基于机器学习的颌骨特征点还原法辅助跨中线颌骨缺损重建  

Landmarks restore,a machine learning algorithm-aided surgical planning for cross-midline maxillo-mandibular defect

作　　者：周子疌 朱向阳[2] 韩婧[1] 刘剑楠[1] 张陈平[1] ZHOU Zi-jie;ZHU Xiang-yang;HAN Jing;LIU Jian-nan;ZHANG Chen-ping(Department of Oromacillofacial Head and Neck Oncology,Shanghai Ninth People's Hospital,College of Stomatology,Shanghai Jiao Tong Uniersity School of Medicine,National Clinical Research Center for Oral Diseases,Shamghai Key Laboratory of Stomatology&Shanghai Research Institute of Stomatology,Shanghai 200011;Institute of Image Communication and Netuork Engineering,Sharghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学医学院附属第九人民医院·口腔医学院口腔颌面-头颈肿瘤科,国家口腔疾病临床医学研究中心,上海市口腔医学重点实验室,上海市口腔医学研究所,上海200011 [2]上海交通大学图像通信与网络工程学院,上海200240

出　　处：《中国口腔颌面外科杂志》2020年第4期323-327,共5页China Journal of Oral and Maxillofacial Surgery

基　　金：国家重点研发计划(2018YFC0807303)。

摘　　要：目的:通过测量正常汉族人群颌骨关键特征点数据,计算并分析上、下颌骨间各特征点的内在联系,为跨中线颌骨缺损的个体化重建提供参考。方法:收集111例华东地区正常成年人颌骨CT数据(Dicom格式),其中男43例、女68例,平均年龄24.3岁。应用手术规划系统(Proplan CMF 3.0)进行上、下颌骨分割,同时标记并收集16个颌骨特征点空间坐标(上颌骨9个、下颌骨7个)。借助MATLAB数学软件,对颌骨特征点坐标进行统计分析;并通过机器学习算法,研究特征点坐标间的相关性。采用SPSS 22.0软件包对男女颌骨外形数据进行t检验,分析性别差异。结果:不同性别的颌骨特征点的线性参数存在显著差异(P<0.05)。除∠b1ab2(男性136.06°,女性132.18°,P<0.05)外,角度变量间无显著性别差异。基于矢量相似度匹配模型完成机器学习算法开发,并将该算法用于1例下颌骨跨中线缺损患者的术前设计。结论:机器学习的颌骨特征点还原法可在复杂颌骨重建过程中提供精确的个体化方案,解决跨中线大范围颌骨缺损病例重建仅凭经验、无参照可依的临床难题。PURPOSE:The distance and angles among characteristic points of normal Chinese jaws were measured and analyzed by an algorithm to explore the morphological correlation between maxilla and mandible,and to guide a personalized preoperative design for cross-midline defects.METHODS:A total of 111 normal maxillary CT data(Dicom format)were collected and delivered to surgical planning software(Proplan CMF 3.0),which included 43 males and 68 females.Each case had 16 key points to mark.The data were analyzed using Mathematical software(MATLAB).The data from male and female jaw bones were analyzed using SPSS 22.0 software package to analyze the difference between genders.RESULTS:There was significant difference(P<0.05)in the size of jaw bones between genders,but no significant difference in angles except for∠b1 ab2(males 136.06°,females 132.18°,P<0.05).A machine learning algorithm was programmed based on vector matching model,and it was clinically applied to a patient with cross-midline mandibular defect.CONCLUSIONS:With the benefit of algorithm-aided preoperative design,a more convenient way can be achieved to tailor a similar maxilla or mandible for patients,especially in cases with cross-midline defects.

关 键 词：颌骨重建 解剖标志点 机器学习 术前规划 

分 类 号：R782.2[医药卫生—口腔医学]