MechE PhD Prospectus Defense: Ahmad Hedayatzadeh Razavi

TITLE: A MULTISCALE COMPUTATIONAL FRAMEWORK FOR PREDICTING BONE FRACTURE HEALING: INTEGRATING MECHANOBIOLOGY, IMMUNE DYNAMICS, AND MACHINE LEARNING

ABSTRACT: Bone fracture healing is a complex regenerative process governed by tightly coupled mechanical, biological, and vascular mechanisms. While experimental studies have provided important insights into these processes, predicting healing trajectories and identifying early indicators of successful repair remain major challenges in orthopedic research. Existing computational models have advanced the understanding of mechanoregulation and tissue differentiation during fracture healing, but many approaches simplify spatial organization, neglect early immune dynamics, or remain computationally expensive for large-scale predictive applications. This research addresses these limitations by developing a multiscale computational framework to model fracture healing by integrating mechanics, biological signaling, and spatial tissue organization. The work begins with the development of a mechanobiology-based finite element model that incorporates unrestricted callus growth, angiogenesis, and mesenchymal stem cell migration to simulate bone healing in a rat femoral osteotomy model. Using experimentally derived inputs from micro-computed tomography and mechanical measurements, the model reproduces tissue differentiation and mechanical stiffness evolution during the early stages of healing, highlighting the central role of vascularization and progenitor cell recruitment in callus development. Building on this mechanistic framework, machine learning surrogate models are developed to enable rapid prediction of healing trajectories across diverse mechanical and biological scenarios. Training multiple algorithms on simulated healing datasets demonstrates that sequence-based models can accurately predict callus stiffness evolution while dramatically reducing computational cost, enabling efficient exploration of implant configurations and mechanical loading conditions. A further extension introduces a spatially resolved three dimensional modeling framework integrating finite element mechanics, strain-sensitive vascularization, and directed progenitor cell migration. This approach reveals that bone regeneration follows spatially organized mechanobiological pathways, with mineralization initiating in mechanically protected periosteal regions and propagating across the fracture callus. Early spatial features of callus formation are shown to contain predictive information about longterm mechanical outcomes. Ongoing work further expands this framework by incorporating inflammatory signaling through a hybrid agent-based model of early immune dynamics that explicitly simulates macrophage polarization, cytokine diffusion, and mesenchymal stem cell recruitment. Together, these efforts establish a predictive computational platform linking cellular signaling, mechanical regulation, and spatial tissue development during fracture repair. Building on these insights, current work also explores fracture callus geometry as a clinically measurable biomarker of healing progression, aiming to translate mechanobiological modeling insights into early-prediction tools that can guide orthopedic treatment and implant design.

COMMITTEE: ADVISOR/CHAIR Professor Ara Nazarian, ME; Professor Emma Lejeune, ME; Professor Michael Albro, ME/MSE/BME

When 12:00 pm - 1:00 pm on 15 April 2026
Building ENG 410, 110 Cummington Mall