干颅骨锥形束CT硬组织标志点手工标注的可靠性及准确性研究  

Accuracy and reliability of craniofacial manual landmark identification on cone beam CT images of dry skulls

作　　者：张楠 杨仲鹏 孟琛达 陈贵[1] 许天民[1] Zhang Nan;Yang Zhongpeng;Meng Chenda;Chen Gui;Xu Tianmin(Department of Orthodontics,Peking University School and Hospital of Stamotology&National Center for Stomatology&National Clinical Research Center for Oral Diseases&National Engineering Laboratory for Digital and Material Technology of Stomatology,Beijing 100081,China)

机构地区：[1]北京大学口腔医学院·口腔医院正畸科,国家口腔医学中心,国家口腔疾病临床医学研究中心,口腔数字化医疗技术和材料国家工程实验室,北京100081

出　　处：《中华口腔正畸学杂志》2023年第2期66-71,共6页Chinese Journal of Orthodontics

基　　金：国家自然科学基金(82071172);国家重大疾病多学科合作诊疗能力建设项目(PKUSSNMP-201902)。

摘　　要：目的评估使用锥形束CT标定颅面部标志点的可靠性及准确性,以便建立三维头影测量系统。方法本研究为横断面研究,使用15个干颅骨,3名正畸医师在无标记锥形束CT影像上标记32个解剖标志点,再使用不锈钢珠标记颅面部标志点作为金标准评价锥形束CT影像上定点的准确性和可靠性。通过统计所有标志点坐标的轴向标准差及平均径向球面误差来评估手工定点的可靠性;两两计算3人分别手工定点坐标、3人手工定点平均坐标及金标准之间的平均误差距离及标准差,以评估手工定点的准确性。结果3人手工定点至金标准的误差距离为(1.599±1.081)mm;3人定点平均值至金标准的误差距离为(1.391±0.821)mm;3人手工定点分别至3人定点平均值的误差距离为(0.778±0.575)mm。平均径向球面误差最大的是下颌升支后缘点,3人手工定点至金标准误差最大的标志点为颧骨点。结论多人手工定点取均值可以提高手工定点准确性;参考系标志点和明确解剖结构的标志点如前鼻棘点、后鼻棘点等可靠性及准确性较高;上、下齿槽座点等平滑曲面上的标志点可靠性较差,需要更严格的三维定义以减小定点误差;三维定义下的颧骨点的手工定点可靠性明显高于传统定义。Objective To evaluate the accuracy and reliability of cone beam CT of craniofacial landmarks,so as to establish the 3D cephalometric system.Methods This is a cross-sectional study.Three orthodontists marked 32 landmarks on unmarked cone beam CT of 15 dry skulls,and stainless steel beads were used to mark landmarks on dry skulls as the gold standard.The reliability of manual landmarking was evaluated by calculating the axial standard deviation and mean radial spherical error(MRSE)of all landmarks.The average error distance between individual landmark detections of each orthodontist and gold standard,between individuals and average detections,and between the average detections and the gold standard were calculated to evaluate the accuracy.Results The average error between detections of individuals and the average landmark detections was(0.778±0.57)mm.It was(1.391±0.821)mm between the average detection and the gold standard.It was(1.599±1.081)mm between individuals and the gold standard.The landmark point of posterior edge of ramus had the largest MRSE and Jugal showed largest error compared to the gold standard.Conclusions Averaging multiple manual detections can elevate accuracy.The landmarks of reference system and definite anatomical structure,like anterior and posterior nasal spines,are reliable and accurate.The reliability of landmarks on smooth surfaces,like point A and point B,is poor,and more strict three-dimensional definition is needed;while Jugal with three-dimensional definition showed less error.

关 键 词：三维头影测量 手工定点 干颅骨 

分 类 号：R783.5[医药卫生—口腔医学]