A multi-objective optimization model for assessing spinal curvature in scoliosis

Introduction and Objectives: Previous studies have shown that the asymmetric postural adjustments had positive effect on the reduction of major spinal curvature in scoliosis. However, those applications did not consider the probable negative effect on the minor region(s) of scoliotic curvature. This study hypothesized that the optimal spinal loading could be achieved by controlled asymmetric load carriage for maximum correction of the affected region and minimum effect(s) on the unaffected region(s) in patients with scoliosis. The objective of this study was to develop an experimental protocol and formulate a multi-objective optimization (MOO) model for determining the optimal asymmetric loading condition for scoliosis. Methods: Subjects with scoliosis were instructed to maintain a standardized and relaxed barefoot erect stance posture. A cross-chest asymmetric load was applied at the level of anterior superior iliac spines to each subject randomly at two sides (left and right) and in six different weights (0, 2.5, 5, 7.5, 10 and 12.5% of body weight). The spinal profile was palpated and identified by stickers adhered to the spinous processes. Posterior-anterior view digital photos were taken to record the respective positions of the markers for determining spinal curvature. The mean spinal Cobb angles were determined. Regression equations predicting the mean Cobb angles with respect to the applied load were set up. A MMO model was formulated to achieve the prioritized multiple goals of optimizing the maximum reduction in the major curvature and minimum effects on the unaffected regions(s). The subject specific optimal asymmetric loading condition was determined. Results: Six young adults (mean age: 20.8 year, mean weight: 49.5 kg) in mild scoliosis participated in the study. The means of major and minor Cobb angles were 17.4° and 12.3° respectively. All of them had major curves at the thoracic region. Three had left curves and three had right curves. Different loads were applied to the left and right sides of the subjects. The mean Cobb angles with loading on the same and opposite sides of the major curves were 18.7° (+1.0°) and 14.2° (-3.2°) respectively. The effective side of the load determined by the MOO model was found to be on the opposite side of the major curve orientation. The optimal load determined by the MOO model was 3.2 kg (6.5% of BW). The predicted means of major and minor Cobb angles were 12.8° and 13.3° respectively. The predicted reduction in mean major Cobb angle was 4.6° (26.4% of initial Cobb angle) and increase in mean minor Cobb angle was only 1.0° (8.1% of initial mean minor Cobb angle). Conclusion: A cross-chest asymmetric load could be applied at the level of anterior superior iliac spines at the effective side opposite to the major curve orientation. The subject specific loading configuration could be determined by a multiobjective optimization model.