Statistical Method for Identification of Deformity Regions and Automatic Scoliosis Diagnosis

The paper introduces a new method to automatically identify deformity regions and provide diagnosis of adolescent idiopathic scoliosis. Starting from the 3D coordinates of the vertebrae and disc centroids, the spine curve is aligned vertically, scaled, and interpolated before performing a general statistical analysis to estimate scoliosis level, as well as to compute local measures such as the mathematical curvature and slope to identify deformed regions. The results are compared to reference Cobb angles computed on the thoracic and lumbar regions, showing a high correlation and easily explain differences in the case of outliers.