Automatic identification of three-dimensional morphometric features of vertebrae

To avoid variability and subjectivity in the measurement of vertebral morphometric features, we developed a system to automatically identify their landmarks and extract three-dimensional (3D) features. On a 3D vertebral model reconstructed from computerised tomography (CT) scans, the symmetry plane was identified. Based on this plane, the coronal planes and the reference frame were determined. A total of 12 morphometric features and 14 landmarks were identified based on these derived planes and frames. The identification generated by the system was validated according to the manual annotations of two orthopaedic experts. The automatically identified landmarks and features agreed with the manual identifications, with average differences less of than 1.91 mm and 1.6 for the landmarks and features, respectively. Paired t-tests showed no statistically significant differences between auto-user variability and inter-user variability in the identified landmarks or the extracted features (p > 0.05). These results indicated that the proposed approach can be an aid in the task of measuring vertebral morphometric features in medicine, anthropological, and forensic investigations.