Data-driven standards for infant skull thickness distributions in computational modeling and analysis.

Abstract

Child abuse remains a global issue, with infants under 1 year of age facing the highest risk of fatality and recurrence if abuse is not detected. Computational modeling is a powerful tool for predicting injury from real-world trauma, offering a means to validate caretaker-reported histories and prevent further abuse. A key challenge and gap, however, lies in capturing the natural anatomical variability within a population to enhance injury prediction accuracy. This study addresses this gap by quantifying skull thickness distributions in a robust sample (n = 266) and establishing data-driven anatomical standards based on similarities in thickness patterns. The study examined age and head circumference as predictors of skull thickness growth. For infants younger than 2.5 months, head circumference was a more reliable predictor than age. Infants under 12 months old were categorized into four age groups-0-1.5, 1.5-5.9, 5.9-10.2, and 10.2-12 months-using natural thickness distribution breaks and a variance optimization routine. No significant sex differences were found in average skull thickness within each cranial bone (left and right parietal, frontal, and occipital), but there were 53 locations with significant sex differences at various stages of development. Symmetry tests suggested that lateral symmetry may be an appropriate assumption for infants under 12 months. Representative thickness distributions for each age group were selected based on similarity scores. This study is the first to apply data-driven methods to categorize infant skull thickness distributions, generating essential guidelines for age- and sex-based models in predicting injury from head trauma in infants.