Published in:
01-03-2014 | Original Article
Posture class prediction of pre-peak height velocity subjects according to gross body segment orientations using linear discriminant analysis
Published in: European Spine Journal | Issue 3/2014
Login to get accessAbstract
Background/purpose
Measurement and classification of standing posture in the sagittal plane has important clinical implications for adolescent spinal disorders. Previous work using cluster analysis on three gross body segment orientation parameters (lower limbs, trunk, and entire body inclination) has identified three distinct postural groups of healthy subjects before pubertal peak growth: “neutral”, “sway-back”, and “leaning-forward”. Although accurate postural subgrouping may be proposed to be crucial in understanding biomechanical challenges posed by usual standing, there is currently no objective method available for class assignment. Hence, this paper introduces a novel approach to subclassify new cases objectively according to their overall sagittal balance.
Methods
Postural data previously acquired from 1,196 pre-peak height velocity (pre-PHV) subjects were used in this study. To derive a classification rule for assigning a class label (“neutral”, “sway-back”, or “leaning-forward”) to any new pre-PHV subjects, linear discriminant analysis was applied. Predictor variables were pelvic displacement, trunk lean and body lean angle. The performance of the newly developed classification algorithm was verified by adopting a cross-validation procedure.
Results
The statistical model correctly classified over 96.2 % of original grouped subjects. In the cross-validation procedure used, over 95.9 % of subjects were correctly assigned.
Conclusions
Based on three angular measures describing gross body segment orientation, our triage method is capable of reliably classifying pre-PHV subjects as either “neutral”, “sway-back”, or “leaning-forward”. The discriminant prediction equations presented here enable a highly accurate posture class allocation of new cases with a prediction capability higher than 95.9 %, thereby removing subjectivity from sagittal plane posture classification.