From the Laboratoire de Radiologie Expérimentale, CNRS UMR 7052, UFR Lariboisière-Saint-Louis, 2 rue Ambroise Paré, 75010 Paris, France (P.P., J.D.L., V.B.); Institute of Medical Physics, University of Erlangen, Erlangen, Germany (K.E., O.M.); LBM/Institut de Biomecanique Humaine Georges Charpak, Arts et Métiers ParisTech, Paris, France (L.D., W.S.); Department of Radiology, Hôpital Notre-Dame, Centre Hospitalier de l'Université de Montréal, Montréal, Québec, Canada (T.M.); Laboratoire de Biomécanique et Mécanique des Chocs-Université Lyon 1-IFSTTAR, Lyon, France (D.M.); Unité de Recherché Clinique Saint-Louis Lariboisère Fernand Widal, Paris, France (E.V.); and Department of Clinical Radiology, The Royal Infirmary, Imaging Science and Biomedical Engineering, University of Manchester, Manchester, England (J.A.).
Radiology. 2016 Sep;280(3):837-47. doi: 10.1148/radiol.2016142796. Epub 2016 Apr 14.
Purpose To evaluate the performance of three imaging methods (radiography, dual-energy x-ray absorptiometry [DXA], and quantitative computed tomography [CT]) and that of a numerical analysis with finite element modeling (FEM) in the prediction of failure load of the proximal femur and to identify the best densitometric or geometric predictors of hip failure load. Materials and Methods Institutional review board approval was obtained. A total of 40 pairs of excised cadaver femurs (mean patient age at time of death, 82 years ± 12 [standard deviation]) were examined with (a) radiography to measure geometric parameters (lengths, angles, and cortical thicknesses), (b) DXA (reference standard) to determine areal bone mineral densities (BMDs), and (c) quantitative CT with dedicated three-dimensional analysis software to determine volumetric BMDs and geometric parameters (neck axis length, cortical thicknesses, volumes, and moments of inertia), and (d) quantitative CT-based FEM to calculate a numerical value of failure load. The 80 femurs were fractured via mechanical testing, with random assignment of one femur from each pair to the single-limb stance configuration (hereafter, stance configuration) and assignment of the paired femur to the sideways fall configuration (hereafter, side configuration). Descriptive statistics, univariate correlations, and stepwise regression models were obtained for each imaging method and for FEM to enable us to predict failure load in both configurations. Results Statistics reported are for stance and side configurations, respectively. For radiography, the strongest correlation with mechanical failure load was obtained by using a geometric parameter combined with a cortical thickness (r(2) = 0.66, P < .001; r(2) = 0.65, P < .001). For DXA, the strongest correlation with mechanical failure load was obtained by using total BMD (r(2) = 0.73, P < .001) and trochanteric BMD (r(2) = 0.80, P < .001). For quantitative CT, in both configurations, the best model combined volumetric BMD and a moment of inertia (r(2) = 0.78, P < .001; r(2) = 0.85, P < .001). FEM explained 87% (P < .001) and 83% (P < .001) of bone strength, respectively. By combining (a) radiography and DXA and (b) quantitative CT and DXA, correlations with mechanical failure load increased to 0.82 (P < .001) and 0.84 (P < .001), respectively, for radiography and DXA and to 0.80 (P < .001) and 0.86 (P < .001) , respectively, for quantitative CT and DXA. Conclusion Quantitative CT-based FEM was the best method with which to predict the experimental failure load; however, combining quantitative CT and DXA yielded a performance as good as that attained with FEM. The quantitative CT DXA combination may be easier to use in fracture prediction, provided standardized software is developed. These findings also highlight the major influence on femoral failure load, particularly in the trochanteric region, of a densitometric parameter combined with a geometric parameter. (©) RSNA, 2016 Online supplemental material is available for this article.
目的 评估 3 种影像学方法（放射学、双能 X 射线吸收法[DXA]和定量计算机断层扫描[CT]）和数值分析有限元建模（FEM）在预测股骨近端失效负荷方面的性能，并确定预测髋部失效负荷的最佳骨密度或几何预测因子。
材料与方法 获得机构审查委员会批准。对 40 对切除的尸体股骨（死亡时患者的平均年龄，82 岁±12[标准差]）进行检查，方法为（a）放射学测量几何参数（长度、角度和皮质厚度），（b）DXA（参考标准）测定面积骨矿物质密度（BMD），（c）定量 CT 采用专用三维分析软件测定体积 BMD 和几何参数（颈轴长度、皮质厚度、体积和惯性矩），（d）定量 CT 基于 FEM 计算失效负荷的数值。80 根股骨通过机械测试断裂，每对股骨中的一根股骨随机分配到单肢站立姿势配置（以下简称站立姿势配置），配对股骨分配到侧位跌倒配置（以下简称侧位配置）。对每种影像学方法和 FEM 进行描述性统计、单变量相关性和逐步回归模型分析，以预测两种配置中的失效负荷。
结果 分别报告站立和侧位配置的统计数据。对于放射学，使用结合皮质厚度的几何参数与机械失效负荷的相关性最强（r2=0.66，P<0.001；r2=0.65，P<0.001）。对于 DXA，使用总 BMD（r2=0.73，P<0.001）和转子下 BMD（r2=0.80，P<0.001）与机械失效负荷的相关性最强。对于定量 CT，在两种配置下，最好的模型都结合了体积 BMD 和惯性矩（r2=0.78，P<0.001；r2=0.85，P<0.001）。FEM 分别解释了 87%（P<0.001）和 83%（P<0.001）的骨强度。通过结合（a）放射学和 DXA 和（b）定量 CT 和 DXA，与机械失效负荷的相关性分别增加到 0.82（P<0.001）和 0.84（P<0.001），对于放射学和 DXA 分别增加到 0.80（P<0.001）和 0.86（P<0.001），对于定量 CT 和 DXA。
结论 基于定量 CT 的 FEM 是预测实验失效负荷的最佳方法；然而，定量 CT 和 DXA 的组合性能与 FEM 一样好。如果开发了标准化软件，定量 CT 和 DXA 的组合可能更易于用于骨折预测。这些发现还强调了在股骨近端失效负荷中，特别是在转子区域，骨密度参数与几何参数的主要影响。
