基于3D人脸数据库的儿童面部发育评估

Assessment of Facial Growth for Children Based on Three-Dimensional Face Database

研究发现，患有遗传性疾病的人脸常伴有面部发育异常的问题，如患有威廉姆斯综合征（WS）和唐氏综合征（DS）的人脸。为此，对儿童面部发育的研究有助于遗传性疾病的早筛查与早诊断。基于目前国内还没有儿童3D人脸数据库的现状，通过3dMDface人脸采集系统，采集人脸5 600例，对原始数据进行预处理使人脸稠密对应，从而构建3D人脸数据库。利用人体测量学标定面部各器官的关键点，建立标准测量指标，建立儿童面部发育评估系统。其中，对浙江地区分别患有DS的人脸55例和患有WS的人脸30例的19项测量指标与正常人脸比较分析。结果发现，DS人脸两眼内宽分布在第90百分位以上的样本为76．3％，鼻长分布在第10百分位以下的样本为61．8％，耳长分布在第lO百分位以下的样本为87．2％，上面深分布在第10百分位以下的样本为89％，颅面变异指数为1．630～2．117，T检验分析为T=10．130，P=．000；WS人脸鼻宽与口裂宽分布在第75百分位以上的样本分别为50％和66．7％，而两耳屏间宽分布在第25百分位以下的样本为63．3％，颅面变异指数为1．714-2．365，T检验分析为T=12．213，P=．000。这表明，该两种特殊人脸与正常人脸差异具有明显的统计学意义。因此，利用该评估系统能正确反映患有遗传性疾病特殊人脸的显著性特征，基于3D人脸数据库建立的面部发育评估系统对儿童生长发育及面部异常的评价具有临床应用价值。

It is acepted that people with inherited diseases, such as WS and DS, have features of facial abnormalities. This work was undertaken to establish 3D face database from birth to 18 years old and the facial growth norms of children of the Chinese Han population using 3D stereo-photogrammetric devices. In this study, 5600 neutral expression subjects were collected including 2803 males and 2797 females. To build 3D face database, original 3D data were preprocessed, and the input faces were required to be well aligned to a reference face. In the first step, the face region was separated from 3D data based on the theory of the largest connectivity. Secondly, the coarse alignment was achieved by approximating a 3D affine transformation between the input face and the reference face. At last, the ICP algorithm was used to align the face more precisely. The database was applied to access the children facial growth by measuring 18 craniofacial indices of 55 DS faces and 30 WS faces, which came from Zhejiang province. Results show that the distribution of 76.3% intercanthal breadth is above the 90＇h percentile, the distribution of 61.8% nose length is under the 10th percentile, the distribution of 87.2% ear length is under the 10th percentile,and the distribution of 89% upper face depth is under the 10th percentile. The CVI is 1. 630 ~2. 117 in DS. For WS, the distribution of 50% nose breadth is above the 75th percentile, the distribution of 66.7% mouth breadth is above the 75th percentile, the distribution of 63.3% bitragal breadth is under the 25th percentile and CVI is 1. 714 -2. 365. These results illustrate that the assessment system can reflect the feature of the special face, therefore the 3D face database provides a dependable 3D facial data for the research of children face. The research of facial growth assessment using the method of anthropometry has a wide range of applications, including children growth, evaluation of the abnormal face.