Matching of lateral cephalograms by means of similarity coefficients

Abstract

Abstract This was a pilot study to establish the possible applications of taxonomic methods (taxonomy = study of the general principles of classification) to the identification of individuals. Similarity coefficients were calculated from data obtained from standardised orthodontic cephalograms and were used to identify individuals by a process of matching. There were two test groups and the individuals in the two groups were radiographed on two separate occasions. The first group consisted of young adults who were radiographed twice with approximately six weeks between radiographs. The second group consisted of orthodontic patients who were radiographed prior to and after the completion of orthodontic treatment. The aims were to test the effects of method error, measurement error and growth and treatment changes on the matching potential of similarity coefficients. The method was found to correctly identify or match all individuals in the first group of young adults, using cephalograms exposed approximately six weeks apart. The differences between the measurements obtained from those cephalograms represented method error and measurement error. In the second group, where the differences between radiographs further included changes due to growth and treatment, the rate of successful matching was not as high. Despite some limitations the method may be of use in two areas — screening and identification of individuals for forensic purposes and secondly for screening orthodontic cephalograms in diagnosis and treatment planning. The similarity coefficient chosen was squared Euclidean distance, with variables standardised to zero mean and unit variance to overcome the dimensionality problem caused by different units of measurement.