MRI-derived skull parameters for early, normative, pediatric cranial evaluation using image synthesis

Introduction: Determining normative cranial progression supports pediatric diagnosis and treatment of developmental diseases and disorders. However, as computed tomography (CT) emits radiation, it is contraindicated for children without substantial deformities. This study uses magnetic resonance imaging (MRI) to evaluate craniometrics traditionally measurable exclusively in CT including maximum cephalic length (CL), maximum cephalic breadth (CB), cephalic index (CI), cranial circumference (CC), intracranial volume (ICV), and average cranial density (ACD). Methods: Paired MRI-CT images were retrospectively obtained for 116 infants (64 male; 52 female; 0-2 years). Patients with any brain/cranial abnormalities were excluded. A custom SwinUNETRv2 within a generative adversarial network schema generated synthetic CT (sCT) from T1-weighted MRI. Then, a novel 3DSlicer extension performed cranial shape analysis on real CTs and sCTs, automatically computing 6 established craniometrics: CL, CB, CI, CC, ICV, and ACD. Image-to-image metrics assessed sCT realism against ground truth CTs. Craniometrics obtained from sCTs and real CTs were assessed for equivalency significance via two one-sided t-tests (TOST) equivalence analysis. Results: All sCT craniometrics were statistically equivalent to real CTs using TOST (−ΔL=-5, ΔU=5, p<0.05). The sCTs had trivially higher CC, CB, CI and ACD values versus true CTs; average CC was 445.9±36.18mm versus true average CC of 437.2±38.34mm, average CB was 115.20±13.70mm versus true average CB of 114.57±13.07mm, average CI was 82.57±5.42mm versus true average CI of 82.03±4.99mm, and ACD was 3.31±1.68mm versus true ACD of 3.27±1.62mm. CL and ICV were fractionally lower on sCTs with average CL values of 139.82±17.32mm against true average CL values of 140.01±17.38mm and an average ICV of 945.21±221.45cm³ against a true average ICV of 950.79±225.76cm³. Conclusion: Significant equivalence between sCT and CT craniometrics indicates sCTs can be useful for the clinical evaluation of craniofacial morphology. The presented approach establishes craniometric parameters in a young, normative, pediatric cohort while simultaneously offering a method to eliminate radiation exposure by computing clinically relevant cranial metrics from T1-weighted MRI.