Automated and manual model creation workflows are equally reliable for virtual mechanical testing of ovine bone and fracture healing.

Abstract

Virtual mechanical testing on image-based bone models (digital twins) provides subject-specific insights about the mechanical behavior of a bone during fracture healing. However, the established workflows for these tests are limited by reliance on commercial software and time-consuming manual procedures needed to create the digital twins. To overcome these barriers to clinical adoption and scalability, we have developed methods for user-independent and automated model generation. This study aimed to: (1) compare four competing methods for digital twin creation (two manual versus two automated approaches), (2) assess the influence of model-creation procedures and choice of material model (single- and dual-zone) on the virtual test results, and (3) evaluate the accuracy of the model-creation techniques through experimental validation of the results. Digital twins were generated from 59 CT scans (33 operated osteotomy fractures, 26 contralateral intact bones). Torsional rigidities were compared between modeling workflows and validated using postmortem physical mechanical test data. There were no significant differences in torsional rigidity between any of the four virtual testing groups and physical testing when a dual-zone material model was implemented for bone and callus. These results confirm that virtual mechanical testing is a reliable alternative to physical mechanical testing for assessing intact and healing long bones, with resilience to variations in digital twin creation methods. Automated model creation was substantially faster than the manual approaches, suggesting that automatic digital twin analysis is the pathway toward future clinical scalability.