Radiographic scoliosis angle estimation: spline-based measurement reveals superior reliability compared to traditional COBB method

Abstract

Introduction and objective Although being standard for scoliosis curve size estimation, COBB angle measurement is well known to be inaccurate, due to a high interobserver variance in end vertebra selection and end plate contour delineation. We propose a stepwise improvement by using a spline constructed from vertebra centroids to resemble spinal curve characteristics more closely. To enhance precision even further, a neural net was trained to detect the centroids automatically. Materials & Methods Vertebra centroids in AP spinal X-ray images of varying quality from 551 scoliosis patients were manually labeled by 4 investigators. With these inputs, splines were generated and the computed curve sizes were compared to the manually measured COBB angles and to the curve estimation obtained from the neural net. Results Splines achieved a higher interobserver correlation of 0.92-0.95 compared to manual COBB measurements (0.83-0.92) and showed 1.5-2 times less variance, depending on the anatomic region. This translates into an average of 1° of interobserver measurement deviation for spline-based curve estimation compared to 3°-8° for COBB measurements. The neural net was even more precise and achieved mean deviations below 0.5°. Conclusion In conclusion, our data suggest an advantage of spline-based automated measuring systems, so further investigations are warranted to abandon manual COBB measurements.

W‐Transformer: Accurate Cobb angles estimation by using a transformer‐based hybrid structure

“…It seemingly like a better way to obtain a curvature, as it focuses on vertebra centroids which can offer the advantage of generating spinal curves that resemble global scoliotic spinal deformity more closely. 38 But, we find that this kind of methods are quite dependent on the accuracy of selected centroids. In our perspective, the data set includes mainly patients of young age, in cases of severe degenerative scoliosis with worsened coronal tilts of vertebra, centerline will remain unchanged while endplates not.…”

“…Second, there are alternative ways to calculate spine curvature. It seemingly like a better way to obtain a curvature, as it focuses on vertebra centroids which can offer the advantage of generating spinal curves that resemble global scoliotic spinal deformity more closely 38 . But, we find that this kind of methods are quite dependent on the accuracy of selected centroids.…”

“…The percentage of Cobb angle detection (SMAPE) in most of the studies is more than twice as ours, which can support the superior effectiveness of our model. Some researches 16,17,38 focus on global spine curvature in order to enhance reliability of Cobb angle estimation through a spine centerline.However,we enhance the accuracy of Cobb angle estimation by improving accuracy of landmark detection, and there are several reasons accounting for it. First of all, we use the algorithm provided by AASCE challenge to estimate Cobb angles, since most automatic measurement methods use this method to calculate Cobb angle when using the same database.…”

Automated Cobb Angle Measurement for Adolescent Idiopathic Scoliosis Using Convolutional Neural Network

“…With the advancement of computer vision technology [ 10 ], such as machine learning [ 11 , 12 ] and deep learning methods [ 13 , 14 , 15 ], a fully automatic process to measure the CA quickly and precisely can potentially overcome the shortcomings mentioned above. Digitalized X-ray images, processed through computerized learning, have previously been used to screen, diagnose, and classify severity of scoliosis [ 12 ].…”

“…Automatic detection of the spine anatomy in X-ray images to select the appropriate end vertebrae is an essential stage for CA measurement. Studies on the vertebrae and spine detection based on machine learning [ 11 , 12 ] and deep learning methods [ 13 , 14 , 15 , 16 ] have been previously reported. However, automatic detection of the vertebrae in X-ray images can be difficult due to overlapping rib and pelvic shadows, as well as contrast differences between the different vertebrae levels.…”