Background: Finite element modeling of human bone provides a powerful tool to evaluate a wide variety of outcomes in a highly repeatable and parametric manner. These models are most often derived from computed tomography data, with mechanical properties related to bone mineral density (BMD) from the x‐ray energy attenuation provided from this data. To increase accuracy, many researchers report the use of quantitative computed tomography (QCT), in which a calibration phantom is used during image acquisition to improve the estimation of BMD. Since model accuracy is dependent on the methods used in the calculation of BMD and density‐mechanical property relationships, it is important to use relationships developed for the same anatomical location and using the same scanner settings, as these may impact model accuracy. The purpose of this literature review is to report the relationships used in the conversion of QCT equivalent density measures to ash, apparent, and/or tissue densities in recent finite element (FE) studies used in common density‐modulus relationships. For studies reporting experimental validation, the validation metrics and results are presented.Results: Of the studies reviewed, 29% reported the use of a dipotassium phosphate (K₂HPO₄) phantom, 47% a hydroxyapatite (HA) phantom, 13% did not report phantom type, 7% reported use of both K₂HPO₄ and HA phantoms, and 4% alternate phantom types. Scanner type and/or settings were omitted or partially reported in 31% of studies. The majority of studies used densitometric and/or density‐modulus relationships derived from different anatomical locations scanned in different scanners with different scanner settings. The methods used to derive various densitometric relationships are reported and recommendations are provided toward the standardization of reporting metrics.Conclusions: This review assessed the current state of QCT‐based FE modeling with use of clinical scanners. It was found that previously developed densitometric relationships vary by anatomical location, scanner type and settings. Reporting of all parameters used when referring to previously developed relationships, or in the development of new relationships, may increase the accuracy and repeatability of future FE models.

Image-based finite element (FE) modeling of bone is a non-invasive method to estimate bone stiffness and strength. High-resolution imaging data as input allows for inclusion of bone microarchitecture but results in large amounts of data unsuitable for traditional FE solvers. Bone-specific mesh-free solvers have been developed over the past 20 years to improve on memory efficiency in simulated bone loading applications. The objective of this study was to provide linear performance benchmarking for a bone-specific, mesh-free solver (FAIM) using µCT and HR-pQCT image data on Mac, Linux, and Windows operating systems using both single- and multi-thread CPU and GPU processing. The focus is on the linear gradient-descent solver using standardized uniaxial loading of bone models from µCT, and first- and second-generation HR-pQCT scans of the radius and tibia. Convergence, speedup, memory, and batch performance tests were completed using CPUs and GPUs on three laboratory-based systems with Windows, Linux, and Mac operating systems. Although varying by system and model size, time-per-iteration was as low as 0.03 s when an HR-pQCT-based radius model (6.45 million DOF) was solved with 3 GPUs. Strong scaling was achieved with GPU and CPU parallel processing, with strong parallel efficiencies when models were solved using 3 GPUs or ≤ 10 CPU threads. Errors in force, strain energy density, and Von Mises stress were as low as 0.1% when a convergence tolerance of 10^-3 or smaller was used. The results of this study indicate that to maximize computational efficiency and minimize model solution times using FAIM software under the standardized tested conditions using µCT, XCT1 and XCT2 HR-pQCT image data, convergence tolerance set to 10^-4, and 10 threads or 2 GPUs are sufficient for efficient solution times. Less strict convergence tolerances will improve solution times but will introduce more error in the outcome measures.