Comparison of 3-dimensional spinal reconstruction accuracy: biplanar radiographs with EOS versus computed tomography

Spine (Phila Pa 1976). 2012 Jul 15;37(16):1391-7. doi: 10.1097/BRS.0b013e3182518a15.


Study design: Experimental study for systematic evaluation of 3-dimensional (3D) reconstructions from low-dose digital stereoradiography.

Objective: To assess the accuracy of EOS (EOS Imaging, Paris, France) 3-dimensional (3D) reconstructions compared with 3D computed tomography (CT) and the effect spine positioning within the EOS unit has on reconstruction accuracy.

Summary of background data: Scoliosis is a 3D deformity, but 3D morphological analyses are still rare. A new low-dose radiation digital stereoradiography system (EOS) was previously evaluated for intra/interobserver variability, but data are limited for 3D reconstruction accuracy.

Methods: Three synthetic scoliotic phantoms (T1-pelvis) were scanned in upright position at 0°, ±5°, and ±10° of axial rotation within EOS and in supine position using CT. Three-dimensional EOS reconstructions were superimposed on corresponding 3D computed tomographic reconstructions. Shape, position, and orientation accuracy were assessed for each vertebra and the entire spine. Additional routine planer clinical deformity measurements were compared: Cobb angle, kyphosis, lordosis, and pelvic incidence.

Results: Mean EOS vertebral body shape accuracy was 1.1 ± 0.2 mm (maximum 4.7 mm), with 95% confidence interval of 1.7 mm. Different anatomical vertebral regions were modeled well with root-mean-square (RMS) values from 1.2 to 1.6 mm. Position and orientation accuracy of each vertebra were high: RMS offset was 1.2 mm (maximum 3.7 mm) and RMS axial rotation was 1.9° (maximum 5.8°). There was no significant difference in each of the analyzed parameters (P > 0.05) associated with varying the rotational position of the phantoms in EOS machine. Planer measurements accuracy was less than 1° mean difference for pelvic incidence, Cobb angle (mean 1.6°/maximum 3.9°), and sagittal kyphosis (mean less than 1°, maximum 4.9°).

Conclusion: The EOS image acquisition and reconstruction software provides accurate 3D spinal representations of scoliotic spinal deformities. The results of this study provide spinal deformity surgeons evidence pertaining to this new upright 3D imaging technology that may aid in the clinical diagnosis and decision making for patients with scoliosis.

Publication types

  • Comparative Study
  • Evaluation Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Humans
  • Imaging, Three-Dimensional*
  • Kyphosis / diagnostic imaging*
  • Phantoms, Imaging
  • Predictive Value of Tests
  • Radiation Dosage
  • Radiographic Image Interpretation, Computer-Assisted*
  • Reproducibility of Results
  • Scoliosis / diagnostic imaging*
  • Severity of Illness Index
  • Software
  • Spine / diagnostic imaging*
  • Tomography, X-Ray Computed / instrumentation
  • Tomography, X-Ray Computed / methods*