EOS imaging and scoliosis: the clinical applicability and intra-rater repeatability of measures.

Objectives: EOS bi-planar imaging enables three-dimensional (3D) reconstructions of the spine and pelvis with segmental vertebral measurements in three planes from a neutral pelvis. This study evaluates the repeatability of these measurements and the accuracy in detecting true changes.

Methods: Twenty patients from four clinical backgrounds (surgical threshold, bracing threshold, micro-dose, and in-brace) were included. EOS bi-planar "full spine" images were modelled and then subsequently re-modelled at least 4 weeks later by the same researcher. All 3D measurements were recorded and compared.

Results: The average modelling interval was 6.7 weeks. Paired measures indicated high agreement, except for planes of maximal curvature (PMC): thoracic (Spearman's = 0.67; p < 0.05) and lumbar (Spearman's = 0.40; p > 0.05). Intraclass correlation coefficients (ICCs) showed excellent agreement, with thoracic and lumbar Cobb angles averaging 0.99. Sagittal measurements ranged from 0.93 (L1/S1 lordosis) to 0.96 (T1/T12 kyphosis). Pelvic parameters ranged from 0.88 (obliquity) to 0.99 (tilt). The transverse profile ranged from 0.82 (apical thoracic rotation) to 0.98 (average lumbar rotation). Repeatability (2.77 × technical error of measurement [TEM]) was ± 4.4° for Cobb angles, ± 7.7° for sagittal profile, ± 5.0° for pelvic parameters, ± 4.8° for transverse profile, and ± 100.4° for automated thoracic and lumbar PMC. With strong outliers excluded, thoracic PMC was ± 16.2° and lumbar PMC was ± 15.5°.

Conclusion: 3D EOS measurements demonstrate excellent intra-rater ICC repeatability despite notable true measurement error that should define future success criteria. Semi-automated modelling provides quick 3D spinal alignment measurements from a neutral pelvis, with this study being the first to report TEM for 3D EOS reconstructions. PMC disagreement indicates the need for further investigation.