Quantitative Benchmarking of CBCT-Derived Finite Element Models Using Digital Image Correlation

Background/Objectives: Image-based finite element analysis (FEA) is increasingly used in dental biomechanics; however, its reliability is often limited by insufficient experimental benchmarking and a lack of standardized workflows. This study aimed to quantitatively benchmark a Cone beam computed tomography-based (CBCT) finite element pipeline using experimentally measured strain in restored human molars. Methods: Extracted human mandibular molars were restored using a total-etch adhesive system and bulk-fill composite resin. Specimen-specific finite element models were generated from CBCT data using a standardized segmentation and meshing workflow. Numerical simulations were compared with experimentally measured strain obtained during mechanical loading using Digital Image Correlation. Agreement between numerical and experimental data was assessed using regression analysis, Bland–Altman analysis, and equivalence testing. Results: A total of 304 spatially clustered paired measurements nested within 16 specimens were analyzed. FEM predictions showed strong correlation with experimental data (r = 0.91–0.97; R2 up to 0.937) and low relative error (~5–6%). The model systematically overestimated deformation by approximately 10–15%. Equivalence was confirmed within ±15% for dentin and composite, and within ±20% for enamel. Bland–Altman analysis revealed proportional bias and heteroscedasticity, particularly in dentin. Conclusions: The proposed CBCT-based finite element workflow demonstrates strong benchmarking agreement with experimental measurements and provides reproducible estimates of mechanical behavior within defined tolerance limits under controlled experimental conditions. Despite systematic overestimation, the model exhibits stable and reproducible behavior under controlled conditions. These findings support the use of experimentally benchmarked, image-based FEA workflows in dental biomechanical research.