Surface topography asymmetry maps categorizing external deformity in scoliosis

Abstract Background context Adolescent idiopathic scoliosis (AIS) affects 2% to 4% of the population and predominantly affects female individuals. The scoliosis researchers and clinical communities use the “Cobb angle” obtained from anterior-posterior radiographs as the standard assessment tool for scoliosis. However, excessive radiation exposure over consecutive visits during the growing years increases the risk of cancer in young patients with AIS. Surface topography (ST) is a noninvasive method that is being investigated as an alternative tool for scoliosis assessment. The necessity of applying markers by skilled operators, which is time consuming and a potential area for errors, is one of the main limitations of these methods. Purpose This study introduces a three-dimensional markerless analysis technique for assessing torso asymmetry in AIS and a system for classifying patients based on this technique. The intra/interobserver and test-retest reliability and validity of the classification system was assessed. Study design A novel three-dimensional analysis technique of ST data of patients with scoliosis and its clinical applications. Methods Full-torso ST scans of 46 patients with AIS (Cobb angle: 34±15°, curve types: Lenke 1, 3, and 5) and five healthy subjects were used for analysis. The best plane of symmetry, dividing the torso into left and right, was calculated for each scan. The deviation between the original torso and its reflection with respect to the best plane of symmetry was illustrated using deviation contour maps. The subjects were visually classified into three main groups and six subgroups based on the number and location of the asymmetry contours. A second baseline scan and a 1-year follow-up scan were analyzed for 15 subjects and reliability of the method was assessed using kappa coefficients. Funding for this research is provided by the Scoliosis Research Society, Women and Children's Health Research Institute, and the Natural Sciences and Engineering Research Council of Canada. Results The intraobserver reliability of the group classification demonstrated excellent agreement with mean kappa coefficient of 0.85. The multiobserver kappa value of 0.62 was attained in the interobserver reliability test conducted among four observers classifying 46 subjects in three groups. The test-retest reliability of the method was assessed. Mean kappa values of 0.99 and 0.83 were achieved for group (three groups) and subgroup (six subgroups) classif