Physiological loading conditions are required for many biomechanical models of the human body. For example, the hip-joint loads during locomotion are essential to predict patient-specific fracture risk or to design hip implants. These loading conditions can be directly measured either ex vivo, which is impossible in clinical settings, or in vivo using invasive methods, such as instrumented prostheses, which is not applicable to healthy patients. A non-invasive and clinically applicable method are musculoskeletal models. However, they are complex and require numerous patient-specific parameters. Therefore, we employ another computational method in our work, which is called “inverse bone remodelling”. It takes advantage of bones’ ability to adapt its microstructure to mechanical stimuli. The background and ongoing research are described in the following. Bones are living organs and constantly adapt to their external loading. Thus, the physiological loading conditions are “imprinted” into the microstructure of the bone. It is possible to depict this microstructure using 3D imaging, e.g., computed tomography (CT). By applying computational models to these images, it is possible to back-calculate the loading conditions just based on the microstructure of the bone, i.e., to read and interpret the imprinted information. This method is referred to as “inverse bone remodelling” (IBR). The first IBR models were created by Fischer et al. in the mid-1990s [1]. They employed 2D finite element (FE) models of the proximal femur to predict the hip joint loadings. Christen et al. adapted the model of Fischer et al. by employing 3D micro-FE models [2]. Although micro-FE models are only computational feasible for smaller bones or sections of bones, they were successfully used by Synek and Pahr to predict hip-joint loading that matched in vivo data measured by instrumented prostheses well [3]. While micro-FE-based IBR predicts the loading conditions well, these models require up to weeks of computational time for a single bone and require high resolution CT images. Therefore, we translated the model of Christen et al. to a homogenized-FE version [4]. These models allow for much faster solving times while using the same mathematical background as the previous model; thus, they are directly comparable. In a recent study, we compared micro-FE-based to homogenized-FE-based IBR models and found generally good agreement between the models, especially in terms of peak loading [5]. Most recently, we started testing the applicability of homogenized-FE-based IBR to clinical CT images, a method that would allow in vivo prediction of patient-specific joint loads. Furthermore, we develop methods that can predict pressure distributions at the joints instead of predicting peak loads. Both methods can then be used in fracture risk assessment, or the development of implant systems that fit patient-specific needs much better than generic models. References [1] Fischer et al. Journal of Biomechanics 28, 1127–1135 (1995) [2] Christen et al. Biomechanics and Modeling in Mechanobiology 11, 483–492 (2011) [3] Synek et al. Biomechanics and Modeling in Mechanobiology 17, 843–852 (2017) [4] Bachmann et al. Annals of Biomedical Engineering (2022) [5] Bachmann et al. Computer Methods and Programs in Biomedicine 107549 (2023)