Assessment of Registration Accuracy and Duration Using AI-assisted Registration Versus Conventional Point-based Registration on CBCT Scans with Heavy Metal Artifacts

Assessment of Registration Accuracy and Duration Using AI-assisted Registration Versus Conventional Point-based Registration on CBCT Scans with Heavy Metal Artifacts | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Assessment of Registration Accuracy and Duration Using AI-assisted Registration Versus Conventional Point-based Registration on CBCT Scans with Heavy Metal Artifacts Yassin S. Alian, Zainab Hafez Abdel Rahman, Mohamed Ahmed Mohamed Tolba, and 1 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9157142/v1 This work is licensed under a CC BY 4.0 License Status: Under Review Version 1 posted 7 You are reading this latest preprint version Abstract Objectives This study assessed the accuracy and duration of AI-assisted registration versus point-based registration in CBCT scans with heavy metal artifacts, which is measured as median gap distance, and duration between AI-assisted registration and conventional point-based registration using coDiagnostiX software on CBCT scans exhibiting heavy metal artifacts. Secondary objectives include evaluating the influence of artifact-to-natural teeth ratio (subgroups: 0 < N < 50, 50 ≤ N < 100, 100 ≤ N) and arch surface area (maxilla vs. mandible) on registration performance in both methods. Materials and Methods 90 CBCT images and intraoral scans were included following specific eligibility criteria, then divided equally among the two registration groups: point-based and AI-assisted registration each group was subdivided into 3 subgroups according to the ratio of artifacts to natural teeth. The CBCT scans were segmented utilizing the AI-assistant feature via coDiagnostiX software (coDiagnostiX® 10.7, Dental Wings Inc., Montreal, Canada). Afterward, point-based registration and AI-assisted registration were performed by a single operator using coDiagnostiX software. Then, the registration accuracy was examined by measuring the median gap distance between the three-dimensional models of CBCT data and the registered intraoral scans. Also, the duration required for registration was calibrated and recorded by a stopwatch. Results The registration accuracy of the AI group was superior to the PR group, where the median gap distance in the PR group was 0.278 mm and 0.202 mm in the AI group, with no statistically significant correlation between the groups p-value > 0.05. SubGroup 1 had the highest accuracy in the PR groups, while Subgroup 3 showed the greatest accuracy in the AI groups. With regards to the influence of the surface area, in the PR groups, the upper arch in group 1 was the most accurate. On the other hand, the lower arch in group 3 had the least accuracy. Concerning the AI groups, the mandible in Subgroup 3 showed the highest accuracy, whereas the maxilla in group 2 highlighted the lowest accuracy. Conclusion AI integration in virtual implant software could enhance and ease the segmentation and registration steps for clinicians with superior accuracy and in a shorter duration compared to conventional point-based registration techniques. Clinical relevance : In oral implantology, metal artifacts from restorations compromise CBCT segmentation and registration, critical for precise digital surgical planning and guided implant placement. AI integration offers clinicians an efficient, operator-independent alternative, reducing errors in edentulous or heavily restored arches and enhancing predictability in implant workflows. These findings support adopting AI tools to improve surgical guide accuracy and clinical outcomes in challenging cases Clinical trial registration number: (NCT06273332), date of registration 5/02/2024,first posted 22/02/2024 AI registration point-based registration segmentation metal artifacts Figures Figure 1 Figure 2 Introduction In digital workflow, the precise registration of Cone Beam Computed Tomography (CBCT) and optical scans is essential to guarantee the accuracy necessary throughout each phase of computer-aided design, thereby facilitating optimal results in the field of oral implantology. Among the essential steps in digital implant planning are segmentation and registration steps. Segmentation has been performed manually, which entails a vast amount of time and effort and an experienced clinician ( 1 ) . In order to overcome these drawbacks, semi-automated segmentation was introduced, which selects particular structures by employing specific Hounsfield/grayscale values as thresholds. So, the quality of the segmented model depends mainly on the quality of the CBCT scan, which accordingly affects both image segmentation and registration, two pivotal steps of accurate surgical planning and guided surgery accuracy. Registration of the model is indispensable as it compensates for the imaging errors that may occur, hence providing accuracy to the digital surgical guide ( 2 , 3 ) One of the commonly used marker-free registration methods is point-based registration (PBR). PBR is carried out manually selecting identical matching points on the data to be registered, through which the software program fetches a possible alignment for the given points. The accuracy of this method relies mainly on operator experience and on the points selected being identical. This method has been used for a long time as a reliable method of registration. However, its precision depends on finding identical reference structures, which is declined with low-quality CBCT imaging and the experience of the operator ( 4 ) In fact, CBCT has been positioned as the preferred method for volumetric image capturing in oral implantology; however, the presence of metallic restorations substantially reduces the CBCT quality. These restorations produce metal artefacts that deform models and conceal anatomical reference points, making it difficult to register precise data. Point registration is common to be followed by surface-based automatic registration to refine the registration process, but this heavily depends on the presence of high-quality segmented models, which is heavily affected by the presence of the artefacts ( 5 ) Recently, the field of dentistry has witnessed the apparent advent of (AI). An approach to address the limitations is suggested by the development of AI-assisted registration. The utilization of AI has facilitated the achievement of a completely automated registration process through machine learning ( 6 ) . The null hypothesis is that point-based registration and AI registration exhibit the same level of accuracy and necessitate a similar duration for registration. Owing to the promising results of AI registration, the current study was set up to assess AI-assisted registration versus conventional point-based registration in cases with metal artifacts. Materials and Methods The study is a retrospective clinical study reported according to the STROBE statement ( https://www.equator-network.org/reporting-guidelines/strobe/ ). This study was reviewed and approved by the research ethical committee at Misr International University (IRB number: MIU-IRB-2324-020) and registered on clinicaltrials.gov (NCT06273332). The purpose of the study was explained to all patients, and an informed consent for using their CBCTs was signed before the conduction of the study. The trial was performed in accordance with the principles of the Declaration of Helsinki. All the study-related information was stored securely and password-protected on the principal investigator's (PI) computer Patients’ files were anonymized. Sample size calculation was performed following Piao et al. 2022 ( 7 ) . A total sample size of 16 patients was calculated under a power of 95% and a type I error of 5%, assuming a mean (standard deviation) for point-based registration (PR group) was 0.29 (0.010) mm and 0.31 (0.009) for AI registration group the sample size was increased to 45 patients. Sample size calculation was performed using G*Power 3.1.9.4 software (Germany). Eligibility criteria entailed full arch CBCT scans of the maxilla, mandible, or both with metallic restorations, alongside intraoral or desktop scans. The upper and lower teeth were parted with the aid of the positioning tool of the CBCT machine to ensure visibility of the occlusal surfaces without any distortion. Cases were excluded if data was missing or incomplete. Metal artifacts and remaining natural teeth were counted for each scan, and cases were grouped based on the ratio of artifacts to natural teeth (N) following Piao et al. 2022 ( 7 ) . The grouping resulted in three groups: Subgroup 1 (0% < N < 50%), subgroup 2 (50% ≤ N < 100%), subroup 3 (100% ≤ N). Furthermore, to assess the influence of surface area on registration accuracy, each case was divided into upper and lower arches. An independent researcher performed the segmentation step by AI using coDiagnostiX software (coDiagnostiX® 10.7, Dental Wings Inc, Montreal, Canada). After AI segmentation, all segmented models underwent point-based and AI-based registration by an independent researcher. In the point-based registration group, five reference points were used following Park et al. 2020 ( 8 ) : the midpoint of the incisal edges of the central permanent incisors, the bilateral cusps of the permanent canines, and the bilateral mesial buccal cusps of the first permanent molars. Then, all registered models were securely stored for subsequent analysis. After registration, three points were determined to measure the distance between the intraoral scan and the 3D model of the CBCT scan (gap distance). These points were: one point at the mesial point angle of the nearest tooth to the midline, and the other two points were at the cusp tip of the last standing teeth bilaterally. The distance was measured by Geomagic Design X software 2020 (3D systems, North Carolina, USA) in millimeters. The duration was divided according to the tasks needed for registration such as scan importing, AI segmentation, PR and AI registration, scan exporting, renaming and saving. The time taken for both registrations was calculated by a stopwatch and recorded Figure (1). Statistical Analysis Numerical data from the experiment were collected and checked for normality using tests of normality (Kolmogorov-Smirnov test and the Shapiro-Wilk tests). Descriptive statistics were calculated for the registration accuracy and duration of registration for both groups. The correlation between the registration accuracy and the number of artifacts was calculated by Spearman’s correlation. Similarly, the correlation between the surface area of the maxilla, mandible and registration accuracy was calculated by Spearman’s correlation. The significance level was set at a p-value ≤ 0.05. IBM SPSS for Windows was used for statistical analysis (SPSS 24 Inc.; Chicago, IL, USA). Results A total of 90 CBCTs were investigated; 45 patients were enrolled in each group for registration accuracy assessment. Normality tests identified that the data are considered non-parametric data. The data was grouped according to the ratio of teeth with artefacts to the remaining natural teeth (N) (Fig. 2 ). Registration Accuracy and Duration Analysis Generally, the AI group showed superior registration accuracy than the PR group in less time, as illustrated in table (1). Furthermore, group 3 in the PR groups highlighted the lowest registration accuracy among all groups. While group 3 in the AI groups had the highest registration accuracy table (2). Table (1) Descriptive statistics for registration accuracy and registration duration of both groups. Descriptive statistics PR Group AI Group number of cases (patients) 45 (72.5%) 45 (72.5%) minimum 0.086 0.044 maximum 1.305 0.723 Mean gap distance (MGD) 0.384 0.2477 SD 0.2992 0.1633 SE 0.0446 0.0243 Median gap distance 0.278 0.202 IQR 0.2115 0.1775 95% CI {0.2941–0.4739} {0.1987–0.2968} Registration Duration Descriptive statistics PR Group AI Group number of cases 32 (51.6%) 38 (61.3%) Minimum 2 min 2 sec 2 min 40 sec Maximum 15 min 53 sec 9 min 50 sec mean 6 min 24 sec 5 min 38 sec SD 3 min 58 sec 1 min 91 sec median 5 min 02 sec 6 min 21 sec IQR 5 min 19 sec 3 min 39 sec 95% CI {4 min 95 sec − 7 min 53 sec} {4 min 73 sec- 6 min 04 sec} Influence of Surface Area on the Registration Accuracy In the PR groups, opposing to other groups, group 3 showed greater registration accuracy in the lower arches than in the upper arches. Furthermore, the greatest accuracy was in the maxilla of group 1 (MGD = 0.164), while the least accuracy was in the mandible of group 3 (MGD = 0.401). However, in the AI groups, only group 1 displayed higher accuracy in the upper arch than in the lower one. Also, in the AI groups, the mandible of group 3 demonstrated the greatest accuracy (MGD = 0.147), whereas the maxilla of group 2 exhibited the least accuracy (MGD = 0.235). Regarding the correlation between upper and lower arches, groups 1 and 2 in the PR groups showed a weak negative correlation. However, group 1 in the AI groups demonstrated a strong negative correlation (⍴ = -0.771), while group 2 showed a weak positive correlation (⍴ = 0.202) with no statistical significance correlation in any group (p-value > 0.05). No correlation coefficient was conducted for the maxilla and the mandible of group 3 using both techniques due to the limited number of cases Table (2). Table (2) Median gap distance in all groups, each separate arch using both the registration technique and the correlation coefficient of them. PR Groups Median Gap Distance Correlation Coefficient (p-value) Subgroup 1 (0% < N < 50%) 0.216 PR maxilla 0.164 -0.371 (0.468) PR mandible 0.236 Subgroup 2 (50%<N < 100%) 0.261 PR maxilla 0.260 -0.228 (0.412) PR mandible 0.261 Subgroup 3 (100% ≤N) 0.387 PR maxilla 0.401 - PR mandible 0.310 AI groups Median Gap Distance Correlation Coefficient (p-value) Subgroup 1 (0% < N < 50%) 0.172 AI maxilla 0.148 -0.771 (0.07) AI mandible 0.182 Subgroup 2 (50%<N < 100%) 0.205 AI maxilla 0.235 0.202 (0.450) AI mandible 0.176 Subgroup 3 (100% ≤N) 0.148 AI maxilla 0.183 - AI mandible 0.147 Discussion The digitization workflow of dental implant planning has significantly improved the accuracy and predictability of implant positioning, thereby yielding superior clinical outcomes. However, the segmentation and registration accuracy can be reduced by the presence of metal restoration which results in metal artifacts in the CBCT scan ( 9 ) . Owing to the advancements and integration of AI in various dental software, our trial aimed to assess the registration accuracy of the AI assistant feature against point-based registration in scans with metal artifacts. The segmentation step was done by the AI assistant feature inherited in the software for both groups to gain consistent segmentation results since it does not depend mainly on the operator ( 9 ) . With regard to the point-based registration procedure, variability in the number and distribution of the available dentition may exert an impact. That’s why the grouping of the included cases was done based on a ratio between the remaining teeth and the number of artifacts. However, a preceding investigation employed a comparable in vitro framework established that point-based registration was not appreciably influenced by the positioning of the edentulous area ( 10 , 11 ) In the current study, point-based registration was carried out by five reference points as previously mentioned to allow as much precision as possible. On the contrary, three reference points were considered enough for point-based registration as mentioned in a previous study ( 12 ) . In scenarios involving the utilization of three points, it is generally advised to position the registration points at maximal distances from one another ( 13 ) . Furthermore, the selection of the quantity and configuration of registration reference points is carried out with respect to the distribution and quantity of existing metallic restorations as well as the quality of (CBCT) imaging ( 14 ) . Concerning the registration accuracy measurement, three points were selected to calibrate the gap distance between the AI-segmented model and the registered scan. The rationale behind the selection and positioning of these points was to get an overall evaluation of the gap distances between both models along the whole arch. In contradiction to our method, Gaber et al. assessed the registration accuracy by two metrics: median surface deviation, which denotes the mean separation, measured in millimeters (mm), between points located on a surface and their respective points on the aligned surface mesh analysis. root mean square: This metric denotes the discrepancies between two aligned surfaces, quantified in millimeters (mm) ( 9 ) . In fact, we can accept the null hypothesis since the current results demonstrated that the two groups had no statistically significant difference in both registration accuracy and duration (p-value > 0.05). Moreover, the mean gap distance (MGD) in the PR group (0.384 mm) surpassed that of the AI group (0.247 mm). Thus, the AI group had a higher registration accuracy than the PR group. The accurate efficacy of artificial intelligence may be ascribed to meticulous AI-driven tooth segmentation, which constitutes an essential phase in the registration procedure. Furthermore, automated registration is devoid of any manual involvement, such as pinpointing landmarks or surfaces. This degree of automation is realized through convolutional neural networks (CNN) that have been honed through supervised learning conducted by individuals possessing a profound level of expertise ( 15 , 16 ) . The weak positive result for the registration of the maxillary arch vs the mandibular arches in the AI group, may be accounted for by the larger surface are of the maxillary teeth compared to the mandibular teeth. Furthermore, the number of metal artifacts and remaining natural teeth can directly affect the registration accuracy. Thus, this variable was evaluated by grouping the included cases according to the ratio between the number of artifacts and the number of remaining teeth (N) ( 7 ) . As highlighted in Table (3), by increasing (N) the mean gap distance increased regardless of the registration technique used, which means that the registration accuracy is reduced. This may return to the presence of metallic components that generate streak artifacts within the CBCT scan, and in cases of significant metal restorations, the identification of dependable reference points for registration emerges as a formidable challenge ( 17 ) . The weak positive result for the registration of the maxillary arch vs the mandibular arches in the AI group, may be accounted for by the larger surface area of the maxillary teeth compared to the mandibular teeth. Further, it is assumed that the difference in surface area between the upper and lower arch is also suggested to affect the registration accuracy. Therefore, we calculated the mean gap distance for each arch in each group in both registration techniques as illustrated in table (4). Only in subgroup 1 was the MGD higher in the mandible than the maxilla using both techniques which align with the aforementioned assumption. However, subgroups 2 and 3 demonstrated superior MGD in the upper arches than in the lower arches. A possible explanation may pertain to the ratio between the number of artifacts and the number of remaining natural teeth where subgroups 2 and 3 had more metal artifacts compared to subgroup 1 ( 18 ) . Besides registration accuracy, the duration of registration is also a relevant factor that should be considered. In accordance with El Garba et al. and Piao et al. our results showed that the time difference between both groups was 2 min and 20 sec. (140 sec.) with less duration for the Ai group. To avoid the experience of the operator effect only one operator did the registration procedure by both techniques. Due to the significant difference in the speed of the internet between different countries, the time of uploading and downloading the file from the AI server 30–40 sec. was deducted from the registration duration. In addition, the features of the computer used affect the speed of the software used. That’s why we used a high-performance computer. Also, using the point-based registration technique in cases of a high number of artifacts may have a direct relation to the increased registration duration as the operator spent more time choosing the registration point accurately ( 7 , 9 ) . One of the strength points of this paper resides in the evaluated dataset, which encompasses scans with diverse dental configurations, thereby facilitating an assessment of the robustness of the artificial intelligence methodology across a multitude of clinical scenarios. Conversely, a notable limitation pertains to the necessity for a degree of caution when generalizing findings to other implant planning software that utilizes a different registration algorithm. Conclusion In conclusion, AI registration may serve as an efficacious alternative to conventional registration techniques. The current findings indicate that the AI assistant feature unveiled surpassing registration accuracy compared to point-based registration. Moreover, PR accuracy is more influenced by the number of artifacts present than the AI. In addition, the AI registration accuracy did not enhance with a larger surface area model. The duration of registration was also shorter with the AI group. Declarations Acknowledgement: NA Authors’ contributions: The authors confirm their contribution to the paper as follows: contributed to the study conception, design, and practical work; Yassin S. Alian & Nehal I. Shobair; data collection was performed by Nehal I. Shobair & Mohamed A. Tolba; Zainab Hafez and Nehal I. Shobair & Mohamed A. Tolba performed manuscript preparation. All authors reviewed the results and approved the final version of the manuscript. Funding: Self-funding. Declarations: Ethical approval and consent to participate : This study was reviewed and approved by the research ethical committee at Misr International University (IRB number: MIU-IRB-2324-020) and registered on clinicaltrials.gov (NCT06273332). The trial was performed in accordance with the principles of the Declaration of Helsinki. All the information needed about the study was given to the patients and informed consent term was obtained from every individual participant included in the study. Conflict of interest: The authors declare that they have no conflict of interest and do not have any financial interest in the companies whose materials are included in this article. Consent for publication : not applicable Availability of data and Materials : Data available on request from the authors. References Leite AF, Vasconcelos KD, Willems H, et al. Radiomics and machine learning in oral healthcare. Proteom Clin Appl. 2020;14(3):e1900040. https://doi.org/10.1002/prca . Singh GD, Singh M. Virtual surgical planning: modeling from the present to the future. J Clin Med. 2021;10(23):5655. https://doi.org/10.3390/jcm10 235655 . Shujaat S, Riaz M, Jacobs R. Synergy between artificial intelligence and precision medicine for computer-assisted oral and maxillofacial surgical planning. Clin Oral Invest. 2023;27(3):897–906. Alhossaini SJ, Neena AF, Issa NO, Abouelkheir HM, Gaweesh YY. Accuracy of markerless registration methods of DICOM and STL files used for computerized surgical guides in mandibles with metal restorations: An in vitro study. The Journal of Prosthetic Dentistry; 2022. Flügge T, Derksen W, Te Poel J, Hassan B, Nelson K, Wismeijer D. Registration of cone beam computed tomography data and intraoral surface scans–A prerequisite for guided implant surgery with CAD/CAM drilling guides. Clin Oral Implants Res. 2017;28(9):1113–8. Chung M, Lee J, Song W et al. (2020) Automatic registration between dental cone-beam CT and scanned surface via deep pose regression neural networks and clustered similarities. IEEE Trans Med Imaging 39:3900–3909. https://doi.org/10.1109/TMI . 2020.30075 20. Piao XY, Park JM, Kim H, Kim Y, Shim JS. Evaluation of different registration methods and dental restorations on the registration duration and accuracy of cone beam computed tomography data and intraoral scans: a retrospective clinical study. Clin Oral Investig Sep. 2022;26(9):5763–71. 10.1007/s00784-022-04533-7 . Park JH, Hwang CJ, Choi YJ, Houschyar KS, Yu JH, Bae SY, Cha JY. Registration of digital dental models and cone-beam computed tomography images using 3-dimensional planning software: Comparison of the accuracy according to scanning methods and software. Am J Orthod Dentofac Orthop Jun. 2020;157(6):843–51. 10.1016/j.ajodo.2019.12.013 . Elgarba BM, Fontenele RC, Ali S, Swaity A, Meeus J, Shujaat S, Jacobs R. Validation of a novel AI-based automated multimodal image registration of CBCT and intraoral scan aiding presurgical implant planning. Clin Oral Implants Res. 2024;35:1506–17. https://doi.org/10.1111/clr.14338 . Jamjoom FZ, Kim DG, McGlumphy EA, Lee DJ, Yilmaz B. Positional accuracy of a prosthetic treatment plan incorporated into a cone beam computed tomography scan using surface scan registration. J Prosthet Dent Sep. 2018;120(3):367–74. 10.1016/j.prosdent.2017.11.019 . Kim BC, Lee CE, Park W, Kang SH, Zhengguo P, Yi CK, Lee SH. Integration accuracy of digital dental models and 3-dimensional computerized tomography images by sequential point‐ and surface‐based markerless registration. Oral Surgery, Oral Medicine, Oral Pathology. Oral Radiol Endodontology. 2010;110:370–8. https://doi.org/10.1016/j.tripleo.2010.03.036 . Jamjoom FZ, Yilmaz B, Johnston WM. Impact of number of registration points on the positional accuracy of a prosthetic treatment plan incorporated into a cone-beam computed tomography scan by surface scan registration: An in vitro study. Clin Oral Impl Res. 2019;30:826–32. https://doi.org/10.1111/clr.13490 . Choi YS, Kim MK, Lee JW, Kang SH. Impact of the number of registration points for replacement of three-dimensional computed tomography images in dental areas using three‐dimensional light‐scanned images of dental models. Oral Radiol. 2014;30:32–7. https://doi.org/10.1007/s11282-013-0136-2 . Al-Jarsha MY, Almezyad O, AlOtaibi N, Naudi KB, Robertson DP, Ayoub AF. The Accuracy of Intraoral Registration for Dynamic Surgical Navigation in the Edentulous Maxilla. Int J Oral Maxillofac Implants Jun. 2024;21(3):21–46. 10.11607/jomi.10531 . Elgarba BM, Van Aelst S, Swaity A, Morgan N, Shujaat S, Jacobs R. Deep learning-based segmentation of dental implants on cone-beam computed tomography images: A validation study. J Dent Oct. 2023;137:104639. 10.1016/j.jdent.2023.104639 . Shaheen E, Leite A, Alqahtani KA, Smolders A, Van Gerven A, Willems H, Jacobs R. A novel deep learning system for multi-class tooth segmentation and classification on cone beam computed tomography. A validation study. J Dent Dec. 2021;115:103865. 10.1016/j.jdent.2021.103865 . Alhossaini SJ, Neena AF, Issa NO, Abouelkheir HM, Gaweesh YY. Accuracy of markerless registration methods of DICOM and STL files used for computerized surgical guides in mandibles with metal restorations: An in vitro study. J Prosthet Dent Nov. 2024;132(5):986–93. 10.1016/j.prosdent.2022.09.017 . Noh H, Nabha W, Cho JH, Hwang HS. Registration accuracy in the integration of laser-scanned dental images into maxillofacial cone-beam computed tomography images. Am J Orthod Dentofac Orthop Oct. 2011;140(4):585–91. 10.1016/j.ajodo.2011.04.018 . Additional Declarations No competing interests reported. Cite Share Download PDF Status: Under Review Version 1 posted Reviewers agreed at journal 04 May, 2026 Reviewers agreed at journal 02 May, 2026 Reviewers invited by journal 30 Apr, 2026 Editor invited by journal 20 Apr, 2026 Editor assigned by journal 30 Mar, 2026 Submission checks completed at journal 27 Mar, 2026 First submitted to journal 27 Mar, 2026 You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-9157142","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":634023975,"identity":"cd19282e-e206-4e24-8702-5f22b3e62e5c","order_by":0,"name":"Yassin S. Alian","email":"","orcid":"","institution":"Ain Shams University","correspondingAuthor":false,"prefix":"","firstName":"Yassin","middleName":"S.","lastName":"Alian","suffix":""},{"id":634023976,"identity":"5db6b532-10c4-4df6-a21d-be803992883c","order_by":1,"name":"Zainab Hafez Abdel Rahman","email":"data:image/png;base64,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","orcid":"","institution":"Misr International University","correspondingAuthor":true,"prefix":"","firstName":"Zainab","middleName":"Hafez Abdel","lastName":"Rahman","suffix":""},{"id":634023977,"identity":"db62e89d-f2af-4d66-b395-3117ad1b2d64","order_by":2,"name":"Mohamed Ahmed Mohamed Tolba","email":"","orcid":"","institution":"Misr International University","correspondingAuthor":false,"prefix":"","firstName":"Mohamed","middleName":"Ahmed Mohamed","lastName":"Tolba","suffix":""},{"id":634023978,"identity":"6cd37de1-8f30-43b2-8336-4ced2dbd66a7","order_by":3,"name":"Nehal I. Shobair","email":"","orcid":"","institution":"Misr International University","correspondingAuthor":false,"prefix":"","firstName":"Nehal","middleName":"I.","lastName":"Shobair","suffix":""}],"badges":[],"createdAt":"2026-03-18 09:09:28","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9157142/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9157142/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":108960171,"identity":"31409f48-a758-4e41-8955-1491a5467791","added_by":"auto","created_at":"2026-05-11 08:38:01","extension":"jpeg","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":33123,"visible":true,"origin":"","legend":"\u003cp\u003eFlowchart of the trial’s steps\u003c/p\u003e","description":"","filename":"floatimage1.jpeg","url":"https://assets-eu.researchsquare.com/files/rs-9157142/v1/b8f7d6f19897b4b34e669669.jpeg"},{"id":108960172,"identity":"3bf4d9f2-9308-4c7c-a9f9-3361eea75f55","added_by":"auto","created_at":"2026-05-11 08:38:01","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":7786,"visible":true,"origin":"","legend":"\u003cp\u003eArches distribution within each group.\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-9157142/v1/270b8aa76137dc0ed0b78290.png"},{"id":108960174,"identity":"4b2bf389-aeb0-4c68-8459-8b3128999ec8","added_by":"auto","created_at":"2026-05-11 08:38:06","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":303380,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9157142/v1/6af7c75c-eb0f-4be2-8a9e-051b149a98c7.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Assessment of Registration Accuracy and Duration Using AI-assisted Registration Versus Conventional Point-based Registration on CBCT Scans with Heavy Metal Artifacts","fulltext":[{"header":"Introduction","content":"\u003cp\u003eIn digital workflow, the precise registration of Cone Beam Computed Tomography (CBCT) and optical scans is essential to guarantee the accuracy necessary throughout each phase of computer-aided design, thereby facilitating optimal results in the field of oral implantology. Among the essential steps in digital implant planning are segmentation and registration steps. Segmentation has been performed manually, which entails a vast amount of time and effort and an experienced clinician \u003csup\u003e(\u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e)\u003c/sup\u003e. In order to overcome these drawbacks, semi-automated segmentation was introduced, which selects particular structures by employing specific Hounsfield/grayscale values as thresholds. So, the quality of the segmented model depends mainly on the quality of the CBCT scan, which accordingly affects both image segmentation and registration, two pivotal steps of accurate surgical planning and guided surgery accuracy. Registration of the model is indispensable as it compensates for the imaging errors that may occur, hence providing accuracy to the digital surgical guide \u003csup\u003e(\u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2\u003c/span\u003e, \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e)\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eOne of the commonly used marker-free registration methods is point-based registration (PBR). PBR is carried out manually selecting identical matching points on the data to be registered, through which the software program fetches a possible alignment for the given points. The accuracy of this method relies mainly on operator experience and on the points selected being identical. This method has been used for a long time as a reliable method of registration. However, its precision depends on finding identical reference structures, which is declined with low-quality CBCT imaging and the experience of the operator \u003csup\u003e(\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e)\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eIn fact, CBCT has been positioned as the preferred method for volumetric image capturing in oral implantology; however, the presence of metallic restorations substantially reduces the CBCT quality. These restorations produce metal artefacts that deform models and conceal anatomical reference points, making it difficult to register precise data. Point registration is common to be followed by surface-based automatic registration to refine the registration process, but this heavily depends on the presence of high-quality segmented models, which is heavily affected by the presence of the artefacts \u003csup\u003e(\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e)\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eRecently, the field of dentistry has witnessed the apparent advent of (AI). An approach to address the limitations is suggested by the development of AI-assisted registration. The utilization of AI has facilitated the achievement of a completely automated registration process through machine learning \u003csup\u003e(\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e)\u003c/sup\u003e. The null hypothesis is that point-based registration and AI registration exhibit the same level of accuracy and necessitate a similar duration for registration. Owing to the promising results of AI registration, the current study was set up to assess AI-assisted registration versus conventional point-based registration in cases with metal artifacts.\u003c/p\u003e"},{"header":"Materials and Methods","content":"\u003cp\u003eThe study is a retrospective clinical study reported according to the STROBE statement (\u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://www.equator-network.org/reporting-guidelines/strobe/\u003c/span\u003e\u003cspan address=\"https://www.equator-network.org/reporting-guidelines/strobe/\" targettype=\"URL\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003cspan type=\"Underline\" class=\"Underline\" name=\"Emphasis\"\u003e).\u003c/span\u003e This study was reviewed and approved by the research ethical committee at Misr International University (IRB number: MIU-IRB-2324-020) and registered on clinicaltrials.gov (NCT06273332). The purpose of the study was explained to all patients, and an informed consent for using their CBCTs was signed before the conduction of the study. The trial was performed in accordance with the principles of the Declaration of Helsinki. All the study-related information was stored securely and password-protected on the principal investigator's (PI) computer Patients\u0026rsquo; files were anonymized.\u003c/p\u003e \u003cp\u003eSample size calculation was performed following \u003cb\u003ePiao et al. 2022\u003c/b\u003e \u003csup\u003e(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/sup\u003e. A total sample size of 16 patients was calculated under a power of 95% and a type I error of 5%, assuming a mean (standard deviation) for point-based registration (PR group) was 0.29 (0.010) mm and 0.31 (0.009) for AI registration group the sample size was increased to 45 patients. Sample size calculation was performed using G*Power 3.1.9.4 software (Germany). Eligibility criteria entailed full arch CBCT scans of the maxilla, mandible, or both with metallic restorations, alongside intraoral or desktop scans. The upper and lower teeth were parted with the aid of the positioning tool of the CBCT machine to ensure visibility of the occlusal surfaces without any distortion. Cases were excluded if data was missing or incomplete.\u003c/p\u003e \u003cp\u003eMetal artifacts and remaining natural teeth were counted for each scan, and cases were grouped based on the ratio of artifacts to natural teeth (N) following \u003cb\u003ePiao et al. 2022\u003c/b\u003e \u003csup\u003e(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/sup\u003e. The grouping resulted in three groups: Subgroup 1 (0% \u0026lt; N\u0026thinsp;\u0026lt;\u0026thinsp;50%), subgroup 2 (50% \u0026le; N\u0026thinsp;\u0026lt;\u0026thinsp;100%), subroup 3 (100% \u0026le; N). Furthermore, to assess the influence of surface area on registration accuracy, each case was divided into upper and lower arches. An independent researcher performed the segmentation step by AI using coDiagnostiX software (coDiagnostiX\u0026reg; 10.7, Dental Wings Inc, Montreal, Canada). After AI segmentation, all segmented models underwent point-based and AI-based registration by an independent researcher. In the point-based registration group, five reference points were used following \u003cb\u003ePark et al. 2020\u003c/b\u003e \u003csup\u003e(\u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e8\u003c/span\u003e)\u003c/sup\u003e: the midpoint of the incisal edges of the central permanent incisors, the bilateral cusps of the permanent canines, and the bilateral mesial buccal cusps of the first permanent molars. Then, all registered models were securely stored for subsequent analysis.\u003c/p\u003e \u003cp\u003eAfter registration, three points were determined to measure the distance between the intraoral scan and the 3D model of the CBCT scan (gap distance). These points were: one point at the mesial point angle of the nearest tooth to the midline, and the other two points were at the cusp tip of the last standing teeth bilaterally. The distance was measured by Geomagic Design X software 2020 (3D systems, North Carolina, USA) in millimeters. The duration was divided according to the tasks needed for registration such as scan importing, AI segmentation, PR and AI registration, scan exporting, renaming and saving. The time taken for both registrations was calculated by a stopwatch and recorded Figure (1).\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003eStatistical Analysis\u003c/h2\u003e \u003cp\u003eNumerical data from the experiment were collected and checked for normality using tests of normality (Kolmogorov-Smirnov test and the Shapiro-Wilk tests). Descriptive statistics were calculated for the registration accuracy and duration of registration for both groups. The correlation between the registration accuracy and the number of artifacts was calculated by Spearman\u0026rsquo;s correlation. Similarly, the correlation between the surface area of the maxilla, mandible and registration accuracy was calculated by Spearman\u0026rsquo;s correlation. The significance level was set at a p-value\u0026thinsp;\u0026le;\u0026thinsp;0.05. IBM SPSS for Windows was used for statistical analysis (SPSS 24 Inc.; Chicago, IL, USA).\u003c/p\u003e \u003c/div\u003e"},{"header":"Results","content":"\u003cp\u003eA total of 90 CBCTs were investigated; 45 patients were enrolled in each group for registration accuracy assessment. Normality tests identified that the data are considered non-parametric data. The data was grouped according to the ratio of teeth with artefacts to the remaining natural teeth (N) (Fig. \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e).\u003c/p\u003e\n\u003ch3\u003eRegistration Accuracy and Duration Analysis\u003c/h3\u003e\n\u003cp\u003eGenerally, the AI group showed superior registration accuracy than the PR group in less time, as illustrated in table (1). Furthermore, group 3 in the PR groups highlighted the lowest registration accuracy among all groups. While group 3 in the AI groups had the highest registration accuracy table (2).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;(1)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eDescriptive statistics for registration accuracy and registration duration of both groups.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Taba\" border=\"1\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eDescriptive statistics\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003ePR Group\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eAI Group\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003enumber of cases (patients)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45 (72.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e45 (72.5%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eminimum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.086\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.044\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emaximum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.305\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.723\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMean gap distance (MGD)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.384\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.2477\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.2992\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.1633\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSE\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0446\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.0243\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMedian gap distance\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.278\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.202\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIQR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.2115\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.1775\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95% CI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e{0.2941\u0026ndash;0.4739}\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e{0.1987\u0026ndash;0.2968}\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eRegistration Duration\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eDescriptive statistics\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003ePR Group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAI Group\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003enumber of cases\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e32 (51.6%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e38 (61.3%)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMinimum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 min 2 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2 min 40 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eMaximum\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15 min 53 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e9 min 50 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emean\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 min 24 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 min 38 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eSD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 min 58 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1 min 91 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003emedian\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 min 02 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6 min 21 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eIQR\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e5 min 19 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e3 min 39 sec\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e95% CI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e{4 min 95 sec\u0026thinsp;\u0026minus;\u0026thinsp;7 min 53 sec}\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e{4 min 73 sec- 6 min 04 sec}\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003ch3\u003eInfluence of Surface Area on the Registration Accuracy\u003c/h3\u003e\n\u003cp\u003eIn the PR groups, opposing to other groups, group 3 showed greater registration accuracy in the lower arches than in the upper arches. Furthermore, the greatest accuracy was in the maxilla of group 1 (MGD\u0026thinsp;=\u0026thinsp;0.164), while the least accuracy was in the mandible of group 3 (MGD\u0026thinsp;=\u0026thinsp;0.401). However, in the AI groups, only group 1 displayed higher accuracy in the upper arch than in the lower one. Also, in the AI groups, the mandible of group 3 demonstrated the greatest accuracy (MGD\u0026thinsp;=\u0026thinsp;0.147), whereas the maxilla of group 2 exhibited the least accuracy (MGD\u0026thinsp;=\u0026thinsp;0.235). Regarding the correlation between upper and lower arches, groups 1 and 2 in the PR groups showed a weak negative correlation. However, group 1 in the AI groups demonstrated a strong negative correlation (⍴ = -0.771), while group 2 showed a weak positive correlation (⍴ = 0.202) with no statistical significance correlation in any group (p-value\u0026thinsp;\u0026gt;\u0026thinsp;0.05). No correlation coefficient was conducted for the maxilla and the mandible of group 3 using both techniques due to the limited number of cases Table\u0026nbsp;(2).\u003c/p\u003e\n\u003cdiv class=\"gridtable\"\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eTable\u0026nbsp;(2)\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eMedian gap distance in all groups, each separate arch using both the registration technique and the correlation coefficient of them.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv class=\"colspec\" align=\"left\"\u003e\u0026nbsp;\u003c/div\u003e\n \u003ctable id=\"Tabb\" border=\"1\"\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth colspan=\"5\" align=\"left\"\u003e\n \u003cp\u003ePR Groups\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003cth colspan=\"4\" align=\"left\"\u003e\n \u003cp\u003eMedian Gap Distance\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eCorrelation Coefficient (p-value)\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubgroup 1\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(0%\u0026thinsp;\u0026lt;\u0026thinsp;N\u0026thinsp;\u0026lt;\u0026thinsp;50%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.216\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePR maxilla\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.164\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e-0.371 (0.468)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePR mandible\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.236\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubgroup 2\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(50%\u0026lt;N\u0026thinsp;\u0026lt;\u0026thinsp;100%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.261\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePR maxilla\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.260\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e-0.228 (0.412)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePR mandible\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.261\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubgroup 3\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(100% \u0026le;N)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.387\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePR maxilla\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.401\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003ePR mandible\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.310\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"5\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eAI groups\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"4\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eMedian Gap Distance\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eCorrelation Coefficient (p-value)\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubgroup 1\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(0%\u0026thinsp;\u0026lt;\u0026thinsp;N\u0026thinsp;\u0026lt;\u0026thinsp;50%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.172\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAI maxilla\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e-0.771 (0.07)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAI mandible\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.182\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubgroup 2\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(50%\u0026lt;N\u0026thinsp;\u0026lt;\u0026thinsp;100%)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.205\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAI maxilla\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.235\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.202 (0.450)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAI mandible\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.176\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e\u003cstrong\u003eSubgroup 3\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003e(100% \u0026le;N)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e0.148\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAI maxilla\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.183\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" align=\"left\"\u003e\n \u003cp\u003e-\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e\u003cem\u003eAI mandible\u003c/em\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e0.147\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n\u003c/div\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e"},{"header":"Discussion","content":"\u003cp\u003eThe digitization workflow of dental implant planning has significantly improved the accuracy and predictability of implant positioning, thereby yielding superior clinical outcomes. However, the segmentation and registration accuracy can be reduced by the presence of metal restoration which results in metal artifacts in the CBCT scan \u003csup\u003e(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/sup\u003e. Owing to the advancements and integration of AI in various dental software, our trial aimed to assess the registration accuracy of the AI assistant feature against point-based registration in scans with metal artifacts.\u003c/p\u003e \u003cp\u003eThe segmentation step was done by the AI assistant feature inherited in the software for both groups to gain consistent segmentation results since it does not depend mainly on the operator \u003csup\u003e(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eWith regard to the point-based registration procedure, variability in the number and distribution of the available dentition may exert an impact. That\u0026rsquo;s why the grouping of the included cases was done based on a ratio between the remaining teeth and the number of artifacts. However, a preceding investigation employed a comparable in vitro framework established that point-based registration was not appreciably influenced by the positioning of the edentulous area \u003csup\u003e(\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e)\u003c/sup\u003e\u003c/p\u003e \u003cp\u003eIn the current study, point-based registration was carried out by five reference points as previously mentioned to allow as much precision as possible. On the contrary, three reference points were considered enough for point-based registration as mentioned in a previous study \u003csup\u003e(\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e)\u003c/sup\u003e. In scenarios involving the utilization of three points, it is generally advised to position the registration points at maximal distances from one another \u003csup\u003e(\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e)\u003c/sup\u003e. Furthermore, the selection of the quantity and configuration of registration reference points is carried out with respect to the distribution and quantity of existing metallic restorations as well as the quality of (CBCT) imaging \u003csup\u003e(\u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eConcerning the registration accuracy measurement, three points were selected to calibrate the gap distance between the AI-segmented model and the registered scan. The rationale behind the selection and positioning of these points was to get an overall evaluation of the gap distances between both models along the whole arch. In contradiction to our method, Gaber et al. assessed the registration accuracy by two metrics: median surface deviation, which denotes the mean separation, measured in millimeters (mm), between points located on a surface and their respective points on the aligned surface mesh analysis. root mean square: This metric denotes the discrepancies between two aligned surfaces, quantified in millimeters (mm) \u003csup\u003e(\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eIn fact, we can accept the null hypothesis since the current results demonstrated that the two groups had no statistically significant difference in both registration accuracy and duration (p-value\u0026thinsp;\u0026gt;\u0026thinsp;0.05). Moreover, the mean gap distance (MGD) in the PR group (0.384 mm) surpassed that of the AI group (0.247 mm). Thus, the AI group had a higher registration accuracy than the PR group. The accurate efficacy of artificial intelligence may be ascribed to meticulous AI-driven tooth segmentation, which constitutes an essential phase in the registration procedure. Furthermore, automated registration is devoid of any manual involvement, such as pinpointing landmarks or surfaces. This degree of automation is realized through convolutional neural networks (CNN) that have been honed through supervised learning conducted by individuals possessing a profound level of expertise \u003csup\u003e(\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe weak positive result for the registration of the maxillary arch vs the mandibular arches in the AI group, may be accounted for by the larger surface are of the maxillary teeth compared to the mandibular teeth. Furthermore, the number of metal artifacts and remaining natural teeth can directly affect the registration accuracy. Thus, this variable was evaluated by grouping the included cases according to the ratio between the number of artifacts and the number of remaining teeth (N) \u003csup\u003e(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e)\u003c/sup\u003e. As highlighted in Table\u0026nbsp;(3), by increasing (N) the mean gap distance increased regardless of the registration technique used, which means that the registration accuracy is reduced. This may return to the presence of metallic components that generate streak artifacts within the CBCT scan, and in cases of significant metal restorations, the identification of dependable reference points for registration emerges as a formidable challenge \u003csup\u003e(\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eThe weak positive result for the registration of the maxillary arch vs the mandibular arches in the AI group, may be accounted for by the larger surface area of the maxillary teeth compared to the mandibular teeth. Further, it is assumed that the difference in surface area between the upper and lower arch is also suggested to affect the registration accuracy. Therefore, we calculated the mean gap distance for each arch in each group in both registration techniques as illustrated in table (4). Only in subgroup 1 was the MGD higher in the mandible than the maxilla using both techniques which align with the aforementioned assumption. However, subgroups 2 and 3 demonstrated superior MGD in the upper arches than in the lower arches. A possible explanation may pertain to the ratio between the number of artifacts and the number of remaining natural teeth where subgroups 2 and 3 had more metal artifacts compared to subgroup 1 \u003csup\u003e(\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e18\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eBesides registration accuracy, the duration of registration is also a relevant factor that should be considered. In accordance with El Garba et al. and Piao et al. our results showed that the time difference between both groups was 2 min and 20 sec. (140 sec.) with less duration for the Ai group. To avoid the experience of the operator effect only one operator did the registration procedure by both techniques. Due to the significant difference in the speed of the internet between different countries, the time of uploading and downloading the file from the AI server 30\u0026ndash;40 sec. was deducted from the registration duration. In addition, the features of the computer used affect the speed of the software used. That\u0026rsquo;s why we used a high-performance computer. Also, using the point-based registration technique in cases of a high number of artifacts may have a direct relation to the increased registration duration as the operator spent more time choosing the registration point accurately \u003csup\u003e(\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e9\u003c/span\u003e)\u003c/sup\u003e.\u003c/p\u003e \u003cp\u003eOne of the strength points of this paper resides in the evaluated dataset, which encompasses scans with diverse dental configurations, thereby facilitating an assessment of the robustness of the artificial intelligence methodology across a multitude of clinical scenarios. Conversely, a notable limitation pertains to the necessity for a degree of caution when generalizing findings to other implant planning software that utilizes a different registration algorithm.\u003c/p\u003e"},{"header":"Conclusion","content":"\u003cp\u003eIn conclusion, AI registration may serve as an efficacious alternative to conventional registration techniques. The current findings indicate that the AI assistant feature unveiled surpassing registration accuracy compared to point-based registration. Moreover, PR accuracy is more influenced by the number of artifacts present than the AI. In addition, the AI registration accuracy did not enhance with a larger surface area model. The duration of registration was also shorter with the AI group.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgement: NA\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthors’ contributions:\u003c/strong\u003e The authors confirm their contribution to the paper as follows:\u003c/p\u003e\n\u003cp\u003econtributed to the study conception, design, and practical work; Yassin S. Alian \u0026amp; Nehal I. Shobair; data collection was performed by Nehal I. Shobair \u0026amp; Mohamed A. Tolba; Zainab Hafez and Nehal I. Shobair \u0026amp; Mohamed A. Tolba performed manuscript preparation. All authors reviewed the results and approved the final version of the manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding:\u0026nbsp;\u003c/strong\u003eSelf-funding.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eDeclarations:\u003c/strong\u003e\u003c/p\u003e\n\u003cul\u003e\n \u003cli\u003e\u003cstrong\u003eEthical approval and consent to participate :\u0026nbsp;\u003c/strong\u003eThis study was reviewed and approved by the research ethical committee at Misr International University (IRB number: MIU-IRB-2324-020) and registered on clinicaltrials.gov (NCT06273332).\u0026nbsp;The trial was performed in accordance with the principles of the Declaration of Helsinki.\u0026nbsp;All the information needed about the study was given to the patients and informed consent term was obtained from every individual participant included in the study.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eConflict of interest:\u0026nbsp;\u003c/strong\u003eThe authors declare that they have no conflict of interest and do not have any financial interest in the companies whose materials are included in this article.\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003eConsent for publication\u003c/strong\u003e: not applicable\u0026nbsp;\u003c/li\u003e\n \u003cli\u003e\u003cstrong\u003e\u0026nbsp;\u0026nbsp;Availability of data and Materials \u0026nbsp;:\u003c/strong\u003e Data available on request from the authors.\u003c/li\u003e\n\u003c/ul\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eLeite AF, Vasconcelos KD, Willems H, et al. Radiomics and machine learning in oral healthcare. Proteom Clin Appl. 2020;14(3):e1900040. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1002/prca\u003c/span\u003e\u003cspan address=\"10.1002/prca\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSingh GD, Singh M. Virtual surgical planning: modeling from the present to the future. J Clin Med. 2021;10(23):5655. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/jcm10 235655\u003c/span\u003e\u003cspan address=\"10.3390/jcm10 235655\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShujaat S, Riaz M, Jacobs R. Synergy between artificial intelligence and precision medicine for computer-assisted oral and maxillofacial surgical planning. Clin Oral Invest. 2023;27(3):897\u0026ndash;906.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlhossaini SJ, Neena AF, Issa NO, Abouelkheir HM, Gaweesh YY. Accuracy of markerless registration methods of DICOM and STL files used for computerized surgical guides in mandibles with metal restorations: An in vitro study. The Journal of Prosthetic Dentistry; 2022.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFl\u0026uuml;gge T, Derksen W, Te Poel J, Hassan B, Nelson K, Wismeijer D. Registration of cone beam computed tomography data and intraoral surface scans\u0026ndash;A prerequisite for guided implant surgery with CAD/CAM drilling guides. Clin Oral Implants Res. 2017;28(9):1113\u0026ndash;8.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChung M, Lee J, Song W et al. (2020) Automatic registration between dental cone-beam CT and scanned surface via deep pose regression neural networks and clustered similarities. IEEE Trans Med Imaging 39:3900\u0026ndash;3909. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1109/TMI\u003c/span\u003e\u003cspan address=\"10.1109/TMI\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e. 2020.30075 20.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePiao XY, Park JM, Kim H, Kim Y, Shim JS. Evaluation of different registration methods and dental restorations on the registration duration and accuracy of cone beam computed tomography data and intraoral scans: a retrospective clinical study. Clin Oral Investig Sep. 2022;26(9):5763\u0026ndash;71. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1007/s00784-022-04533-7\u003c/span\u003e\u003cspan address=\"10.1007/s00784-022-04533-7\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePark JH, Hwang CJ, Choi YJ, Houschyar KS, Yu JH, Bae SY, Cha JY. Registration of digital dental models and cone-beam computed tomography images using 3-dimensional planning software: Comparison of the accuracy according to scanning methods and software. Am J Orthod Dentofac Orthop Jun. 2020;157(6):843\u0026ndash;51. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ajodo.2019.12.013\u003c/span\u003e\u003cspan address=\"10.1016/j.ajodo.2019.12.013\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElgarba BM, Fontenele RC, Ali S, Swaity A, Meeus J, Shujaat S, Jacobs R. Validation of a novel AI-based automated multimodal image registration of CBCT and intraoral scan aiding presurgical implant planning. Clin Oral Implants Res. 2024;35:1506\u0026ndash;17. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/clr.14338\u003c/span\u003e\u003cspan address=\"10.1111/clr.14338\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJamjoom FZ, Kim DG, McGlumphy EA, Lee DJ, Yilmaz B. Positional accuracy of a prosthetic treatment plan incorporated into a cone beam computed tomography scan using surface scan registration. J Prosthet Dent Sep. 2018;120(3):367\u0026ndash;74. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.prosdent.2017.11.019\u003c/span\u003e\u003cspan address=\"10.1016/j.prosdent.2017.11.019\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKim BC, Lee CE, Park W, Kang SH, Zhengguo P, Yi CK, Lee SH. Integration accuracy of digital dental models and 3-dimensional computerized tomography images by sequential point‐ and surface‐based markerless registration. Oral Surgery, Oral Medicine, Oral Pathology. Oral Radiol Endodontology. 2010;110:370\u0026ndash;8. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tripleo.2010.03.036\u003c/span\u003e\u003cspan address=\"10.1016/j.tripleo.2010.03.036\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eJamjoom FZ, Yilmaz B, Johnston WM. Impact of number of registration points on the positional accuracy of a prosthetic treatment plan incorporated into a cone-beam computed tomography scan by surface scan registration: An in vitro study. Clin Oral Impl Res. 2019;30:826\u0026ndash;32. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1111/clr.13490\u003c/span\u003e\u003cspan address=\"10.1111/clr.13490\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChoi YS, Kim MK, Lee JW, Kang SH. Impact of the number of registration points for replacement of three-dimensional computed tomography images in dental areas using three‐dimensional light‐scanned images of dental models. Oral Radiol. 2014;30:32\u0026ndash;7. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s11282-013-0136-2\u003c/span\u003e\u003cspan address=\"10.1007/s11282-013-0136-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAl-Jarsha MY, Almezyad O, AlOtaibi N, Naudi KB, Robertson DP, Ayoub AF. The Accuracy of Intraoral Registration for Dynamic Surgical Navigation in the Edentulous Maxilla. Int J Oral Maxillofac Implants Jun. 2024;21(3):21\u0026ndash;46. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.11607/jomi.10531\u003c/span\u003e\u003cspan address=\"10.11607/jomi.10531\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eElgarba BM, Van Aelst S, Swaity A, Morgan N, Shujaat S, Jacobs R. Deep learning-based segmentation of dental implants on cone-beam computed tomography images: A validation study. J Dent Oct. 2023;137:104639. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jdent.2023.104639\u003c/span\u003e\u003cspan address=\"10.1016/j.jdent.2023.104639\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eShaheen E, Leite A, Alqahtani KA, Smolders A, Van Gerven A, Willems H, Jacobs R. A novel deep learning system for multi-class tooth segmentation and classification on cone beam computed tomography. A validation study. J Dent Dec. 2021;115:103865. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.jdent.2021.103865\u003c/span\u003e\u003cspan address=\"10.1016/j.jdent.2021.103865\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAlhossaini SJ, Neena AF, Issa NO, Abouelkheir HM, Gaweesh YY. Accuracy of markerless registration methods of DICOM and STL files used for computerized surgical guides in mandibles with metal restorations: An in vitro study. J Prosthet Dent Nov. 2024;132(5):986\u0026ndash;93. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.prosdent.2022.09.017\u003c/span\u003e\u003cspan address=\"10.1016/j.prosdent.2022.09.017\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eNoh H, Nabha W, Cho JH, Hwang HS. Registration accuracy in the integration of laser-scanned dental images into maxillofacial cone-beam computed tomography images. Am J Orthod Dentofac Orthop Oct. 2011;140(4):585\u0026ndash;91. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003e10.1016/j.ajodo.2011.04.018\u003c/span\u003e\u003cspan address=\"10.1016/j.ajodo.2011.04.018\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":false,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"bmc-oral-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ohea","sideBox":"Learn more about [BMC Oral Health](http://bmcoralhealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ohea/default.aspx","title":"BMC Oral Health","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true},"keywords":"AI registration, point-based registration, segmentation, metal artifacts","lastPublishedDoi":"10.21203/rs.3.rs-9157142/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9157142/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjectives\u003c/h2\u003e \u003cp\u003eThis study assessed the accuracy and duration of AI-assisted registration versus point-based registration in CBCT scans with heavy metal artifacts, which is measured as median gap distance, and duration between AI-assisted registration and conventional point-based registration using coDiagnostiX software on CBCT scans exhibiting heavy metal artifacts. Secondary objectives include evaluating the influence of artifact-to-natural teeth ratio (subgroups: 0\u0026thinsp;\u0026lt;\u0026thinsp;N\u0026thinsp;\u0026lt;\u0026thinsp;50, 50\u0026thinsp;\u0026le;\u0026thinsp;N\u0026thinsp;\u0026lt;\u0026thinsp;100, 100\u0026thinsp;\u0026le;\u0026thinsp;N) and arch surface area (maxilla vs. mandible) on registration performance in both methods.\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003e90 CBCT images and intraoral scans were included following specific eligibility criteria, then divided equally among the two registration groups: point-based and AI-assisted registration each group was subdivided into 3 subgroups according to the ratio of artifacts to natural teeth. The CBCT scans were segmented utilizing the AI-assistant feature via coDiagnostiX software (coDiagnostiX\u0026reg; 10.7, Dental Wings Inc., Montreal, Canada). Afterward, point-based registration and AI-assisted registration were performed by a single operator using coDiagnostiX software. Then, the registration accuracy was examined by measuring the median gap distance between the three-dimensional models of CBCT data and the registered intraoral scans. Also, the duration required for registration was calibrated and recorded by a stopwatch.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe registration accuracy of the AI group was superior to the PR group, where the median gap distance in the PR group was 0.278 mm and 0.202 mm in the AI group, with no statistically significant correlation between the groups p-value\u0026thinsp;\u0026gt;\u0026thinsp;0.05. SubGroup 1 had the highest accuracy in the PR groups, while Subgroup 3 showed the greatest accuracy in the AI groups. With regards to the influence of the surface area, in the PR groups, the upper arch in group 1 was the most accurate. On the other hand, the lower arch in group 3 had the least accuracy. Concerning the AI groups, the mandible in Subgroup 3 showed the highest accuracy, whereas the maxilla in group 2 highlighted the lowest accuracy.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eAI integration in virtual implant software could enhance and ease the segmentation and registration steps for clinicians with superior accuracy and in a shorter duration compared to conventional point-based registration techniques.\u003c/p\u003e\u003ch2\u003eClinical relevance\u003c/h2\u003e \u003cp\u003e: In oral implantology, metal artifacts from restorations compromise CBCT segmentation and registration, critical for precise digital surgical planning and guided implant placement. AI integration offers clinicians an efficient, operator-independent alternative, reducing errors in edentulous or heavily restored arches and enhancing predictability in implant workflows. These findings support adopting AI tools to improve surgical guide accuracy and clinical outcomes in challenging cases\u003c/p\u003e\u003ch2\u003eClinical trial registration number:\u003c/h2\u003e \u003cp\u003e(NCT06273332), date of registration 5/02/2024,first posted 22/02/2024\u003c/p\u003e","manuscriptTitle":"Assessment of Registration Accuracy and Duration Using AI-assisted Registration Versus Conventional Point-based Registration on CBCT Scans with Heavy Metal Artifacts","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-05-11 08:37:47","doi":"10.21203/rs.3.rs-9157142/v1","editorialEvents":[{"type":"communityComments","content":0},{"type":"reviewerAgreed","content":"63617909550971832009367376522628941352","date":"2026-05-04T09:25:05+00:00","index":"hide","fulltext":""},{"type":"reviewerAgreed","content":"132059290103940345876840499796693357669","date":"2026-05-03T00:00:52+00:00","index":"hide","fulltext":""},{"type":"reviewersInvited","content":"","date":"2026-04-30T07:11:41+00:00","index":"","fulltext":""},{"type":"editorInvited","content":"","date":"2026-04-20T12:30:44+00:00","index":"","fulltext":""},{"type":"editorAssigned","content":"","date":"2026-03-30T13:30:50+00:00","index":"","fulltext":""},{"type":"checksComplete","content":"","date":"2026-03-27T19:08:30+00:00","index":"","fulltext":""},{"type":"submitted","content":"BMC Oral Health","date":"2026-03-27T19:05:06+00:00","index":"","fulltext":""}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"bmc-oral-health","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":false,"externalIdentity":"ohea","sideBox":"Learn more about [BMC Oral Health](http://bmcoralhealth.biomedcentral.com/)","snPcode":"","submissionUrl":"https://www.editorialmanager.com/ohea/default.aspx","title":"BMC Oral Health","twitterHandle":"BMC_series","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"em","reportingPortfolio":"BMC Series","inReviewEnabled":true,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"9e007858-4b0c-4b7c-8474-4ce0ada5f52c","owner":[],"postedDate":"May 11th, 2026","published":true,"recentEditorialEvents":[{"type":"reviewerAgreed","content":"63617909550971832009367376522628941352","date":"2026-05-04T09:25:05+00:00","index":37,"fulltext":""},{"type":"reviewerAgreed","content":"132059290103940345876840499796693357669","date":"2026-05-03T00:00:52+00:00","index":36,"fulltext":""},{"type":"reviewersInvited","content":"4","date":"2026-04-30T07:11:41+00:00","index":"","fulltext":""}],"rejectedJournal":[],"revision":"","amendment":"","status":"under-review","subjectAreas":[],"tags":[],"updatedAt":"2026-05-11T08:37:47+00:00","versionOfRecord":[],"versionCreatedAt":"2026-05-11 08:37:47","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9157142","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9157142","identity":"rs-9157142","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}