Identification of statistical critical area of proximal femur when a lateral fall happens.

Osteoporotic hip fracture is considered a worldwide health problem in the elderly population, with a negative social and economic impact characterized by an increase of hospital cost and a worsening of quality of life. Diagnostic tools widely used for assessment of osteoporosis include Dual-energy X-ray absorptiometry (DXA), Quantitative Computed Tomography (QCT), FRAX. Although DXA is the gold standard for osteoporosis diagnose, it is not enough reliable for fracture prediction. For this reason, the Finite Element (FE) model rises as a complementary tool to tackle the fracture prediction. In this sense, the evaluation of critical fracture regions on FE model, is often based on visual identification of mapped field introducing under/over estimation of the local nature of fracture event. The present thesis provides an innovative rigorous methodology applied on FE proximal femur models generated from DXA images, to identify statistically significant difference between fracture and non-fracture group through the application of Random Field Theory (RFT) and its topological extension of statistical process based on Statistical Parametric Map (SPM). The “spm1d” software has been used to compute statistical tests. The investigated groups include 111 osteoporotic subjects: 62 fractured patients and 49 controls. Fracture region (neck/trochanter), type of tissue (trabecular/cortical) and gender (female/male) were considered in the analyses (two-sample t test and two-way ANOVA with repeated measures) to explore the impact of these factors on the dependent variables Major Principal Stress (MPS) and Major Principal Strain (MPE) obtained by FE simulations. In particular, the significant elements of FE model detected by performing the two-sample t test, are implicated in the second level of analysis to take into consideration the interaction between factors. The results showed the variable MPS as the main significant parameter to discriminate the investigated groups. In relation to the zone detected as statistically significant, it was observed that not necessarily corresponds to the fracture region. In addition, the elements of the FE model belonging to the regions identified as critical represent only a tiny percentage of neck/trochanter area. To verify the reliability of this method of analysis, a comparison between the classification made considering actual data belonging to a reduction of neck and trochanter region and one more conservative, that consider all elements of neck and trochanter was performed. The results obtained demonstrate an improvement of predictive power when a reduction of critical regions is considered, reaching a 79% and 89% in the classification of neck and trochanter fracture, respectively. The implications of these findings suggest an important advance: on one hand in clinical practice, about prevention and early diagnosis and, on the other hand, in biomechanical field concerning the bone behaviour in relation to the event of hip osteoporotic fracture.