In this article the authors present a method for the backbone recognition and modelling. The process of recognition combines some classical techniques (Hough transformation, GVF snakes) with some new (authors present a method for initial curvature detection, which they call the Falling Ball method). The result enables us to identify high-quality features of the spine and to detect the major deformities of backbone: the intercrestal line, centre sacral vertical line, C7 plumbline; as well as angles: proximal thoracic curve, main thoracic curve, thoracolumbar/lumbar. These features are used for measure in adolescent idiopathic scoliosis, especially in the case of treatment. Input data are just radiographic images, meet in everyday practice.