RESUMO
OBJECTIVE: Apply elliptic Fourier analysis to find shape differences among the hypodivergent, normodivergent, and hyperdivergent growth patterns in skeletal classes I, II and, III in mandibular and maxillary curves and evaluate the discriminatory capacity of these differences. MATERIALS AND METHODS: A total of 626 adult patients were included: 354 Brazilian patients (52 with tomographic information and 302 with radiographic information) and 272 Colombian patients with radiographic information. Lateral views were selected. The maxillary and mandibular curves were digitized. Elliptic Fourier analysis was employed considering with 20 harmonics as well as filtering size, rotation, and translation properties. One-way non-parametric MANOVA was employed to determine differences. A confusion matrix tool was employed to analyze the discriminatory capacity of the model. RESULTS: Significant shape differences in the mandibular and maxillary contours were found among the hypodivergent, normodivergent, and hyperdivergent growth patterns in classes I, II, and III (p < 0.05). The accuracies obtained from the confusion matrix were respectively 74.1, 79.5, and 90.1% in classes I, II, and III in the mandibular curves and respectively 71.9, 73.9, and 75% in classes I, II, and III in the maxillary curves. CONCLUSIONS: Elliptic Fourier analysis can be used to find shape differences with an acceptable discriminatory capacity, especially in the mandible contour. Maxillary and mandibular bone curves each significantly defined facial biotypes regardless of the size and position properties. CLINICAL RELEVANCE: This exploration offers a way to quantify mandibular morphology for the construction of an economic mandibular prediction system applicable to the Latin American population.