Date on Master's Thesis/Doctoral Dissertation


Document Type

Doctoral Dissertation

Degree Name

Ph. D.


Mechanical Engineering

Degree Program

Mechanical Engineering, PhD

Committee Chair

Voor, Michael John

Committee Co-Chair (if applicable)

Bradshaw, Roger D

Committee Member

Bradshaw, Roger D

Committee Member

Berfield, Thomas Austin

Committee Member

El-Baz, Ayman Sabry

Author's Keywords

Bone modeling; FEA; biomechanical testing; image processing; validation


The most common hip fracture in the elderly occurs as a result of a fall to the side with impact over the greater trochanter resulting in a fracture of the proximal femur. The fracture usually involves the femoral neck or the intertrochanteric region. It has recently been determined that the fracture crack of a hip fracture typically initiates on the superior-lateral cortex of the femoral neck and then propagates across the femoral neck, resulting in a complete fracture. The strength of the superior-lateral cortex of the femoral neck is likely determined by the combined properties of the generally thin cortex (outer layer) and the underlying trabecular bone in this region. The objective of this study was to determine the relative effects of increasing or decreasing the thickness of these bone tissues on the overall failure strength of the proximal femur. The clinical significance of this work relates to hip fracture risk with various potential treatment options to improve either cortical or trabecular bone quality. A human femur obtained from a 68 year old female donor was scanned using computed tomography at 60-micron voxel resolution and a series of high-resolution finite element models were generated. The models were constructed with a base-element dimension of 120 microns and models included a basic model with cortical and trabecular thicknesses representative of the cadaver specimen from the original scan. Other models used a standardized algorithm to either dilate or erode the trabecular and cortical bone compartments of the femoral neck so that a total of nine models were created including the basic model. Each model was used to simulate a fall-to-the-side loading condition with appropriate boundary and loading conditions as used in previous models and experiments. An experimental test of the cadaver femur was also performed with three strain gauges placed on the proximal femur: on the superior-lateral cortex, on the inferior-medial cortex, and on the medial cortex positioned distal to the lesser trochanter. This femur was loaded at a rate of 100 mm/s until fracture of the femoral neck using a standard fall-to-the-side setup and the applied load and gauge strains were recorded. The femur neck fractured at a load of 2140 N. To validate the basic finite element model, the strain gauge strains at the load levels of 1000 N and 2000 N were compared to the calculated strains from the basic model at the same loads and same location as the gauge on the cadaver femur. After the basic model was validated, a failure criterion was determined as the volume percentage of the elements in the model that had exceeded 7000 µε at the failure load corresponding to the load at which the cadaver femur failed. Subsequently, this failure criterion was applied to the other eight models as a parametric analysis to estimate the increase or decrease in failure strength caused by the changes in cortical and trabecular thickness. The validation test results showed that the basic finite element model calculated strain on the superolateral cortex was within 2.1% of the experimentally measured strain at 1000 N loading. The validated basic model was then used to determine that the percentage of finite elements (by volume of the model) in excess of 7000 µε at the failure load was 4.2%. This failure criterion was then used to estimate the failure load for the other eight models with different combinations of either thicker (+120 µm) or thinner (-120 µm) cortex and trabeculae in the femoral neck. The calculated failure loads ranged from 324 N for the model with thinned cortex and thinned trabeculae to 3336 N for the model with thickened cortex and thickened trabeculae. The model with normal cortex and thickened trabeculae had a failure load of 3242 N, which is only 2.8% less than the strongest case. The largest single parameter effect on proximal femoral strength is realized by an increase in trabecular thickness. This is somewhat surprising considering that cortical bone is typically stronger than cancellous bone. However, the spatial arrangement of trabecular bone and the buttress support it provides to the thin cortex apparently plays an important role in the ability of a global increase in thickness to have a significant beneficial effect.