Radiologists interpret the normality of vertebras within cervical spine x-rays for determining the presence of osteoporosis. Features relating to vertebra fracture such as contrasting anterior and posterior heights are used in the vertebra normality assessment. Vertebra distortion along the anterior boundary can be used as an indicator of vertebra normality. As the vertebra becomes less normal in appearance, the vertebra boundary increasingly deviates from the general rectangular shape. This research introduces a radius of curvature-based approach to quantify the distortion in vertebra boundary shape for anterior osteophytes classification of vertebras within cervical spine x-ray images. An adaptive histogram analysis technique is used to enhance the vertebras within cervical spine x-ray images to facilitate vertebra feature extraction. A least squares technique is used for calculating the radii of curvature along the vertebra boundary. Radius of curvature and gray-level based features are computed using the anterior boundary point position with the minimum radius of curvature as the reference. A total of 118 cervical spine vertebras are used for training and testing various classifiers including a standard back propagation neural network, K-Means algorithm, quadratic discriminant classifier and LVQ3. The results from those classifiers are reported for recognizing vertebras as normal or abnormal.