Simultaneous, radiation-free registration of the dentoalveolar position and the face by combining 3D photography with a portable scanner and impression-taking

The position of the maxillary dentulous or edentulous dentoalveolar arches in relation to the extraoral soft tissues is usually determined by using facebows and cast models that are positioned in an articulator after registration. To correlate the soft tissue and facial anatomy, auxiliary lines are marked on the models to transfer the patient’s situation as well and as realistically as possible [1]. This method, however, is susceptible to errors and may result in inaccuracies due to varying soft tissue situations, movements (e.g. grimacing), material properties in terms of shrinking and secondary deformation [2‐4]. Three-dimensional (3D) photography is already used for various indications in dentistry and cranio-maxillofacial surgery, including esthetical dental rehabilitation of incisors, as a pre-interventional visualization tool to supplement the recorded information, treatment planning and follow-up documentation in orthognathic surgery [5‐8]. This sort of mobile or stationary surface imaging is non-invasive and is becoming an additional gold standard tool for documentation and planning, especially in craniofacial surgery [9‐12]. Several mobile systems have shown to be a valid and reliable solution with a reasonable cost–benefit ratio alongside the established expensive stationary systems of the last decade due to ongoing technical developments [11, 13, 14].

In terms of surface matching combining two different capturing methods, the combination of cone-beam computed tomography (CBCT) and 3D photogrammetry or scanned dental casts has proven to be a reliable and feasible method. An overview of various investigations was provided by Mangano and colleagues [15‐17]. This results in good accuracy of the dental arch positioning and/or soft tissue illustration [18], which is necessary in pre-interventional planning of orthognathic surgery or orthodontic treatment and might facilitate the planning and simulation of a full mouth restoration. But of course, CBCT is associated with radiation and therefore should be restricted to defined indications with respect to the radiation protection law and current guidelines.

As a consequence, Bechtold et al. have described a radiation-free integration of a virtual maxillary dentoalveolar arch model into a facial scan in ten steps using a stationary photogrammetry system. This was found to have comparable precision to 3D-data derived from CBCT images alone [19]. In cases of an edentulous jaw Schweiger et al. as well as Hassan et al. presented a virtual workflow for complete dentures for which also facial scans were used. Their workflow aligns the digitalized dental arches according to the facial scan and provides valuable information to evaluate the tooth arrangements, however, without a definite intra-extraoral registration [20, 21].

The aim of this presented study was to analyse and describe a solution and workflow to register the intraoral position of the maxillary dentoalveolar arch simultaneously to the extraoral 3D photography with an intra-extraoral geometry using a portable 3D scanner. This would enable a virtual and radiation-free registration of the intraoral dental situation to the extraoral facial anatomy. The provided workflow could be used for prosthetic/orthodontic/orthognathic planning and post-interventional follow-ups and provides a recommendation for a straightforward geometry design and a step-by-step explanation.

In a first step we analysed the accuracy of the alignments between the original, virtual .stl file and the scanned .stl file of the two geometries (cross vs. sphere) applying the RMSE analysis. The sphere geometries (n = 5; mean: 0.24 mm; range: 0.23–0.28 mm) showed significantly better results than the cross geometries (n = 5; mean: 0.36 mm; range: 0.33–0.40 mm; p < 0.008), (Fig. 7 and 8a, Table 1).

Ten healthy, Caucasian participants (four females and six males) with a mean age of 29.2 years (range: 27–32 years) were included in the clinical application and transfer. From all participants a facial scan was performed with simultaneous intraoral maxillary impression (Fig. 2). All impressions and scans were adequate in quality and could be used for further analyses. The two 3D files could be aligned digitally after extraoral digitalization of the impression tray in every case (Fig. 5). Once the geometries were scanned, there was no statistically significant difference in RMSE analysis between the cross and the sphere geometries (p = 0.70, Fig. 8b, Table 2).

The consecutive exemplary alignment of a digitalized dental cast model along the gumline of the scanned impression and the positioning of the mandibular model along the occlusion points in maximal intercuspation was also possible in all cases, resulting in a complete virtual model indicating the three-dimensional position of upper and lower jaws in relation to the extraoral face (Fig. 6).

The tenfold repetition of the virtual alignment workflow showed a mean RMSE of 0.27 mm (range: 0.17–0.40 mm) with a standard deviation of 0.078 mm and a variance of 0.006 mm².

Radiation-free solutions for intra-extraoral registrations are wanted in times of CAD/CAM-assisted surgery as well as increasing awareness and interest to health and radiation safety. Further, simultaneous registration and virtual and plaster-free workflows would reduce time and increase accuracy. Accuracy of facial plaster casts varies between 0.95 and 3.55 mm according to Holberg et al. [27]. This might be due to the reported finding that the influence of facial movements is greater than the technical influence in terms of technical error [28]. Grimacing is another common reason for insufficient quality for both direct 3D acquisition and indirect impression-taking as well as model or impression scanning [29, 30]. A quiet room with monotone walls and surroundings is therefore recommended for every kind of (3D) image-taking.

In addition, facial 3D photography has reached a high level of accuracy and reproducibility, even with portable devices [11, 13, 14]. Additionally, intraoral scanners have become a standardized and promising tool and the direct data capturing in terms of scanning/digitalization of the impression achieves more accurate results than the indirect/conventional way by creating a corresponding plaster model [22]. But a whole arch scan might be susceptible for more deviation in accuracy and should be restricted to ten units without wide edentulous areas [31, 32]. The direct scanning of dental arches takes longer than a conventional impression. Further, application is restricted to adults and to patients with regular mouth opening. The scanning time and the dimensions of the intraoral scanners are still too long and big for regular use in children or even newborns for diagnostic purposes or full virtual feeding-plate planning and production [33]. Therefore, our workflow for simultaneous, radiation-free intra-extraoral registration remains dependent on conventional impression-taking.

The idea of digital facebows that combine intra- and extraoral registrations using spherical geometries have been described and patented before [34]. Our geometries, which have been designed independently of the mentioned patent, have less contact to the lips and are in our opinion more easily transferred to the clinical setting. Bechtold et al. described a ten-step workflow for simultaneous intra-extraoral registration using a stationary photogrammetry system [19]. In contrast to their technique, our modified impression tray was much smaller and easier to design than their extraoral registration geometry and we only needed six steps for virtual segmentation and alignment. In contrast, we did not perform a control analysis of the maxillary or mandibular dentoalveolar arch position in correlation to the extraoral facial anatomy with a CBCT or comparable methodology after virtual alignment, something that is a common procedure in the literature [35]. There is no ethical approval granted by the Ethical Committee of the Technical University of Munich to perform a CBCT of our enrolled healthy participants. Therefore, this presented study focused on the accuracy of the two attached and scanned geometries as well as the feasibility of our virtual workflow and showed a low variance of alignments after a tenfold workflow repetition. The reduction of information when only performing six steps instead of ten seems to have only minor or even no impact. Here, the extraoral geometry showed best results in the RMSE analysis when the spherical geometry was used. This is in concordance to good results in the navigation-assisted surgery, where the intraoperative registration devices commonly also have spherical geometries for optimized tracking in the three-dimensional space. Spherical geometries can be detected from multiple angles easily [36]. We wanted to compare the standard geometry to the cross geometry, because automated registration and positioning of the geometry is wanted in a further step in our diagnostics and treatment planning for children with cleft lip and palate. A cross-like geometry has shown best results in this automated step (data not published) and would have been the missing link for fully automated generation of CAD/CAM-assisted appliances for nasoalveolar moulding (NAM) therapy as described earlier [37, 38]. Furthermore, a cross-like geometry seems to be more suitable for the alignment due to definite edges which can be used for reference marker positioning. However, our analysis showed that the spherical geometry is detected better by the scanner used in our clinical practice due to the technical scanning algorithm – the cross was also fully scanned but the edges seemed to be radiused. Since the scanner always needs a swing, e.g. for scanning the nose completely, the advantages of the detection of a spherical geometry compared to an edged geometry are pushed into the background. Once scanned, there were no statistically significant differences in RMSE analysis between the two kinds of geometries. For this purpose, we therefore need to perform more analysis on the basis of this feasibility study to improve the missing cornerstone. Next steps will be the design of individualized impression trays with an integrated threaded basis in order to abolish the need for an additional attachment of it to further optimize the CAD procedure.

Lin et al. and Jayaratne et al. compared the accuracy of low-dose cone beam CT scan protocols with the 3dMD system and obtained an RMS error between 0.74 ± 0.24 and 1.8 ± 0.4 mm [35, 39]. The precision of other stationary 3D camera systems is reported to be good, with the mean absolute differences for the VECTRA system lying within 1.2 mm and less than 1 mm by using 3dMD [40, 41]. These reported results are more precise than a deviation of 2 mm. RMS error values larger than 2 mm are considered unreliable according to the literature [11, 35]. Our tenfold repetition of alignment and the consecutive analysis of RMSE of the superimposed models showed a mean deviation of 0.27 mm with a standard deviation of 0.078 and a variance of 0.006. For documentation and illustration for the patient, this deviation is clinically negligible. Virtual surgery planning (VSP) is reported to be feasible, reliable and accurate. But nevertheless, the difference between the virtual plan and the postoperative result still ranges between 1 and 2 mm or up to ±12.5° in mandibular reconstructions using the free fibula flap and in VSP orthognathic surgery [42‐45].

Nevertheless, studies comparing 3D photos compare only the “theoretical truth” with all the inaccuracies of the used systems [46]. Further, no technique enables a precise simulation and prediction of the postoperative result, yet. Within the reported and known limitations we therefore think that our results are clinically acceptable and relevant [47].

Simultaneous, radiation-free intra-extraoral registration is possible and we described a six-step approach to solving this interesting and promising procedure, which can be applied in many fields in modern documentation and treatment planning. Our results implied a superiority of spherical geometry for extraoral registration.