Finite element (FE) models of the proximal femur are often used to study hip fracture. To interpret the results of these models, it is important to know whether the models accurately predict fracture location and/or type. This study evaluated the ability of automatically generated, CT scan-based linear FE models of the proximal femur to predict fracture location and fracture type. Fracture location was defined as the specific location of the fracture. Fracture type was a categorical variable defined as either a cervical or a trochanteric fracture. FE modeling and mechanical testing of 18 pairs of human femora were performed under two loading conditions, one similar to joint loading during single-limb stance and one simulating impact from a fall. For the stance condition, the predicted and actual fracture locations agreed in 13 of the 18 cases (72% agreement). For the fall condition, the predicted and actual fracture locations agreed in 10 of the 15 cases where the actual fractures could be identified (67% agreement). The FE models correctly predicted that only cervical fractures occurred in the stance configuration. For the fall configuration, FE-predicted and actual fracture types agreed in 11 of the 14 cases that could be compared (9 trochanteric, 2 cervical; 79% agreement). These results provide evidence that CT scan-based FE models of the proximal femur can predict fracture location and fracture type with moderate accuracy.
Keywords: Femur; Hip fracture; Finite element; Osteoporosis