Correlation Between a Novel Surface Topography Asymmetry Analysis and Radiographic Data in Scoliosis

Amin Komeili; Lindsey Westover; Eric C Parent; Marwan El-Rich; Samer Adeeb

doi:10.1016/j.jspd.2015.02.002

Correlation Between a Novel Surface Topography Asymmetry Analysis and Radiographic Data in Scoliosis

Spine Deform. 2015 Jul;3(4):303-311. doi: 10.1016/j.jspd.2015.02.002. Epub 2015 Jun 11.

Authors

Amin Komeili¹, Lindsey Westover², Eric C Parent³, Marwan El-Rich¹, Samer Adeeb¹

Affiliations

¹ Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta T6G2V4, Canada.
² Department of Mechanical Engineering, University of Alberta, Edmonton, Alberta T6G2G8, Canada. Electronic address: lwestove@ualberta.ca.
³ Department of Physical Therapy, Faculty of Rehabilitation Medicine, University of Alberta, Edmonton, Alberta T6G2G4, Canada.

PMID: 27927474
DOI: 10.1016/j.jspd.2015.02.002

Abstract

Study design: Cross-sectional study.

Objective: To investigate the correlation between parameters extracted from a three-dimensional (3D) asymmetry analysis of the torso and the internal deformities of the spine presented on radiographs, including 1) curve number, direction and location; 2) location of the apical vertebra; and 3) curve severity.

Summary of background data: Surface topography (ST) is used to assess external torso deformities and may predict important characteristics of the underlying spinal curves. ST does not expose patients to radiation and could safely be used clinically for scoliosis patients. Most ST indices rely on anatomical landmarks on the torso and 2D measurements.

Methods: The ability of a 3D markerless asymmetry technique to predict radiographic characteristics was assessed for 100 scoliosis patients with full torso ST scans. Twenty-four additional patients were used for validation. The number, direction, and location of curves were determined by three examiners using ST deviation color maps. The inter-method percentage of agreement and Kappa coefficient were estimated for each measure. Linear regression predicted the vertical location of the apical vertebra from ST. Curve severity (mild, moderate, severe) was predicted with a decision tree analysis using ST parameters.

Results: The average percentage of agreement was 62%, 66%, and 23% for single, double, and triple curves, respectively. Curve direction was always correctly identified. The average percentages of agreement for curve location were 63%, 92%, and 62% for proximal thoracic, thoracic/thoracolumbar (T-TL), and lumbar (L) curves, respectively. Apical vertebra location was predicted with R² = 0.89 for T-TL and R² = 0.58 for L curves. ST parameters classified curve severity for T-TL and L curves with 73% and 59% accuracy, respectively.

Conclusions: The method presented here improves upon current ST techniques by using the entire torso surface and both a visual and quantitative representation of the asymmetry to better capture the torso deformity.

Keywords: Adolescent idiopathic; Asymmetry analysis; Radiography; Scoliosis; Surface topography.