To predict the sagittal spinal parameters as measured in a 3D model of the spine using the 2D radiographic measurements. Bi-planar low-dose stereoradiography images of 73 right thoracic AIS patients were processed to generate 3D models of the spine and pelvis. T1-T12 kyphosis, L1-S1 lordosis, and pelvic rotation were calculated using these 3D models. With the same X-rays, T1-T12 kyphosis, L1-S1 lordosis, thoracic and lumbar frontal curves, and pelvic rotation (calculated from the frontal and sagittal distances between the femoral heads) were manually measured on the X-rays by two independent observers. 3D sagittal parameters were predicted from only 2D sagittal parameters (simple regression) and from 2D sagittal parameters, 2D frontal parameters, and pelvic rotation (multiple regression). The simple and multiple regression models were compared for efficiency and accuracy of prediction. Comparing single and multiple regression models, multiple regression improved the prediction of the 3D sagittal parameters for kyphosis (R2 = 0.78-0.86) and lordosis (R2 = 0.88-0.92) measurements when compared to simple regression. The impact of pelvic rotation was significant when 2D kyphosis was higher than 40° and thoracic curve was less than 60° or 2D kyphosis was less than 40° and thoracic curve was higher than 60°, p < 0.05. Lordosis of 60° and higher were more prone to measurement error when pelvic rotation was present, p < 0.05. Both pelvic rotation and frontal deformity affect the accuracy of the 2D sagittal measurements of the scoliotic spine. We suggest the importance of the 3D considerations in sagittal evaluation of AIS.
Read full abstract