Teenager idiopathic scoliosis screening method based on three-dimensional curvature asymmetry

The back point cloud data was collected through the depth camera and the three-dimensional curvature asymmetry was calculated, which solved the problems of long time, high cost, large radiation and low accuracy of idiopathic scoliosis screening in adolescents, achieving rapid and accurate screening Methods: The screening results were more correlated with the Cobb angle, with an accuracy rate of 95%.

CN119055193APending Publication Date: 2024-12-03PEKING UNION MEDICAL COLLEGE HOSPITAL
View PDF 0 Cites 5 Cited by

Patent Information

Application Number
CN202411177074.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-26
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

In the prior art, the screening method for idiopathic scoliosis in adolescents has problems such as time-consuming, high cost, high radiation and low accuracy, and depends on the proficiency of the doctor and overestimating sensitivity and specificity.

Method used

The back point cloud data was collected by using a depth camera, and the asymmetric index was obtained by calculating the three-dimensional curvature asymmetry to achieve rapid and radiation-free scoliosis screening. The PDflow method was used to perform denoising treatment, and the back was calculated by combining the average curvature and Gaussian curvature. Asymmetry index, classification of severity of curvature.

Benefits of technology

A low-cost, radiation-free and high accuracy screening of idiopathic scoliosis in adolescents was achieved, with fast screening speed, and the relationship between the results and the Cobb angle was more convincing, with an accuracy rate of 95%, and a sensitivity of 91.8%. , the specificity is 97.6%.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119055193A_ABST
    Figure CN119055193A_ABST
Patent Text Reader

Abstract

The invention discloses a method for screening teenager idiopathic scoliosis based on three-dimensional curvature asymmetry, which is characterized in that calculation is carried out by using back point cloud data shot by a depth camera, and an asymmetry index is obtained by quantifying the three-dimensional curvature asymmetry of the back and is used for screening patients with scoliosis. According to the method, errors caused by trunk rotation are avoided, the screening speed is high, radiation is avoided, the screening method is simple and easy to operate, screening groups also include common population in schools and scoliosis patients in hospitals, and the relation between data results and Cobb angles is more persuasive. The method is low in cost, free of radiation and high in accuracy, and teenager idiopathic scoliosis screening can be completed in a short time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of medical detection, and in particular to a method for screening adolescent idiopathic scoliosis. Background Art

[0002] Idiopathic scoliosis is a lateral curvature of the spine with a Cobb angle of at least 10 degrees in the absence of an underlying congenital or neuromuscular abnormality. It is more common in adolescents, especially females, between the ages of 10 and the end of their growth spurt and skeletal maturity. Although most adolescents with scoliosis do not experience clinical symptoms, it may progress to rib deformity and respiratory restriction, causing significant cosmetic and emotional distress in some patients. When the scoliotic curve reaches 20 degrees, bracing should be used, while when the angle reaches 45 degrees or more, surgery is required.

[0003] Currently, the "gold standard" for diagnosing scoliosis is a standing full spine X-ray examination, and a Cobb angle of the coronal spinal curvature greater than 10 degrees can be confirmed. However, the harm of X-ray radiation to the body makes it impossible to use it as a large-scale screening tool. Therefore, a non-invasive, rapid and radiation-free screening tool is needed to evaluate adolescent idiopathic scoliosis. Therefore, it is practical and important to study the estimation and determination of scoliosis based on the information of the back surface of the human body.

[0004] Chiara et al. concluded that the identification of 3D curves passing through the external positions of vertebral condyles is one of the most relevant applications of the theory of symmetry lines applied to the back model. Moreover, once the symmetry line is known and the thickness of the soft part is taken into account, the midline of the spine can be easily estimated. In all subjects, the symmetry line divides the back into two "quasi-mirror" parts, each describing an involuntary orthogonal transformation of the other. Kyphosis, scoliosis, or lordosis can occur during adolescence and cause the spine to deviate from the sagittal plane or even torsion. These methods can be divided into three categories: methods based on skin markers, methods based on parallel sections, and methods based on adaptive sections.

[0005] In the paper "A new system for measuring three-dimensional back shape in scoliosis" published in the journal EUROPEAN SPINE JOURNAL, Berryman et al. proposed a low-cost, radiation-free scoliosis assessment and monitoring system that uses a structured light device to capture the 3D back and evaluate the spinal status by calculating the back asymmetry based on the spinous process curve, lumbar dimples, and the 3D back. The parallel fringe pattern projected onto a three-dimensional object is distorted by the shape of the object. This distortion contains information about the distance (height) of the object's surface away from the plane. When a digital camera acquires an image of such an object, the pixel position of the object gives its x and y dimensions, and the phase of the fringe pattern at each pixel contains height information perpendicular to the image plane. Therefore, the problem of finding the height distribution is simplified to the problem of finding the phase distribution of the fringes on the object. The method used here to calculate the phase distribution is Fourier transform profilometry (FTP). However, the variability of the Cobb angle measurement itself is considerable, ranging from 5 to 10 degrees between raters. In patients with extreme obesity or heavy musculature, the relationship between ISIS2 lateral asymmetry and the radiographic Cobb angle will be limited because identification of bony landmarks for marking will be more difficult. It will also be limited in patients with congenital curves with little rotation.

[0006] Drerup's paper "Rasterstereographic measurement of scoliotic deformity" published in the journal Scoliosis reconstructed the 3D spinal midline with spinous process curves and lumbar vertebrae, which was positively correlated with the Cobb angle, and scoliosis can be evaluated using back asymmetry. First, raster stereo image acquisition and reconstruction were performed to obtain a reconstructed back perspective view, and then the raster stereo image shape analysis was performed using curvature information, showing the curvatures detected on the reconstructed back model. According to the principal curvatures, they are indicated as convex or concave areas. In terms of differential geometry, this type of representation is named mean curvature. Another algorithm for discriminating saddle curvature is represented by Gaussian curvature. Finally, the spinal model was constructed as well as the positioning device and surface markers. Although the agreement between the radiological and reconstructed curves was sufficient, the prediction of the radiological Cobb angle showed significant deviations from the radiological measurements.

[0007] Kokabu et al., in their paper “Three-dimensional depth sensor imaging to identify adolescent idiopathic scoliosis: a prospective multicenter cohort study” published in the journal Scientific Reports, proposed that asymmetry of the entire back is positively correlated with the Cobb angle and that asymmetry of the back can be used to assess scoliosis. Back asymmetry was calculated by bending forward 90 degrees, the Adam flexion test. In this study, the correlation coefficient between the Cobb angle and the asymmetry index was 0.85, which exceeded the reported value (0.677) for the Adam flexion test using a spinal screening instrument. When predicting an AIS greater than 10 degrees, the ROC was 0.98, the sensitivity was 0.97, and the specificity was 0.93, indicating that the system had outstanding discriminatory power in determining whether the subject required additional radiological examinations to confirm the diagnosis of scoliosis. In addition, if the asymmetry index was >1.268 three times in a row, the posterior probability of a positive screening greater than 10 degrees was 98.9%. When a subject was suspected of having AIS with an asymmetry index greater than 1.268 using this system, the subject would undergo two additional tests, which eliminated the necessity of radiography for the diagnosis of AIS. Although this is OK when determining the correlation between Cobb measurements and asymmetry index, it may overestimate the true sensitivity and specificity when used as a screening tool in primary care hospitals or schools where most students do not have scoliosis.

[0008] The current clinical screening technology for adolescent idiopathic scoliosis still has the following shortcomings:

[0009] 1) When clinicians perform scoliosis screening, they need to use spinal screening tools, which places high demands on clinicians and takes a long time, affecting the efficiency of large-scale screening and being too dependent on the doctor's proficiency.

[0010] 2) Due to the cost of X-rays and the adverse effects of radiation, large-scale screening cannot be carried out using X-rays.

[0011] 3) Current non-radiation methods for detecting scoliosis have their own limitations, such as overestimation of the relationship between the Cobb angle and overestimation of the true sensitivity and specificity. Summary of the invention

[0012] In order to solve the problems of current screening technology, such as long time consumption, high cost, large radiation and low accuracy, the present invention provides a low-cost, radiation-free, high-accuracy method for screening adolescent idiopathic scoliosis that can be completed in a short time.

[0013] The present invention is implemented by the following technical solutions:

[0014] A screening method for adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry uses back point cloud data captured by a depth camera for calculation. By quantifying the three-dimensional curvature asymmetry of the back, an asymmetry index is obtained to screen scoliosis patients.

[0015] The following steps are involved:

[0016] S1: The depth camera is used to collect back point cloud data in two postures: standing posture and adam forward bending posture.

[0017] S2: By analyzing the depth image, the key joints of the human body are identified, so as to accurately extract the key parts and crop the rectangular area of ​​the back.

[0018] S3: The PDflow method is used to apply denoising technology to the point cloud. The mean curvature and Gaussian curvature values ​​of the points in the point cloud are calculated.

[0019] S4: Calculate the back asymmetry index.

[0020] S5: Based on the asymmetry index calculated above, the severity of scoliosis is classified.

[0021] The present invention uses the back point cloud data captured by the depth camera for calculation, and obtains the asymmetry index by quantifying the three-dimensional curvature asymmetry of the back to screen scoliosis patients. The present invention solves the error caused by trunk rotation, and the screening speed is fast, there is no radiation, the screening method is simple and easy to operate, and the screening group also includes the general population in schools and scoliosis patients in hospitals. The relationship between the data results and the Cobb angle is more convincing.

[0022] The present invention further verified the feasibility of the system in a screening. Compared with the preliminary results obtained by doctors using spinal screening tools and physical examinations, the system results of the present invention showed an accuracy rate of more than 0.95. The specific results are shown in the following table:

[0023]

[0024]

[0025] The above data show that the screening accuracy of the present invention is as high as 95%, the sensitivity is 91.8%, and the specificity is 97.6%, which can provide a potential low-cost, radiation-free, and highly accurate method for screening adolescent idiopathic scoliosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is an overall flow chart of the screening method of the present invention. DETAILED DESCRIPTION

[0027] With reference to the accompanying drawings, the present invention comprises the following steps:

[0028] S1: Collect point cloud data of the subject's back in standing and adam forward bending positions

[0029] S2: Both Azure Kinect DK cameras used use infrared light projection and time-of-flight principles to obtain depth information about the human body and the surrounding environment. This depth information is then used to identify the human body in the scene, effectively distinguishing it from the surrounding environment. By examining the connected areas and object shapes in the depth image, the potential presence of the human body can be determined. The depth image is then analyzed to identify the key joints of the human body, thereby achieving accurate extraction of key parts such as the shoulders, neck, and spine. The key joints refer to the shoulders, neck, and spine. Accurate extraction of the key joints is the rectangular area of ​​the back. The specific steps are as follows

[0030] ① Data capture: The depth camera captures 3D depth data and RGB image data of the human body. These data provide detailed information about the human body in space.

[0031] ② Image processing: Use the sensor data from the Azure Kinect SDK for preprocessing, including filtering out noise and correcting distortion to ensure data accuracy.

[0032] ③ Skeleton tracking: The SDK uses a machine learning model to identify the human skeleton and extract the position of joints from the captured data. This model can identify multiple joints of the human body, such as the head, shoulders, elbows, wrists, hips, knees, and ankles.

[0033] ④ Generate skeleton model: Once the joint positions are identified, the SDK generates a 3D skeleton model in which each joint has a unique ID and 3D coordinates. These coordinates can be used to track the movements and postures of the human body.

[0034] ⑤Continuous tracking: In continuous frames, the SDK can continuously track the position of each joint and update the skeleton model to achieve real-time human motion capture.

[0035] S3: To optimize the acquired data, the PDflow method was used, which involves applying denoising techniques to the point cloud. PDflow performs the denoising process by learning the distribution of the noisy point set and subsequently by noise separation. This step helps to improve the quality of the captured data and optimize it for subsequent analysis and evaluation.

[0036] For individuals with scoliosis, changes in spinal curvature may result in rib prominence, uneven shoulders, and asymmetric back muscles. These deviations affect the curvature of the back plane, and this change can be used to quantify the degree of back asymmetry. In the field of differential geometry, the mean curvature H of a surface S is considered an "extrinsic" measure that characterizes the local curvature of the surface in the surrounding space. On the other hand, the Gaussian curvature K is an "intrinsic" measure of curvature that relies only on the distance measurement on the surface itself, without considering its embedding in space. Therefore, any non-stretching transformation of the surface will not change its Gaussian curvature. Therefore, using these two curvatures can avoid the influence of uncontrollable body rotation when evaluating back asymmetry, thereby achieving a more stable assessment of asymmetry.

[0037] S4: To calculate the Gaussian curvature and mean curvature of a specific point in the back point cloud, it is necessary to calculate the Euclidean distance to the nearest point to determine the index of the nearest point. A covariance matrix C is then generated by pairing the current point with its neighboring points with a radius not greater than r. The process involves subtracting the mean from the coordinates of these neighboring points to create a covariance matrix C that measures the discreteness of the data points along different axes. Finally, the principal curvature kn is obtained by calculating the eigenvalues ​​and eigenvectors derived from the covariance matrix. In the main direction of the tangent plane, the Gaussian curvature K and the mean curvature H are the product and average of the maximum principal curvature kmax and the minimum principal curvature kmin, respectively.

[0038] K=k max ·k min (1)

[0039]

[0040] The asymmetry index of the back can then be calculated. From the rectangular area extracted from the point cloud, a comprehensive analysis of the curvature difference from top to bottom along the left and right midline of the back, i.e., the spine line, is performed. A point p located on the midline Lm is chosen for consideration. In this case, a series of points are selected from the lateral projection point, i.e., the lateral back, from left to right. These points are then fitted using a curve of length L0, denoted as Lp. Starting from point p, the process sequentially evaluates the Gaussian curvature and the mean curvature difference of points equidistant from Lp. It is worth noting that the experimental configuration uses a local neighborhood radius r of 32. At the same time, the step size u of each calculation along the tangent is set to 4.

[0041] It should be noted that separate weights are assigned to H and K because the values ​​of H and K have different rates of change in the local neighborhood of the posterior point cloud at a given radius. This distinction leads to the asymmetry index calculation formula for Lp:

[0042]

[0043] in H li =H(x p -i·u), H ri =H(x p +i·u), K li =K(x p -i·u), K ri =K(x p +i·u);

[0044] H li ,K li They represent the mean curvature and Gaussian curvature of the i-step to the left of point pp, respectively. ri ,K ri denote the mean curvature and Gaussian curvature i steps to the right of point p, respectively. Meanwhile, α and β serve as weights assigned to the mean curvature difference and Gaussian curvature difference, respectively. The asymmetry index of the individual back is calculated as the mean of the A values ​​of m points along the midline Lm.

[0045]

[0046] Subsequently, the final asymmetry index A was derived by combining the measurements of the Astand (α=0.6, β=0.4) and Adams forward bend Abend (α=0.8, β=0.2) postures.

[0047] In summary, the proposed three-dimensional curvature-based back asymmetry index calculation method is as follows.

[0048]

[0049]

[0050] S5: Based on this asymmetry index, the severity of scoliosis can be classified. Among the more than 1,000 people who have been screened, according to the final parameters, when the asymmetry index is greater than 6, it corresponds to a Cobb angle greater than 10 degrees (which can be considered as scoliosis), and when the index is less than 6, it is considered healthy and without scoliosis. The following table shows the asymmetry index values ​​corresponding to several patients with different degrees of scoliosis.

[0051] Cobb angle Asymmetry Index 15.6 8.24 22.6 10.01 39.4 15.22

Claims

1. A screening method for adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry, characterized in that: The following steps are involved: S1: The depth camera is used to collect the back point cloud data of the subject in two postures: standing posture and adam forward bending posture; S2: By analyzing the depth image, the key joints of the human body are identified, so as to accurately extract the key parts and cut out the rectangular area of ​​the back; S3: The PDflow method is used to apply denoising technology to point clouds, and based on the formula K = k max ·k min and Calculate the mean curvature and Gaussian curvature values ​​of points in the point cloud; S4: Calculate the back asymmetry index; S5: Based on the asymmetry index calculated above, the severity of scoliosis is classified.

2. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1, characterized in that: In step S4, the formula The back asymmetry index was calculated, where K represents Gaussian curvature, H represents mean curvature, and A represents asymmetry index.

3. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1, characterized in that: The depth camera in step S1 is Azure Kinect DK camera, which uses infrared light projection and time-of-flight principle to obtain depth information about the human body and the surrounding environment and identify the human body in the scene.

4. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1, characterized in that: Step S2 checks the connection area and object shape in the depth image to determine the potential existence of the human body, analyzes the depth image, identifies the key joints of the human body, which are the shoulders, neck and spine, and accurately extracts the key joints, which are the rectangular areas of the back.

5. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 4, characterized in that: The specific steps for accurately extracting key joint points are as follows: 3.

1. Data Capture: The depth camera captures 3D depth data and RGB image data of the human body. These data provide detailed information about the human body in space; 3.

2. Image processing: Use the sensor data from Azure Kinect SDK for preprocessing, including filtering out noise and correcting distortion to ensure data accuracy; 3.

3. Skeleton tracking: The SDK uses a machine learning model to identify the human skeleton and extract the position of joints from the captured data. This model can identify multiple joints of the human body, including the head, shoulders, elbows, wrists, hips, knees, and ankles; 3.

4. Generate skeleton model: Once the joint positions are identified, the SDK will generate a 3D skeleton model, in which each joint has a unique ID and 3D coordinates, which can be used to track the body's movements and postures; 3.

5. Continuous tracking: In consecutive frames, the SDK can continuously track the position of each joint and update the skeleton model to achieve real-time human motion capture.

6. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1, characterized in that: Step S3: In order to optimize the acquired data, the PDflow method was used to improve the quality of the captured data and optimize it for subsequent analysis and evaluation.

7. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1, characterized in that: Step S3: Calculate the Gaussian curvature and mean curvature of a specific point in the back point cloud, that is, the lateral projection point of the midline Lm, that is, a series of points in the lateral back from left to right, and its local neighborhood, calculate the Euclidean distance to the nearest point, determine the index of the nearest point, and generate a covariance matrix C by pairing the current point with its neighboring points with a radius not greater than r. The process includes subtracting the mean from the coordinates of these neighboring points to create a covariance matrix C that measures the discreteness of the data points along different axes. Finally, the principal curvature kn is obtained by calculating the eigenvalues ​​and eigenvectors derived from the covariance matrix. In the main direction of the tangent plane, the Gaussian curvature K and the mean curvature H are the product and average of the maximum principal curvature kmax and the minimum principal curvature kmin, respectively; K=k max ·k min (1) 8. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1 or 2, characterized in that: Step S4: Calculate the asymmetry index of the back. From the rectangular area extracted from the point cloud, a comprehensive analysis is performed on the curvature difference from the top to the bottom along the left and right midlines of the back, i.e., the spine line. A point p located on the midline Lm is selected for consideration. In this case, a series of points are selected from the lateral projection point, i.e., the lateral back, from left to right. Then, these points are fitted using a curve of length L0, denoted as Lp. Starting from point p, the process sequentially evaluates the Gaussian curvature and mean curvature differences of points equidistant from Lp. The local neighborhood radius r is 32. At the same time, the step length u of each calculation along the tangent is set to 4. Since the values ​​of H and K have different rates of change in the local neighborhood of the posterior point cloud at a given radius, separate weights are assigned to H and K. This distinction leads to the asymmetric index calculation formula for Lp: in H li =H(x p -i·u), H ri =H(x p +i·u), K li =K(x p -i·u), K ri =K(x p +i·u); H li ,K li They represent the mean curvature and Gaussian curvature of the i-step to the left of point pp, respectively. ri ,K ri denote the mean curvature and Gaussian curvature i steps to the right of point p, respectively. Meanwhile, α and β are the weights assigned to the mean curvature difference and Gaussian curvature difference, respectively. The asymmetry index of the individual back is calculated as the mean of the A values ​​of m points along the midline Lm; Subsequently, the final asymmetry index A was derived by combining the measurements of the Astand (α=0.6, β=0.4) and Adams forward bend Abend (α=0.8, β=0.2) postures.

9. The method for screening adolescent idiopathic scoliosis based on three-dimensional curvature asymmetry according to claim 1, characterized in that: S5: Based on the asymmetry index, when the asymmetry index A is greater than 6, it corresponds to a Cobb value greater than 10 degrees, which is considered to be scoliosis. When the index is less than 6, it is considered to be healthy and without scoliosis.

Citation Information

Cited By

  • Spinal line generation method, spinal line generation device and electronic equipment

    CN120130998A

  • Spine line generation method, spine line generation device, and electronic device

    CN120130998B

  • Multi-dimensional physiological data automatic teenager idiopathic scoliosis intelligent screening system

    CN121242513A

  • Method for monitoring abnormal spine posture of teenager based on behavior feature recognition

    CN121392758A

  • A method for monitoring abnormal posture of teenagers' spine based on behavior feature recognition

    CN121392758B