Investigating clinically relevant methods of assessing the quality of three-dimensional surface scan data in dentistry

Title: Investigating clinically relevant methods of assessing the quality of three-dimensional surface scan data in dentistry Motivation: Having an awareness of the quality of three-dimensional (3D) scan data produced by a dental scanner can affect the clinician’s treatment plan, and potentially whether they choose to invest in a 3D scanner for clinical use. Assessing the quality of 3D scan data can therefore be of great value both to practitioners and scanner manufacturers. Statement of problem: Assessing the quality of 3D data is a challenge; assessing the quality of 3D data for clinical use is even more so. As a result, there is no standardised method of assessing, or reporting, the quality of 3D data within the field of dentistry. This research aimed to investigate methods of assessing the trueness and precision of scanners with clinical application in mind. Method and summary of publications: This research resulted in four publications. All data collection was undertaken in vitro. Digital processing, measurements and analyses were all performed digitally, in most cases using automated methods. Chapter 2 compares full arch edentulous scans produced by six intraoral scanners [IOS], with focus on identifying full arch error which may not be identified if analysed using commonly used methods such as measuring the mean distance deviation between a scan-pair. The proposed method suggests reporting the unsigned distance of the median value of the upper 1% most deviating aspect of a repeated scan. Chapter 3 investigates three methods of analysing scan data produced by two IOSs, applying the upper-bound method presented in the previous chapter in a more approachable manner, by reporting the percentage of a scan deviating beyond 0.1mm. This paper also presents the use of a virtual key point method, by which topologically similar key points can be robustly identified across differing meshes. The virtual key point method is further investigated in Chapter 4, in which the accuracy of the virtual occlusion of an IOS is investigated. By using virtual key points, the proportion of error produced by the arch scan and the proportion of error which is introduced during the occlusion stage are identified. Lastly, in Chapter 5 the key point method is used to investigate the precision of physical interocclusal records and the difference in precision between virtual articulation of dental models using un-clamped, scanned bite records, and traditional, manual articulation done by an experienced technician. Having determined the accuracy with which an IOS can record occlusion in the previous chapter, we use the same scanner to digitally record the technician’s manual articulations and, from there, identify the portion of error which was introduced by the manual articulation. Results: Chapter 2 finds that the upper-bound method may provide a clinically useful metric with which to gain an insight into the precision of full arch scans produced using different IOSs. The results indicate that three of the six scanners investigated would likely produce scans appropriate for clinical use where the full arch is required, while the latter three produced errors deviating beyond 0.3mm, hence proving to be unlikely to be appropriate for clinical use. Chapter 3 concludes that the Primescan produces significantly truer scans than the Omnicam, regardless of the method used to analyse the scan data. Furthermore the ‘standard’ analysis method might incorrectly infer that Omnicam produced clinically acceptable full arch scans. The proposed novel methods, measuring the intermolar-width and proportion beyond 0.1mm, may give a clinically relevant insight into the quality of scan data. These novel analyses reveal clinically unacceptable limitations for Omnicam. Our findings in Chapter 4 conclude that while the virtual bite records were relatively precise ((never deviating beyond 0.022mm) the error produced during the creation of the full arch scan negatively impacts the virtual occlusion. Lastly, Chapter 5 indicates that the digital articulation method using un-clamped, scanned bite records is significantly more precise than the traditional articulation method when considering precision (or lack of dislocation) along the anteroposterior axis. Discussion: All papers, with the exception of Chapter 5, investigate aspects of IOSs ability to accurately record full arch scans. All investigations presented are in vitro; as a result one may assume the quality of in vivo scan data will be worse than the findings reported herein. Chapter 3, which investigates the quality of full arch data produced by the Omnicam and Primescan scanners, highlights the importance of using the appropriate measurement method during investigation, as seen in Figure 3:2. Chapters 4 and 5 highlight the variable nature of digital methods, with the IOS virtual occlusion feature being negatively affected by error produced in the full arch scan, making the method less likely to be clinically reliable. Whereas the virtual method in Chapter 5 indicates that virtual articulation, using a high quality, clamp-less laboratory scanner is likely to reproduce articulation more precisely than traditional, physical methods. Conclusion: This thesis concludes that efforts to gain a clinically relevant insight into the quality of scan data are challenging. It finds finds that there is no one-size-fits-all when assessing 3D data in a clinically relevant manner but suggests some newer methods that go some way to addressing this. Standard surface comparison methods, borrowed from Engineering and used extensively in dental research, almost invariably produce overly optimistic results. Given that dental audiences are generally less well versed in mathematical 3D analysis, there is a real risk that clinical applicability of some digital techniques may be advocated in error. The findings also show that the quality of 3D data within digital dentistry varies widely. A paradigm shift from digital dentistry being considered as multiple methods all producing data and clinical work of similar quality, to digital dentistry being considered an umbrella term covering a spectrum of workflows, all of highly varying quality, is needed.