Predicting Craniofacial Growth: A Geometric Morphometric Approach

Objectives: To evaluate differences in craniofacial ontogenetic patterns among facial types and assess the predictive capacity of geometric morphometric methods for craniofacial growth.
Methods: A set of 39 2D craniofacial landmarks were collected on longitudinal series of lateral cephalographs from the AAOF Legacy Collection, Fels Longitudinal Study, and Michigan Growth Study. Using mandibular plane angle (MP-SN), individuals were classified as hyperdivergent (〉39 degrees; 37 males, 45 females), normodivergent (28-39 degrees; 205 males, 266 females), or hypodivergent (〈28 degrees; 89 males, 68 females) at age 11.5 for girls and 13.5 for boys. Landmark configurations were superimposed using a generalized Procrustes analysis. Ontogenetic shape trajectories for each facial type were calculated using separate multivariate regressions (pooled by individual) of the Procrustes shape coordinates on chronological age. Resulting shape vectors were visualized to determine different ontogenetic patterns among facial types. Additionally, each trajectory was applied to the youngest landmark configurations (average=6.1yrs) for 9 test individuals (3 from each facial type; 5 females, 4 males) and scaled (“grown”) to the age at which the oldest radiographs were taken (average=19.0yrs). The predicted configurations were compared to the actual adult morphologies using Procrustes distance.
Results: Regardless of starting shape, the trajectories distinguish the hypodivergent group from the hyperdivergent group by forward facial and mandibular rotation, a posteriorly positioned and taller mandibular ramus, and a shorter, rounder mandibular symphysis (with additional nuanced craniofacial differences). The normodivergent group exhibits an intermediate shape. Estimated adult configurations were relatively accurate with 8 of the 9 actual adult configurations closest to the predicted morphology using the corresponding facial type trajectory.
Conclusions: This study highlights the nuanced but real differences among facial types and demonstrates the promising predictive capabilities of these methods which aid in our long-term goal of providing individualized predictions of craniofacial growth, ultimately improving decisions in treatment timing.