Biotypic classification of facial profiles using discrete cosine transforms on lateral radiographs.

Objective: The aim of the study was to use discrete cosine transforms to graph soft tissue curves in lateral cephalometric radiographs and, with the obtained mathematical values, to group these curves by both traditional biotypes and cluster systems, in order to evaluate discriminatory capacity in terms of accuracy.

Design: A sample of 625 lateral radiographs of adult patients (319 women and 306 men) was classified by facial biotype based on the ANB angle and mandibular plane angle. The curves of the facial profile were digitized with 50 equidistant points and discrete cosine transform was applied to analyze these curves mathematically for the determination of the accuracy of the classification of traditional biotypes. Phylogram cluster analysis was then performed for hierarchical grouping and accuracy was determined through cross-validation.

Results: Grouping by biotype was performed for men and women separately. Although significant, accuracy did not surpass 71.4%. In the groups by clusters, significant results were achieved when performing four analyses for men and two for women. The best accuracy regarding classification power and qualitative distinction was 89.5% for men and 94% for women.

Conclusions: Discrete cosine transforms using a cluster system had greater discriminatory capacity in terms of accuracy compared to traditional grouping considering the ANB angle and mandibular plane angle. This exploration can be useful for the creation of a soft-tissue facial reconstruction software for the Latin American population.