Computational finite element models of the skull and cervical spine derived from computed tomography (CT) scans are a promising tool for predicting and preventing traumatic brain injury. For pediatric heads and spines, the immaturity of the skull and spine tissue produces model geometries that are functions of the threshold used to produce them. This study seeks to provide a technique for selecting a threshold that develops a geometrically accurate representation of both the pediatric skull and cervical spine. To develop the model, two pediatric postmortem human subjects (PMHSs) were scanned at parameter settings that maximized resolutions;​ these parameters are not routinely used in clinical practice due to excess radiation exposure. The CT data sets were exported to AmiraTM 3.0 (TGS, Inc., San Diego, CA), where 3D isosurface images could be rendered. After imaging, the mandibles of the PMHSs were dissected and detailed anatomical measurements of the jaw were acquired. Using these measurements as a theoretical standard, the errors in the anatomical measurements taken from the rendered isosurfaces were minimized to find the threshold values that produce the least error. The threshold that produced the least error was used to produce 3D models of the skull and cervical spine. The results produced a thresholding technique that produces geometrically accurate skull and cervical spine models.