Quantitative Three-Dimensional Characterization of Facial Dysmorphism in Self-Identified African-American Individuals With Prenatal Alcohol Exposure

Abstract

Background

The objective of this study was to characterize facial dysmorphology in a self-identified African-American cohort of children with prenatal alcohol exposure (PAE) and to evaluate the utility of computer vision-based methods for quantitative assessment and screening of facial form.

Methods

A total of 307 self-identified African-American individuals were evaluated by expert dysmorphologists and classified according to PAE status as unexposed controls (controls), individuals with prenatal alcohol exposed without fetal alcohol syndrome (FAS)-associated physical features (AE), or individuals with prenatal alcohol exposed with FAS-associated physical features (AE-FAS). Three-dimensional (3D) facial images were analyzed using volumetric and measurement-based approaches. Statistical shape modeling was applied to support surface-based analysis, cluster identification, and group discrimination testing.

Results

(i) Facial growth in the AE-FAS group was significantly reduced. (ii) We observed a subgroup of AE clinically without FAS-associated physical features who displayed an affinity to the AE-FAS group in the unscaled models. (iii) Sex-stratified analyses revealed alcohol-associated alterations in sexual dimorphism of facial morphology, with a substantially stronger association between PAE and reduced facial growth in females than in males. (iv) Unique patterns of facial dysmorphology were exhibited in the AE group, showing underdevelopment of the zygomatic arch. (v) Using closest mean, support vector machine, and linear discriminant analysis, discrimination between AE-FAS and controls was moderate.

Conclusions

3D facial analysis facilitates recognition of potential sex-related differences in facial morphology associated with PAE in African-American children, a population underrepresented in the clinical literature. The distinct patterns of dysmorphology observed may contribute to improved phenotypic characterization and diagnostic support.