摘要
目的利用计算机断层扫描技术探索耻骨联合表面形态学变化与年龄的相关性,并基于机器学习算法建立成人骨龄推断模型。方法收集10~90岁陕西省汉族腹部CT样本649例作为训练集,构建耻骨联合三维模型,选择7个形态学特征,即腹侧缘、背侧缘、联合面下端、联合面上端、腹侧斜面、耻骨结节和联合面沟嵴,建立新的耻骨联合形态特征评分系统,利用6种机器学习回归算法分别建立男性和女性年龄推断模型。分别使用85例临床CT样本(男性35例、女性50例)、96例死后CT样本(post-mortem CT samples,PMCT)(男性51例、女性45例)、82例真实耻骨联合样本(男性40例、女性42例)对各模型进行检验,选择最优的年龄推断模型。结果特征等级与年龄具有较强相关性(r>0.700,P<0.001)。男性样本中AdaBoost模型表现最佳,在临床CT、PMCT、真实耻骨联合三维重建模型上检验,得到较低的平均绝对误差(mean absolute error,MAE)分别为5.23、7.04和7.55岁,在40~70岁的年龄区间内MAE都小于10岁。女性样本中GBR模型表现最好,在临床CT、PMCT、真实耻骨联合三维重建模型上检验,得到较低的MAE,分别为5.16、5.02和5.71岁,在10~70岁的年龄区间内MAE都小于10岁。结论本研究构建的年龄推断模型具有一定的可靠性和准确性,可用于成人年龄推断。
Objective To explore the correlation between morphological changes on the surface of the pubic symphysis and age using computed tomography,and to develop an adult bone age estimation model using machine learning algorithms.Methods A total of 649 abdominal CT samples from Chinese Han population in Shaanxi Province aged 10 to 90 years were collected as the training set.After acquiring the 3D reconstructed model of the pubic symphysis,7 morphological features(ventral margin,dorsal margin,lower extremity of symphysial surface,upper extremity of symphysial surface,ventral beveling,pubic tubercle,and ridges and furrows on the symphysial surface)were selected with reference to the existing scoring system in order to establish a new scoring system for the morphological features of the pubic symphysis,and six machine learning regression algorithms were used to establish age estimation models for males and females,respectively.Each model was tested using 85 samples of clinical CT(35 males and 50 females),96 samples of post-mortem CT(PMCT)(51 males and 45 females),and 82 samples of real pubic symphysis(40 males and 42 females),respectively,to select the optimal age inference model.Results The scores obtained for the features all had a strong correlation with age(r>0.700,P<0.001).The AdaBoost model performed best in the male samples,with lower mean absolute error(MAE)of 5.23,7.04,and 7.55 years after testing on the clinical CT 3D reconstruction model,PMCT 3D reconstruction model,and real pubic symphysis 3D reconstruction model,respectively.The MAE was less than 10 years in the age interval of 40 to 70 years.The GBR model performed best in the female samples and also yielded lower MAE after examination on the clinical CT 3D reconstruction model,the PMCT 3D reconstruction model,and the real pubic symphysis 3D reconstruction model,which were 5.16,5.02,and 5.71 years,respectively.The MAE was less than 10 years in the age interval of 10 to 70 years.Conclusion The age inference model constructed in this study has a degree of reliability and accuracy.
作者
熊剪
曹永杰
马永刚
杨孝通
张吉
黄平
万昌武
XIONG Jian;CAO Yong-jie;MA Yong-gang;YANG Xiao-tong;ZHANG Ji;HUANG Ping;WAN Chang-wu(School of Forensic Medicine,Guizhou Medical University,Guiyang 550004,Guizhou Province,China;Academy of Forensic Science/Shanghai Key Laboratory of Forensic Medicine/Key Laboratory of Forensic Science of Ministry of Justice/Shanghai Professional Technical Service Platform of Forensic Science,Shanghai 200063,China;School of Basic Medical Sciences,Nanjing Medical University,Nanjing 211166,Jiangsu Province,China;Department of Medical Imaging,3201 Hospital Affiliated to Xi’an Jiaotong University,Hanzhong 723000,Shaanxi Province,China;School of Forensic Medicine,Shanxi Medical University,Taiyuan 030001,Shanxi Province,China)
出处
《复旦学报（医学版）》
CAS
CSCD
北大核心
2023年第5期731-742,共12页
Fudan University Journal of Medical Sciences
基金
国家自然科学基金(81722027)
上海市法医学重点实验室项目(17DZ2273200)
上海市司法鉴定专业技术服务平台项目(16DZ2290900)
中央级公益性科研院所项目(GY2020G-2)。
关键词
法医人类学
耻骨联合
CT
机器学习
年龄推断
成人
forensic anthropology
pubic symphysis
CT
machine learning
age estimation
adult
