Vatzia Konstantina, Fanariotis Michail, Bugajski Maciej, Fezoulidis Ioannis V, Piagkou Maria, Vlychou Marianna, Triantafyllou George, Vezakis Ioannis, Botis George, Papadodima Stavroula, Matsopoulos George, Vassiou Katerina
Faculty of Medicine, University of Thessaly, Biopolis, 41110 Larissa, Greece.
Department of Radiology, Sykehuset Telemark HF, 3710 Skien Telemark, Norway.
Diagnostics (Basel). 2025 Jun 27;15(13):1649. doi: 10.3390/diagnostics15131649.
This study aimed to assess the potential of sternal morphometric parameters derived from multidetector computed tomography (MDCT) for sex estimation in a contemporary Greek population. A secondary objective was to develop and evaluate statistical and machine learning models based on these measurements for forensic identification. : Sternal measurements were obtained from chest MDCT scans of 100 Greek adults (50 males, 50 females). Morphometric variables included total sternum length, surface area, angle, and index (SL, SSA, SA, and SI); manubrium length, width, thickness, and index (MBL, MBW, MBT, and MBI); sternal body length, width, thickness, and index (SBL, SBW, SBT, and SBI); and xiphoid process length and thickness (XPL and XPT). Logistic regression and a Random Forest classifier were applied to assess the predictive accuracy of these parameters. : Both models showed high classification performance. Logistic regression identified MBL and SBL as the most predictive variables, yielding 91% overall accuracy, with 92% sensitivity and 90% specificity. The Random Forest model achieved comparable results (91% accuracy, 88% sensitivity, 93% specificity), ranking SSA as the most influential feature. : MDCT-derived sternal morphometry provides a reliable, non-invasive method for sex estimation. Parameters such as MBL, SBL, and SSA demonstrate strong discriminatory power and support the development of population-specific standards for forensic applications.
本研究旨在评估源自多排螺旋计算机断层扫描（MDCT）的胸骨形态测量参数在当代希腊人群中进行性别估计的潜力。第二个目标是基于这些测量结果开发并评估用于法医鉴定的统计模型和机器学习模型。从100名希腊成年人（50名男性，50名女性）的胸部MDCT扫描中获取胸骨测量数据。形态测量变量包括胸骨总长度、表面积、角度和指数（SL、SSA、SA和SI）；柄的长度、宽度、厚度和指数（MBL、MBW、MBT和MBI）；胸骨体长度、宽度、厚度和指数（SBL、SBW、SBT和SBI）；以及剑突长度和厚度（XPL和XPT）。应用逻辑回归和随机森林分类器来评估这些参数的预测准确性。两个模型均显示出较高的分类性能。逻辑回归确定MBL和SBL为最具预测性的变量，总体准确率为91%，敏感性为92%，特异性为90%。随机森林模型取得了类似的结果（准确率91%，敏感性88%，特异性93%），将SSA列为最具影响力的特征。MDCT衍生的胸骨形态测量为性别估计提供了一种可靠的非侵入性方法。诸如MBL、SBL和SSA等参数显示出强大的鉴别力，并支持开发针对法医应用的特定人群标准。
