Identification of the relapsing Class III face: Chernoff faces in orthodontics

Thirty Class III patients, treated by a standard non-extraction technique, provided the material for this study. Records were examined at the beginning of treatment and a minimum of 2 years out of all retention. On the basis of their post-retention occlusions, children were sub-divided into three groups: stable, borderline, and relapsed. Using this classification, a four-variable discriminant model had been generated which correctly indicated the eventual outcome from the pretreatment data. A graphical technique not previously described in orthodontics and suitable for displaying multivariate data: 'Chernoff faces', provided a pictorial illustration of the discriminant values for each individual. These stylized, cartoon faces were examined to determine whether any common, characteristic facial patterns existed at the beginning of treatment in children of a single outcome group. Whilst the information given by the faces was easier to assimilate than tabulated data, no typical stable or relapsing face could be identified using the features chosen. The reason for this and further possible applications of Chernoff faces to orthodontics are discussed.