Automated identification of anatomical anomalies in the hyoid region of cleft lip and palate patients

Abstract

IntroductionSystematic identification of anatomical variants and anomalies in large patient populations remains problematic. The purpose of this study is to determine whether anatomical anomalies of the hyoid region can be quantitatively characterized, and corresponding subjects can be identified within a large cohort of cleft lip and palate (CLP) patients MethodsSoftware was developed (er3d.com) comprising Procrustes transformation (PT) followed by shape state extraction and thin plate spline description. Validation was achieved by simulating specific affine and non-affine geometric changes utilizing a set of 3D printed geometries and assessing expected outcomes. Subsequently, scans derived from cone beam computed tomography (CBCT) of three primary groups were analyzed including control subjects, unilateral CLP, and bilateral CLP patients with each affected group divided into either growing (<18 years of age) or non-growing (adult) subgroups. Cervical neck structures including the hyoid bone and contiguous vertebra were identified on CBCT scans and corresponding anatomical landmarks were analyzed. ResultsSize dominated the system with CLP patients displaying smaller hyoid regions in comparison to controls (90% variance reduction through PT; ANOVA, 6 linear distances, p ≤ .001; PCA eigenvector I = 79%). CLP patients displayed hyoid regions that were shorter and wider compared with control patients that were higher and narrower (TPS visualization). Shape states data were used to identify anatomical outliers graphically with landmarks #3, #4, #6 (hyoid body to the third cervical vertebra) showing greatest variability while three bilateral CLP patients are readily identifiable as outliers. ConclusionsThese data support the hypothesis that CLP affects size of the hyoid apparatus while shape is less affected, but contrasts patients based on height and width of the hyoid region. The methodology developed and applied here demonstrates a potential approach for rapid and automatic identification of patients with the most extreme anomalous hyoid anatomy within a large sample.