[Trueness of 4 three-dimensional facial scanners: an in vitro study].

Objective: To investigate the trueness of 4 three-dimensional (3D) facial scanners and to evaluate the applicability of their clinical use. Methods: An art head model was used as the scanning object, and it was scanned by Handyscan 3D scanner in an enclosed environment with a fixed light source to obtain the reference digital model. Three fixed 3D facial scanners (A: 3dMDface; B: Facego Pro; C: RDS Facescan) and a portable hand-held 3D facial scanner (D: Revopoint POP 2) were used to scan the art head model 10 times, and 10 models of each scan group were obtained. The face of the reference model was divided into 16 regions according to anatomy and muscle distributions in the Geomagic Wrap software with saved boundary curves of whole face and each region. The test models were also divided into 16 regions through the curves above after registered with the reference model through "Best fit" function. The root-mean-square error (RMS) of the complete test models and their segmented regions compared with the reference model and its corresponding regions were calculated by 3D comparison function. The smaller the RMS, the higher the accuracy. One-way ANOVA and SNK post-test were used for statistical analysis. Results: RMS of complete test models scanned by A, B, C, D scanners were (0.295±0.005), (0.216±0.053), (0.059±0.012) and (0.103±0.026) mm (F=123.81, P<0.001), respectively. There was significant difference between any two groups (P<0.05). For each facial region, the group D had the best trueness in nasal region, lip region, left orbital region and right orbital region [RMS were (0.079±0.032), (0.061±0.019), (0.058±0.021), (0.081±0.032) mm, respectively], while the group C had the best trueness in frontal region, left buccal region, right buccal region, left zygomatic region, right zygomatic region, left parotideomasseteric region, right parotideomasseteric region, left temporofacial region, right temporofacial region, mental region, left infraorbital region and right infraorbital region [RMS were (0.039±0.011), (0.034±0.007), (0.033±0.007), (0.066±0.023), (0.038±0.022), (0.070±0.030), (0.067±0.024), (0.063±0.029), (0.045±0.023), (0.063±0.006), (0.039±0.010), (0.046±0.008) mm, respectively]. Conclusions: On the basis of art head model scanning, although the overall average deviation between the scanning model and the reference models obtained by the four kinds of 3D facial scanners were small, the portable handheld 3D facial scanner (D) has better accuracy than the fixed 3D facial scanners (A, B, C) in the orbital area, nasal area, lip area and areas with rich features.目的： 通过体外模型扫描研究比较4种三维面部扫描仪的扫描正确度，评价其临床适用性。 方法： 以美术头颅模型为扫描对象，在封闭、光源亮度固定的环境中，使用便携式计量级三维扫描仪（Handyscan 3D）进行扫描，将扫描数据作为参考模型；使用3种固定式三维面部扫描仪（A：3dMDface；B：Facego Pro；C：RDS Facescan）和1种便携式手持三维面部扫描仪（D：Revopoint POP 2）分别扫描头颅模型10次，获得4个扫描组模型各10个。使用Geomagic Wrap软件，按解剖和肌肉走行将参考模型面部数据划分为16个区域，保存全脸和各区域边界曲线。分别将4个扫描组模型与参考模型通过最佳匹配功能配准后，使用已建立的边界曲线将各扫描模型面部数据划分为16个区域，通过三维比较功能计算各扫描组模型与参考模型全脸及各区域偏差的均方根误差（root-mean-square error，RMS），作为各三维面部扫描仪扫描正确度（RMS越小扫描正确度越高）。使用单因素方差分析和事后检验（SNK法）分析各组差异。 结果： 对于全脸，A、B、C、D组RMS分别为（0.295±0.005）、（0.216±0.053）、（0.059±0.012）和（0.103±0.026）mm（F=123.81，P<0.001），两两比较差异均有统计学意义（P<0.05）。对于面部各区域，鼻区、唇区、左眶区、右眶区D组扫描正确度均为4组中最高［RMS分别为（0.079±0.032）、（0.061±0.019）、（0.058±0.021）、（0.081±0.032）mm］，而额面区、左颊区、右颊区、左颧区、右颧区、左腮腺咬肌区、右腮腺咬肌区、左颞面区、右颞面区、颏区、左眶下区、右眶下区C组扫描正确度均为4组中最高［RMS分别为（0.039±0.011）、（0.034±0.007）、（0.033±0.007）、（0.066±0.023）、（0.038±0.022）、（0.070±0.030）、（0.067±0.024）、（0.063±0.029）、（0.045±0.023）、（0.063±0.006）、（0.039±0.010）、（0.046±0.008）mm］。 结论： 体外模型扫描显示，4种三维面部扫描仪的扫描模型与参考模型的全脸偏差较小；相对于固定式三维面部扫描仪，便携式手持三维面部扫描仪在眶区、鼻区、唇区等特征较复杂区域的扫描正确度较高。.