of claim 2, further comprising: the training image data set comprises image data of the right position and the side position of the vertebra part; and calculating the first Cobb angle, the second Cobb angle and the third Cobb angle respectively for the original image data of the right position and the side position, acquiring end vertebrae corresponding to the first Cobb angle, the second Cobb angle and the third Cobb angle, and carrying out three-dimensional recognition on scoliosis of the vertebra part.
7. A scoliosis end vertebral identification system for performing the scoliosis end vertebral identification method of any of claims 1-6, comprising: The image data labeling module is used for inputting the original image data of the vertebra part into the vertebra identification dividing model to obtain labeled image data after the vertebra identification; The first Cobb angle calculation module is used for calculating cone angles between every two of all cones of thoracic vertebrae and lumbar vertebrae in the marked image data, taking the maximum value of the cone angles as a first Cobb angle, and taking two cones corresponding to the first Cobb angle as a first Cobb angle upper end cone and a first Cobb angle lower end cone; The second Cobb angle calculation module is used for judging the overall shape of the spine part, calculating a second Cobb angle and a third Cobb angle according to the overall shape, and acquiring end vertebrae corresponding to the second Cobb angle and the third Cobb angle; and the apical vertebral calculation module is used for sequentially calculating the apical vertebral or the apical intervertebral disc between the end vertebrae.
8. A computer readable storage medium storing computer code which, when executed, performs the method of any one of claims 1 to 6.

Description

Scoliosis end vertebra identification method and system Technical Field The invention relates to the technical field of vertebral body identification, in particular to a scoliosis end vertebral identification method and system. Background With the rapid development of computer and medical technology, deep learning is one of research hotspots in the field of artificial intelligence at present in the application of image recognition and clinical parameter calculation, and an automatic recognition method applying the deep learning technology has been applied to aspects such as measurement and arrangement. Scoliosis is a frequently occurring occurrence of the spine and is a hot spot in spine stability research. On standing position X-ray films, vertebral bodies with the largest inclination angles of the elbow side and the tail side of the spine, namely Cobb angle end vertebrae (end vertebrae, EV), are respectively drawn into straight lines along the upper end endplate of the upper end vertebrae and the lower end endplate of the lower end vertebrae, and the included angle of the two lines or the intersection angle of vertical lines, namely the Cobb angle. The angle is a numerical value for evaluating and quantifying the scoliosis degree, and has the guiding value of clinical treatment. The Cobb angle is an indispensable important parameter in both clinical trials and clinical treatment evaluations. In the actual measurement process, orthopedics often need to review a large number of radiological images, the process of vertebral body identification and angle measurement is time-consuming, and the diagnosis quality is affected by subjectivity and energy. The intelligent cone identification and angle measurement system using the artificial intelligence technology is beneficial to saving measurement and calculation time for doctors, and in the film reading process, the doctors can automatically measure and calculate required data only by confirming the recommended results given by the artificial intelligence system. Thus, automatic identification of the vertebral body segment type and position number, and automatic angular measurement are critical to use in clinical practice and in clinical research. The measurement of the Cobb angle needs to determine the end vertebra position at first, the traditional method measurement needs a clinician to manually select the upper end vertebra and the lower end vertebra of the lateral curvature spine, then the manual measurement is carried out, in the process of large-batch data processing, the workload is huge, the measurement accuracy cannot be controlled, the data heterogeneity is high, and the requirement on an imaging physician is higher because the manual measurement method needs to obtain a standing position standard righting plate. At present, an intelligent end vertebra identification algorithm based on a spine three-dimensional image has been presented, the identification algorithm has high identification efficiency and high calculation accuracy, but in the use process, a subject is required to firstly perform CT scanning, the spine three-dimensional reconstruction is performed by using a CT image, then the end vertebra is identified after the spine model registration of an X-ray image is performed, the flow is complex, the registration operation workload is large, and the radiation dose received by the subject is high. Disclosure of Invention The invention aims to provide a scoliosis end vertebra identification method and system. And automatically identifying the standing position X-ray film based on the artificial intelligent identification system, and positioning the coordinates of the end vertebrae in the scoliosis image. Compared with the traditional manual labeling method, the end-cone identification method based on the artificial intelligent image identification system has the advantages of high measurement speed and high identification precision, is more suitable for large-batch data processing, and compared with the end-cone identification method based on the three-dimensional model, the end-cone identification method based on the three-dimensional model adopts the standing X-ray film as the identification object, does not need CT scanning for three-dimensional modeling of a subject, reduces radiation exposure dose, and simplifies the operation flow of a user. The above object of the present invention is achieved by the following technical solutions: A scoliosis end vertebra identification method, comprising the following steps: s1, inputting original image data of a vertebra part into a vertebra identification dividing model to obtain labeled image data after vertebra identification; S2, calculating cone angles between every two of all cones of the thoracic vertebrae and the lumbar vertebrae in the marked image data, wherein the maximum value of the cone angles is used as a first Cobb angle, and two cones corresponding to the first Cobb angle are used as a fi