The Open Bone Journal

2009, 1 : 16-22
Published online 2009 September 11. DOI: 10.2174/1876525400901010016
Publisher ID: TOBONEJ-1-16

Discrimination of Cervical and Trochanteric Hip Fractures Using Radiography-Based Two-Dimensional Finite Element Models

Jérôme Thevenot , Pasi Pulkkinen , Janne E. M. Koivumäki , Volker Kuhn , Felix Eckstein and Timo Jämsä
Department of Medical Technology, University of Oulu, P.O. Box 5000, 90014 Oulu, Finland.

ABSTRACT

Introduction: Predictors of fracture risk differ between cervical and trochanteric hip fractures. The aim of this experimental study was therefore to investigate whether two-dimensional (2D) finite element (FE) models, generated from standard radiographs, are able to predict and discriminate fracture types, originating from a simulated fall on the greater trochanter. Methods: A semi-automatic custom algorithm was applied to segment cortical and trabecular bone contours from radiographs of 49 female cadaver femora (mean age 80.7±10.3 years). Two types of 2D FE models were generated, either one or four material properties assigned to the trabecular bone. The cartilage and soft tissue were also simulated, and the boundary conditions were mimicking the experiment. VonMises stress distributions within the trabecular bone were evaluated and the regions of maximum continuous stress patterns were determined. Results: The best fracture type prediction was obtained with the criterion that a cervical fracture was predicted if the maximum stress in trabecular bone was located at the superior part of the femoral neck and the maximum continuous stress pattern through the neck region; and in all other cases a trochanteric fracture was predicted. The two different models predicted 79.6% and 85.7% of the fracture cases correctly, in comparison with the actual failure type. Conclusion: Our results suggest that the cervical and trochanteric hip fractures can be discriminated with a satisfactory level of accuracy, using a relatively simple radiographybased 2D model. Based on the current experimental findings, the predictive power of these models should now be tested in clinical studies.