Fractures of the acromion are a common complication following surgical procedures of the shoulder due to changes in joint biomechanics. To improve understanding of acromial stresses and evaluate surgical procedures, image-based finite element models (FEMs) may be used. Image-based FEMs are dependent on accurate volumetric bone mineral density (vBMD), as this relates to mechanical properties in FEMs. Image reconstruction kernel alters vBMD;​ however, the effect on FEM output during simulated loading of the acromion has not been reported. The objective of this study was to compare predicted forces from FEMs derived from two common kernels and four density-modulus relationships to experimental forces in cadaveric scapulae (n = 10). Scapular FEMs were generated from CT scans reconstructed using bone sharpening (BONE) and standard (STD) kernels using four density-modulus relationships. Displacements were applied corresponding to experimental data collected on cadaveric specimens and forces were compared for each FEM. Specimen-specific percentage errors were as low as 1 % when using BONE kernel vBMD as input. Across all FEMs, the most accurate density-modulus relationship had a lower mean absolute percentage error (40 %) compared to the other three relationships compared (275 %, 281 %, 547 %), which greatly overestimated experimental forces. Across all models, those derived with STD kernel vBMD (40 %) had lower mean absolute percentage error relative to BONE kernel vBMD (42 %). This study highlights the relative accuracy of current density-modulus relationships using vBMD from two common reconstruction kernels. More accurate density-modulus relationships that account for variations in kernel parameters are required for FEM estimates of acromial forces and fracture predictions. Current models are not able to replicate experimental forces in cadaveric scapulae.