椎弓足螺絲(pedicle screw)是用來矯正脊椎側彎的植入物，而椎弓足螺絲會因受到彎曲負荷導致斷裂或者椎體因骨質疏鬆而造成螺絲鬆脫等問題。此外，椎弓足螺絲斷裂與彎曲強度有關，螺絲鬆脫則與抗拉強度有關，又這兩種強度皆與螺紋設計有很大的關係。於是我們利用田口品質法(Taguchi Method)和類神經網路(Artificial Neural Networks)來得到螺紋設計參數之貢獻度。本研究目的是建立ㄧ個公平且合理的比較標準，評估何種方法能得到最準確之貢獻度，並以此結果解釋隱藏於類神經網路中的黑箱情形。
　　本次研究將資料代入田口品質法，以L25直交表中的設計參數組合，利用有限元素法(Finite Element Method)模擬出椎弓足螺絲的彎曲強度和抗拉強度，再使用變異數分析(Analysis of Variation)計算出每個因子的貢獻度。另外，也將直交表算出的結果匯入類神經網路，並訓練出十個模型以運用在各個方法之中，而計算貢獻度的方法有偏微分法(Partial derivative)、修正偏微分法(Modified PaD)、Garson權重法(Garson weight method)、Olden權重法(Olden weight method)、輪廓法(Profile method)、最小因子取代法(Minimum substitution)、因子權重移除法(Weight elimination)、最大因子取代法(Maximum substitution)、隨機因子取代法(Random substitution)、標準化數值對調法(Data permutation)與全範圍取代法(Full range substitution)。以變異數分析作為評估標準，找出重要因子的正確性、重要因子對輸出之正負相關、因子間排名正確度與相對貢獻度比例，藉此相互比較各種方法之優缺點。
　　本研究在計算貢獻度方面之結果，不論是彎曲強度分析還是抗拉強度分析，皆為全範圍取代法最接近變異數分析的結果。

　　Pedicle screw can be used for fixation of hip trochanteric fractures. It should have higher bending strength and pullout strength to resist breakage and loosening. Both of the strengths are relative to thread designs. We could get the optimal design factors of pedicle screw from neuro-genetic method which involves artificial neural networks(ANN) and genetic algorithm(GA). But the “black box” problem of ANN was not disclosed. The purpose of this study is to establish a fair standard of comparison and illuminate the black box by calculating contribution of each factor with different methods.
　　In our study, the bending and pullout functions of the pedicle screws were first simulated by finite element models(FEM). With L25 orthogonal arrays, the result of Taguchi robust design method with analysis of variation(ANOVA) is considered as the standard. The partial derivative(PaD), modified PaD, Garson weight method, Olden weight method, profile method, minimum substitution, weight elimination, maximum substitution, random substitution, data permutation and full range substitution in ANN are compared to find which one’s contribution is the most accurate. The performance of these methods was evaluated in four aspects, including selecting the important factors, finding the action direction of the important factors, ranking and relative contribution proportion.
　　In addition, using full range substitution could calculate accurate contribution of the design factors in bending and pullout analyses.

[1]H. Kurtaran, B. Ozcelik, and T. Erzurumlu, "Warpage optimization of a bus ceiling lamp base using neural network model and genetic algorithm," Journal of materials processing technology, vol. 169, pp. 314-319, 2005.
[2]K. Panneerselvam, S. Aravindan, and A. N. Haq, "Hybrid of ANN with genetic algorithm for optimization of frictional vibration joining process of plastics," The International Journal of Advanced Manufacturing Technology, vol. 42, pp. 669-677, 2009.
[3]X. Cui, M. Abbod, Q. Liu, J.-S. Shieh, T. Chao, C. Hsieh, et al., "Ensembled artificial neural networks to predict the fitness score for body composition analysis," The journal of nutrition, health & aging, vol. 15, pp. 341-348, 2011.
[4]M. Gevrey, I. Dimopoulos, and S. Lek, "Review and comparison of methods to study the contribution of variables in artificial neural network models," Ecological Modelling, vol. 160, pp. 249-264, 2003.
[5]J. D. Olden, M. K. Joy, and R. G. Death, "An accurate comparison of methods for quantifying variable importance in artificial neural networks using simulated data," Ecological Modelling, vol. 178, pp. 389-397, 2004.
[6]M. H. Shojaeefard, M. Akbari, M. Tahani, and F. Farhani, "Sensitivity analysis of the artificial neural network outputs in friction stir lap joining of aluminum to brass," Advances in Materials Science and Engineering, vol. 2013, 2013.
[7]H. Matsuzaki, Y. Tokuhashi, F. Matsumoto, M. Hoshino, T. Kiuchi, and S. Toriyama, "Problems and solutions of pedicle screw plate fixation of lumbar spine," Spine, vol. 15, pp. 1159-1165, 1990.
[8]Q. Zhu, W. W. Lu, A. D. Holmes, Y. Zheng, S. Zhong, and J. C. Leong, "The effects of cyclic loading on pull-out strength of sacral screw fixation: an in vitro biomechanical study," Spine, vol. 25, pp. 1065-1069, 2000.
[9]S. D. Cook, S. L. Salkeld, T. Stanley, A. Faciane, and S. D. Miller, "Biomechanical study of pedicle screw fixation in severely osteoporotic bone," The Spine Journal, vol. 4, pp. 402-408, 2004.
[10]J. L. Stambough, F. El Khatib, A. M. Genaidy, and R. L. Huston, "Strength and fatigue resistance of thoracolumbar spine implants: an experimental study of selected clinical devices," Journal of Spinal Disorders & Techniques, vol. 12, pp. 410-414, 1999.
[11]P.-Q. Chen, S.-J. Lin, S.-S. Wu, and H. So, "Mechanical performance of the new posterior spinal implant: effect of materials, connecting plate, and pedicle screw design," Spine, vol. 28, pp. 881-886, 2003.
[12]B. W. Cunningham, J. C. Sefter, Y. Shono, and P. C. McAfee, "Static and cyclical biomechanical analysis of pedicle screw spinal constructs," Spine, vol. 25, pp. 1S-12S, 2000.
[13]C. C. Hsu, C. K. Chao, J. L. Wang, S. M. Hou, Y. T. Tsai, and J. Lin, "Increase of pullout strength of spinal pedicle screws with conical core: biomechanical tests and finite element analyses," Journal of Orthopaedic Research, vol. 23, pp. 788-794, 2005.
[14]M. R. Mikles, F. A. Asghar, E. P. Frankenburg, D. S. Scott, and G. P. Graziano, "Biomechanical study of lumbar pedicle screws in a corpectomy model assessing significance of screw height," Journal of spinal disorders & techniques, vol. 17, pp. 272-276, 2004.
[15]G. D. Garson, "Interpreting neural-network connection weights," AI Expert, vol. 6, pp. 46-51, 1991.
[16]Y. Dimopoulos, P. Bourret, and S. Lek, "Use of some sensitivity criteria for choosing networks with good generalization ability," Neural Processing Letters, vol. 2, pp. 1-4, 1995.
[17]S. Lek, A. Belaud, P. Baran, I. Dimopoulos, and M. Delacoste, "Role of some environmental variables in trout abundance models using neural networks," Aquatic Living Resources, vol. 9, pp. 23-29, 1996.
[18]M. Scardi and L. W. Harding, "Developing an empirical model of phytoplankton primary production: a neural network case study," Ecological modelling, vol. 120, pp. 213-223, 1999.
[19]J. D. Olden and D. A. Jackson, "Illuminating the “black box”: a randomization approach for understanding variable contributions in artificial neural networks," Ecological modelling, vol. 154, pp. 135-150, 2002.
[20]M. Gevrey, I. Dimopoulos, and S. Lek, "Two-way interaction of input variables in the sensitivity analysis of neural network models," Ecological modelling, vol. 195, pp. 43-50, 2006.
[21]P. Heckerling, B. Gerber, T. Tape, and R. Wigton, "Entering the Black Box of Neural Networks A Descriptive Study of Clinical Variables Predicting Community-Acquired Pneumonia," Methods Inf Med, vol. 42, pp. 287-296, 2003.
[22]Q. Liu, X. Cui, M. F. Abbod, S.-J. Huang, Y.-Y. Han, and J.-S. Shieh, "Brain death prediction based on ensembled artificial neural networks in neurosurgical intensive care unit," Journal of the Taiwan Institute of Chemical Engineers, vol. 42, pp. 97-107, 2011.
[23]J. C. Gower, "A general coefficient of similarity and some of its properties," Biometrics, pp. 857-871, 1971.
[24]J. Podani, "Extending Gower's general coefficient of similarity to ordinal characters," Taxon, pp. 331-340, 1999.
[25]C. J. Willmott and K. Matsuura, "Advantages of the mean absolute error (MAE) over the root mean square error (RMSE) in assessing average model performance," Climate Research, vol. 30, p. 79, 2005.
[26]I. Dimopoulos, J. Chronopoulos, A. Chronopoulou-Sereli, and S. Lek, "Neural network models to study relationships between lead concentration in grasses and permanent urban descriptors in Athens city (Greece)," Ecological modelling, vol. 120, pp. 157-165, 1999.
[27]J. Montano and A. Palmer, "Numeric sensitivity analysis applied to feedforward neural networks," Neural Computing & Applications, vol. 12, pp. 119-125, 2003.
[28]T. Tchaban, M. Taylor, and J. Griffin, "Establishing impacts of the inputs in a feedforward neural network," Neural Computing & Applications, vol. 7, pp. 309-317, 1998.
[29]S. Lek, M. Delacoste, P. Baran, I. Dimopoulos, J. Lauga, and S. Aulagnier, "Application of neural networks to modelling nonlinear relationships in ecology," Ecological modelling, vol. 90, pp. 39-52, 1996.
[30]S. J. Kemp, P. Zaradic, and F. Hansen, "An approach for determining relative input parameter importance and significance in artificial neural networks," ecological modelling, vol. 204, pp. 326-334, 2007.