Stephan Carl N, Amidan Brett, Trease Harold, Guyomarc'h Pierre, Pulsipher Trenton, Byrd John E
Joint POW/MIA Accounting Command, Central Identification Laboratory, 310 Worchester Avenue, Building 45, Hickam Air Force Base, HI, 96853.
J Forensic Sci. 2014 Mar;59(2):306-13. doi: 10.1111/1556-4029.12324. Epub 2013 Dec 6.
This paper describes a computerized clavicle identification system primarily designed to resolve the identities of unaccounted-for U.S. soldiers who fought in the Korean War. Elliptical Fourier analysis is used to quantify the clavicle outline shape from skeletons and postero-anterior antemortem chest radiographs to rank individuals in terms of metric distance. Similar to leading fingerprint identification systems, shortlists of the top matching candidates are extracted for subsequent human visual assessment. Two independent tests of the computerized system using 17 field-recovered skeletons and 409 chest radiographs demonstrate that true-positive matches are captured within the top 5% of the sample 75% of the time. These results are outstanding given the eroded state of some field-recovered skeletons and the faintness of the 1950's photofluorographs. These methods enhance the capability to resolve several hundred cold cases for which little circumstantial information exists and current DNA and dental record technologies cannot be applied.
本文描述了一种计算机化的锁骨识别系统，其主要设计目的是确定在朝鲜战争中失踪的美国士兵的身份。椭圆傅里叶分析用于从骨骼和生前胸部后前位X光片中量化锁骨轮廓形状，以便根据度量距离对个体进行排名。与领先的指纹识别系统类似，提取最匹配候选人的入围名单以供后续人工视觉评估。使用17具现场回收的骨骼和409张胸部X光片对该计算机化系统进行的两项独立测试表明，75%的情况下，真阳性匹配出现在样本的前5%中。考虑到一些现场回收骨骼的腐蚀状态以及20世纪50年代荧光照片的模糊程度，这些结果非常出色。这些方法增强了解决数百起几乎没有间接信息且无法应用当前DNA和牙科记录技术的悬案的能力。
