Objective To establish and validate a biomechanical modeling method based on micro-magnetic resonance imaging (μMRI) and microstructure segmentation for noninvasively assessing microstructure behavior of the proximal femur. Methods Firstly, μMRI images were obtained from the femoral samples, and bone microstructures were segmented through regionized image processing to create the μMRI finite element model (μMRI model). Finite element analysis was performed utilizing a lateral fall posture simulation, and stress and strain were calculated. Secondly, the accuracy of μMRI image segmentation of bone microstructure was verified using micro-computed tomography (μCT), and the accuracy of μMRI model calculation results was verified using a finite element model constructed based on μCT (μCT model). Finally, the accuracy of bone surface strain calculated by μMRI model was verified through in vitro mechanical loading experiments simulating lateral falls and strain gauge measurements. Results The bone microstructure parameters BV/TV calculated by μMRI model and μCT model were significantly correlated (r=0.89, P<0.05). The maximum and minimum principal stress/principal strain percentiles calculated by μMRI model and μCT model were highly correlated (R2>0.9). The strain calculated by μMRI-FEM was highly correlated with the strain measured by mechanical experiments (R2=0.82). Conclusions The micro finite element model based on μMRI segmentation of bone microstructure can accurately characterize the micro-mechanical behavior of the proximal femur. This study provides an important tool for non-invasive assessment of hip femur microstructure degeneration and osteoporosis fracture risk in vivo.