Martin-de-Las-Heras Stella, Tafur Daniel, Bravo Manuel
Department of Forensic Medicine and Forensic Odontology.
Acta Odontol Scand. 2014 Jul;72(5):331-6. doi: 10.3109/00016357.2013.826383. Epub 2013 Aug 26.
To develop a quantitative method to compare 3D overlays from dental casts with experimental bitemarks by using geometric morphometric analysis.
Thirteen upper and lower dental casts and corresponding simulated bitemarks were 3D-scanned to generate comparison overlays with DentalPrint software(©). This study considered the inter-canine distance and four incisal angles. A matrix was created to compare all possible combinations of matches and non-matches between models and bites, i.e. 169 combinations (13 models × 13 bites), of which 13 were true matches. For each combination, the percentage difference was calculated between the variables in the model and the same variables in the bitemark. Logistic regression was used to obtain a predictive model (algorithm) for a match, calculating the discriminative values (area under the ROC curve, sensitivity and specificity) for each measure and for the logistic model.
Statistically significant discriminative power was found for all single (angle or distance) and combined (logistic model) variables, with lower 95% CI limits > 0.50 for areas under the ROC curves and sensitivity/specificity values > 50% in both maxilla and mandible.
This quantitative method has sufficient discriminative power to be utilized in forensic cases.
通过几何形态测量分析，开发一种定量方法，用于比较牙模的三维叠加图与实验咬痕。
对13副上下颌牙模及相应的模拟咬痕进行三维扫描，以使用DentalPrint软件（©）生成比较叠加图。本研究考虑了犬齿间距离和四个切牙角。创建了一个矩阵，以比较模型与咬痕之间所有可能的匹配和不匹配组合，即169种组合（13个模型×13个咬痕），其中13种为真实匹配。对于每种组合，计算模型中的变量与咬痕中相同变量之间的百分比差异。使用逻辑回归获得匹配的预测模型（算法），计算每种测量方法以及逻辑模型的判别值（ROC曲线下面积、敏感性和特异性）。
发现所有单一（角度或距离）和组合（逻辑模型）变量均具有统计学显著的判别力，上颌和下颌的ROC曲线下面积的95%置信区间下限均>0.50，敏感性/特异性值均>50%。
这种定量方法具有足够的判别力，可用于法医案件。
