Colman K L, Janssen M C L, Stull K E, van Rijn R R, Oostra R J, de Boer H H, van der Merwe A E
Department of Medical Biology, Section Clinical Anatomy and Embryology, Academic Medical Center, University of Amsterdam, The Netherlands.
Department of Medical Biology, Section Clinical Anatomy and Embryology, Academic Medical Center, University of Amsterdam, The Netherlands.
Forensic Sci Int. 2018 May;286:268.e1-268.e8. doi: 10.1016/j.forsciint.2017.12.029. Epub 2017 Dec 23.
Sex estimation techniques are frequently applied in forensic anthropological analyses of unidentified human skeletal remains. While morphological sex estimation methods are able to endure population differences, the classification accuracy of metric sex estimation methods are population-specific. No metric sex estimation method currently exists for the Dutch population. The purpose of this study is to create Dutch population specific sex estimation formulae by means of osteometric analyses of the proximal femur. Since the Netherlands lacks a representative contemporary skeletal reference population, 2D plane reconstructions, derived from clinical computed tomography (CT) data, were used as an alternative source for a representative reference sample. The first part of this study assesses the intra- and inter-observer error, or reliability, of twelve measurements of the proximal femur. The technical error of measurement (TEM) and relative TEM (%TEM) were calculated using 26 dry adult femora. In addition, the agreement, or accuracy, between the dry bone and CT-based measurements was determined by percent agreement. Only reliable and accurate measurements were retained for the logistic regression sex estimation formulae; a training set (n=86) was used to create the models while an independent testing set (n=28) was used to validate the models. Due to high levels of multicollinearity, only single variable models were created. Cross-validated classification accuracies ranged from 86% to 92%. The high cross-validated classification accuracies indicate that the developed formulae can contribute to the biological profile and specifically in sex estimation of unidentified human skeletal remains in the Netherlands. Furthermore, the results indicate that clinical CT data can be a valuable alternative source of data when representative skeletal collections are unavailable.
性别估计技术经常应用于对身份不明的人类骨骼遗骸的法医人类学分析中。虽然形态学性别估计方法能够适应人群差异，但计量学性别估计方法的分类准确性是因人群而异的。目前还没有适用于荷兰人群的计量学性别估计方法。本研究的目的是通过对股骨近端进行骨测量分析，创建适用于荷兰人群的性别估计公式。由于荷兰缺乏具有代表性的当代骨骼参考人群，因此从临床计算机断层扫描（CT）数据中获得的二维平面重建图像被用作代表性参考样本的替代来源。本研究的第一部分评估了股骨近端十二项测量的观察者内和观察者间误差，即可靠性。使用26根干燥的成人股骨计算测量技术误差（TEM）和相对TEM（%TEM）。此外，通过一致性百分比确定干燥骨骼测量与基于CT的测量之间的一致性，即准确性。只有可靠且准确的测量数据被保留用于逻辑回归性别估计公式；一个训练集（n = 86）用于创建模型，而一个独立测试集（n = 28）用于验证模型。由于多重共线性水平较高，仅创建了单变量模型。交叉验证的分类准确率在86%至92%之间。高交叉验证分类准确率表明，所开发的公式有助于构建生物学特征，特别是在荷兰对身份不明的人类骨骼遗骸进行性别估计方面。此外，结果表明，当没有代表性的骨骼样本时，临床CT数据可以成为有价值的数据替代来源。
