IntroductionAs scoliosis is a three-dimensional (3D) local and global deformation of the spine [5,9,10,23], 3D computerized rendering of scoliotic vertebrae is required for surgical planning as well as for finite element modeling of this spinal pathology [3,16].Because of the high irradiating dose [17] and of the patient having to adopt a lying position, computed tomography (CT), although quite accurate [7, 14,20], is an inappropriate 3D reconstruction method when dealing with scoliotic patients. Thus alternative radiological 3D reconstruction techniques, using 3D surface rendering algorithms, are necessary in order to provide the surgeon with the same 3D information. Such alternative techniques are all based on the stereoradiography method, which is also called biplanar radiography.Stereoradiographic 3D reconstruction techniques are divided into two classes. The first of these use an algorithm that allows reconstruction only where there are stereo-corresponding points, i.e., points for which the proAbstract The medical imaging techniques that allow a three-dimensional (3D) surface rendering reconstruction, which is usually required by the clinician when dealing with scoliotic patients, are computed tomography (CT) and stereoradiography. However, CT cannot provide a 3D rendering of the whole spine because of the high irradiating dose, while the stereoradiographic 3D reconstruction techniques, which use an algorithm derived from the direct linear transformation (DLT), are usually limited in accuracy because of the small number of corresponding anatomical landmarks identifiable on both radiographs. The purpose of the present study is to validate a recent biplanar 3D surface rendering reconstruction technique on scoliotic vertebrae. This technique, called "non stereo-corresponding points" (NSCP), has already been tested on non-pathologic dry cervical vertebrae and frozen lumbar specimens, and the results have proved very encouraging. Since scoliosis is a 3D deformity of the vertebrae and of the global spine, such a technique could be a very useful clinical tool for the diagnostic, follow-up and surgical planning when dealing with scoliotic patients. The validation of the NSCP technique on scoliotic patients was performed on 58 scoliotic vertebrae in 14 patients, by comparison with the CT scan 3D rendering technique. The results of this study show mean errors of 1.5 mm. On the basis of this study, we can conclude that the NSCP 3D reconstruction technique is a definite improvement over existing techniques, and can serve as a useful diagnosis tool in scoliosis. However, the results of the technique still need to be optimized for use in geometrical modeling.