Artificial Intelligence for Root Canal Segmentation on Radiographic Images: A Scoping Review

Artificial Intelligence for Root Canal Segmentation on Radiographic Images: A Scoping Review | 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 Systematic Review Artificial Intelligence for Root Canal Segmentation on Radiographic Images: A Scoping Review Maryam Ghiasi, Ava Zaboli, Mina Mahdian, Sanjay M. Mallya This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5952801/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objectives: The number, size, patency, and location of pulp canals are critical in endodontic treatment planning. This information is currently obtained through visual radiographic assessment, which is time-consuming and labor-intensive. Artificial intelligence (AI) could automate this task via accurate segmentation of root canals providing efficiency and consistency. This scoping review maps existing literature on the use of AI to automate root canal segmentation on radiographic images. Materials and Methods: We searched MEDLINE (Ovid), Embase, Scopus, and Web of Science for relevant studies up to January 8, 2025. Studies that used AI for root canal segmentation were included. Study selection was not limited by design, language, or date. Commentaries, retracted articles, and inaccessible full-text articles were excluded. Titles and abstracts were screened based on eligibility criteria, and the full-text of potentially relevant studies was assessed. Screening and data extraction were conducted in duplicate by independent reviewers, with disagreements resolved via consensus or a third-reviewer if necessary. Results: Out of 836 articles identified, 35 studies met the eligibility criteria and were retained for synthesis. Data extraction focused on the country of origin, study design, imaging modalities, obturation status, type of teeth analyzed, AI models used, and results. Modalities included were mostly cone beam computed tomography (CBCT, 51%), followed by panoramic (17%) and periapical (14%). AI-based models, particularly those employing CNNs, reported accuracies ranging from 0.73 to 0.99 and sensitivities from 0.72 to 1. These models were effective across all imaging modalities with most studies reporting improved diagnostic precision and reduced time compared with manual methods. Conclusions: AI-based root canal segmentation has clinical value by increasing accuracy in identifying root canal anatomy prior to treatment. This will preserve clinicians' time and reduce the risk of treatment failure. This review highlights current status of this technological application and identifies areas to refine these technologies for broad clinical application to enhance patient outcomes in endodontic care. Clinical Relevance: The application of AI in root canal segmentation offers significant clinical benefits by improving the accuracy and efficiency of identifying root canal anatomy. This can lead to better treatment planning, reduced procedure times, and lower risk of endodontic failure. As AI technology continues to evolve, its integration into endodontic practice has the potential to enhance patient outcomes and streamline clinical workflows. Figures Figure 1 Figure 2 Figure 3 Figure 4 1. Introduction The number, size, patency, and location of pulp canals are critical in endodontic treatment planning. Accurate knowledge of these parameters is essential to ensure successful endodontic outcomes, as the inability to locate, debride and disinfect all canals is a common cause of endodontic failure(1). Variations in canal anatomy, which can be influenced by factors such as age, ethnicity, and root anatomy, further complicate this task(2,3). A comprehensive three-dimensional understanding of pulp morphology is essential for successful endodontic treatment outcomes(4). Traditional methods for pulp canal identification and segmentation involve manual or semi-automated approaches, both of which are time-consuming and labor-intensive. Manual segmentation, in particular, requires extensive user interaction and can take considerable time, depending on the complexity of the cases(5). Artificial intelligence (AI) offers a promising solution to these challenges. AI consists of a variety of technologies that emulate intelligent human behavior through experiential learning(6). Deep learning (DL), a subset of AI, uses neural networks with multiple hidden layers that function similarly to biological neurons, learning by identifying features from input data(7). Convolutional neural networks (CNNs), a subset of neural networks, have been particularly successful in processing complex data structures like images(8). In healthcare, AI and DL are increasingly being used to enhance disease prediction, diagnosis, and management(7,9,10). In dentistry, AI has shown potential in various applications, including the detection of caries, periapical lesions, cracks, root fractures, and the determination of working length and tooth labeling(11–15). However, the application of AI in endodontics, particularly for root canal segmentation, remains relatively underexplored. Existing reviews have had limited scope(16). AI-driven approaches can potentially automate the segmentation of root canals from radiographic images, providing efficiency, consistency, and accuracy. Given the rapidly evolving nature of AI technology and its growing application in dental practice, it is timely to conduct a comprehensive scoping review to map the literature on AI-driven solutions for automated root canal segmentation. This scoping review aims to map the existing literature on the use of AI to automate root canal segmentation on radiographic images. It provides an updated overview of the current state of research in this area, highlights the gaps in existing research, and propose future research directions, emphasizing the potential of AI to transform endodontic practice and enhance treatment results. 2. Materials and Methods This scoping review was carried out using a predefined protocol in accordance with the PRISMA Extension for Scoping Reviews guidelines(17,18). Protocol development and review conduct were guided by best practice principles outlined by the Cochrane Handbook for Systematic Reviews of Interventions. There were no deviations from the protocol. Our approach followed the five-step framework originally developed by Arksey and O’Malley(19), and later refined by Levac et al(20). The framework involves: (1) formulating the research question, (2) identifying relevant studies, (3) selecting the studies, (4) charting the data, and (5) collating, summarizing, and reporting the findings. Due to the novelty and limited research available on this topic, resulting in a heterogeneous body of literature, the scoping review method was deemed the most suitable. This study did not require ethical approval. 2.1 Information sources and search strategy An experienced medical information specialist developed and refined the search strategies through an iterative process in collaboration with the main reviewer. The MEDLINE (Ovid), Embase, Scopus, and Web of Science were searched for all studies published since the database's inception up to January 8, 2025. These four health sciences major electronic databases were comprehensively searched using Medical Subject Headings (MeSH), keywords, truncations, adjacency functions, and Boolean operators. Additionally, Google Scholar was searched using the same keywords applied in the above-mentioned electronic databases to find further relevant sources of gray literature. The complete search strategy used to locate articles in electronic databases is detailed in Supplementary material 1. Following the electronic database search, the reference lists of the included articles were manually searched for relevant articles. 2.2 Eligibility criteria Studies were included in the review if they investigated or reported on application of artificial intelligence for dental canal segmentation. Studies that did not meet the inclusion criteria were excluded from this review. Inclusion criteria: All types of radiographic images including cone-beam computed tomography (CBCT), panoramic, periapical, and bitewing are included. Study selection is not limited to any specific study design; observational and interventional studies are included. There is no restriction on the date of publication or language of the studies. There is no limitation on the type of teeth investigated. There is no limitation on the status of the obturation of the root canals (i.e., both obturated and unobturated canals are included). Exclusion criteria: Studies not utilizing artificial Intelligence are excluded. Studies not using radiographic images to assess the root canals are excluded. Studies not focusing on pulp canal segmentation are excluded (e.g., studies reporting on tooth segmentation without focusing on root canals are excluded). Studies not having canal segmentation as their objective with no details on the segmentation methods and results are excluded. Commentaries, retracted articles, and inaccessible full-text articles are excluded. Review articles are excluded if they do not contain relevant studies that are not already included in our review. 2.3 Study selection Study selection was performed independently by two reviewers (MG and AZ). Initially, titles and abstracts were screened to remove irrelevant studies (stage 1) using the Covidence web application(21). For potentially relevant studies, full texts were retrieved and comprehensively assessed to make sure they were compliant with the eligibility criteria (stage 2). References from all included publications that were identified as potentially relevant or which the reviewers were unclear of its inclusion were also considered for full-text screening. When there was uncertainty, a third-reviewer with expertise in subject matter was consulted. A pilot-test screening was done for 20 studies before the main screening in each stage to train and calibrate the reviewers to the eligibility criteria and ensure consistent application of the criteria. Reasons for the exclusion of each study were documented. 2.4 Charting the data Data extraction was independently performed by two reviewers (MG and AZ) using the Covidence web application a pre-developed extraction form. Any disagreements were resolved through discussion and consensus, or by involving a third reviewer if necessary. The data mapping process involved two main steps. First, the data was organized by examining the included studies, classifying them, and sorting them according to the research objectives. Second, an explanatory summary of the data was generated to derive insights. Reviewers thoroughly reviewed and charted all included full-text articles, with the data being summarized narratively and compiled into tables. Extracted data consisted of the title, authorship, year of publication, country of origin, study design, imaging modalities, obturation status, type of teeth analyzed, AI models used, and results. 3. Results 3.1 Study selection Our electronic search initially identified 836 articles. After examining the titles and abstracts, 79 articles were selected for a full-text review. Of these, 13 were excluded for not investigating root canal/pulp specifically, 14 were excluded for not performing segmentation, and 6 were excluded as canal segmentation was not the objective of the studies without providing any details on the segmentation methods and results. Additionally, 8 review articles were excluded because their references did not include any new eligible study not already covered in our review. Four studies were excluded for the following reasons: not using AI in their methods, being retracted, being a commentary, and the full-text not being found. This left 35 studies that met the eligibility criteria and were included in the final synthesis. Data extraction was then performed on these eligible articles, and the results are presented in this paper. The screening and selection process of the studies is illustrated in Fig. 1 , adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowchart. 3.2 Characteristics of included studies The included studies were published between 2011 and 2024, with one study from 2011(22), another from 2012(23), and the remaining studies published from 2019 to 2024. Most of the studies originated from China (26%), followed by Turkey (17%) and the United States (11%). Figure 2 provides a detailed breakdown of the number of studies by country of origin. The characteristics of the 35 included articles are summarized in Table 1 . 3.3 Content analysis Two studies were conducted prospectively and ex vivo(4,24), while the remaining 33 were conducted retrospectively in vivo. Most of the studies used CBCT images (51%) to train their AI models, followed by panoramic (17%) and periapical images (14%). Figure 3 illustrates the distribution of studies based on the diagnostic techniques used for AI model training. Regarding the obturation status of the teeth used for training, 34% of the studies did not specify whether they used obturated or unobturated canals, 23% used obturated canals, 26% used unobturated canals, and 17% used both types. Most of the studies did not specify the type of teeth used for AI model training. Figure 4 presents the number of studies categorized by each tooth type. Two older studies, published in 2011 and 2012, employed histogram-based AI models: Histogram-based Quick Fuzzy C-means Clustering(22) and a Histogram-based Enhanced Version(23). The remaining 33 studies utilized CNNs. Regarding the ground truth, 7 studies did not specify their ground truth sources. Three studies used micro-CT(4,24,25), while the remaining studies relied on dental experts—including dentists, endodontists, and dentomaxillofacial radiologists—for clinical confirmation or manual segmentation of the root canals. Outcome metrics The performance of the AI models for canal segmentation was primarily evaluated using the Dice Similarity Coefficient (DSC), also known as the F1 score, which ranged from 0.6 to 0.98 ( 22 studies(4,5,24–43)). The majority of the studies reported several metrics. Sensitivity values ranged from 0.72 to 1 (11 studies(26,28,29,31,35,39,40,44–47)), while specificity values varied from 0.23 to 1 (7 studies(26,28,40,44–47)). Accuracy was reported between 0.73 and 0.99 (15 studies(23,26,28,38,40,42–44,46–52)), and precision ranged from 0.66 to 1 (12 studies(27,31,32,35,36,39,40,42,46,47,49,53)). The area under the curve (AUC) spanned from 0.58 to 0.99 (7 studies(26–28,31,40,46,50)). Less frequent metrics were also reported including Positive Predicte Value(29,44), Negative Predictive Value(44), Hausdorff distance (HD)(25,33,53), and inter-over union (IoU)(31,41–43). Lin et al.(25) reported that only thick root canals could be segmented, unlike the thin root canals, such as those in the apical third and lateral root canals. Lee et al.(40) showed that certain features, including canal visibility, were less distinguishable. Szilágyi et al.(22) indicated that while their algorithm performed automatically in most cases, some decisions required interactive input. Wang et al.(33) demonstrated that their AI workflow reduced the time needed to obtain a 3D model of the tooth and root canal from 6 hours to 2 minutes, resulting in satisfactory outcomes for complex root canal treatments. Similarly, Song et al(5), showed that automatic segmentation took 13–33 s while manual segmentation needs 15–25 min. Schneider et al.(54) found that deeper, more complex models did not necessarily outperform less complex alternatives. Rohrer et al.(37) used a tiling approach on panoramic images, resulting in a performance improvement of + 294%; the mean F1-score increased from 0.33 to 0.9. Gardiyanoglu et al.(38) reported a fair DSC value of 0.78 for successful root canal treatments but noted that cases with inadequate root canal fillings and gutta-perchas superimposed with neighboring teeth's roots caused relatively lower DSC values. Orhan et al.(45) found a sensitivity of 82.1% for root canal fillings, but only 23.0% for the assessment of voids in the root canal. Four studies investigated c-shaped canal segmentation in mandibular second molars(29,46,47,50). One of them(47) showed that the neural network's classification performance, when augmented with GAN-synthesized periapical images, improved compared to using real data alone. Hasan et al.(52) demonstrated that imbalanced datasets with noise removal led to YOLOv5x's prediction accuracy dropping to 72%, while balancing and noise removal led to all models (YOLOv5s, YOLOv5x, and YOLOv7) performing at over 95% accuracy. Mean average precision (mAP) improved from 52–92% following balancing and denoising. 4. Discussion This study is the most comprehensive review of AI models for root canal segmentation conducted to date, encompassing a wide range of studies and providing a detailed understanding of the current landscape in this field. Our analysis of studies published between 2011 and 2024 highlights significant progress and ongoing challenges in the development and application of AI in endodontics. It is also noteworthy that the majority of studies in this field have originated from China (26%), followed by Turkey (17%) and the United States (11%). Several factors might explain why China leads in publications, including strong government support for research and development, a large domestic market that drives demand for advanced dental technologies, widespread availability of data, effective collaboration between industry and academia, and their investment in education and research infrastructure(55). Understanding the regional strengths and focus areas can help in fostering international collaborations and addressing any gaps in research coverage. However, the development of deep learning algorithms trained on regional morphology may impact the global applicability of such tools. Baydar et al.(35) and Ari et al.(39) both reported on the effectiveness of these models in both unobturated and obturated canals separately. Performance metrics such as F1 scores, sensitivity, and precision were higher for obturated canals, which may suggest that the AI models are particularly adept at identifying high contrast objects such root canal fillings. This should be further investigated to confirm that the observed differences are not due to chance. Lee et al.(40) reported low AUC (0.58) and specificity (0.23), which were attributed to the difficulty in distinguishing less visible canal features. This suggests that their DCNN model has room for improvement, and refining machine learning algorithms and incorporating advanced image processing in future studies could enhance both feature detection and endodontic prognostication accuracy. A notable gap in the literature is the lack of consideration for patient age. None of the studies accounted for the impact of age on AI performance, despite the known tendency for dental canals to become more calcified with age, potentially affecting model accuracy. Future research should investigate this variable to ensure the robustness of AI models across different age groups. The majority of studies did not specify the type of teeth used for training, implying that their datasets included various tooth types and conditions. This generalizability is beneficial for developing AI models applicable to a wide range of clinical scenarios. However, focusing on specific tooth types could also allow for more targeted model development and potentially higher accuracy for application in those particular tooth types. Additionally, while some studies demonstrated the potential of AI to significantly reduce the time required for clinical tasks, such as Wang et al.'s reduction from 6 hours to 2 minutes for 3D modeling(33) and Song et al.’s reduction from 15–20 minutes to 13–33 seconds for segmentation (manual versus automatic), it is essential to balance speed with accuracy. The findings by Schneider et al.(54) that more complex models do not always yield better results emphasize the importance of optimizing model complexity. Future studies should strive for more standardized and detailed methodologies, along with greater transparency in their reporting. Standardized reporting would allow for a more homogeneous set of studies, facilitating a more comprehensive analysis of various features, such as different CNN models, obturated versus unobturated canals, and the types of teeth used in the datasets. By establishing uniform criteria for data collection, annotation, and evaluation metrics, researchers can ensure that studies are comparable and replicable, which is critical for validating AI models and their clinical utility(56–58). Moreover, detailed reporting should include comprehensive descriptions of the datasets, including patient demographics, imaging modalities, and the ground truth used for training and validation. This level of detail will be essential in identifying the factors that contribute to the success or failure of AI models, enabling more accurate interpretation of results. In addition, future research should consider stratifying results based on clinically relevant subgroups, such as age, sex, and specific dental conditions, to assess the robustness and generalizability of AI models across diverse patient populations. This stratification could reveal insights into how different variables impact model performance and help in developing more tailored AI solutions for various clinical scenarios. Collaborations among researchers, clinicians, computer science experts, and institutions can further enhance the quality of AI research in endodontics. Collaborative efforts can lead to larger, more diverse datasets and the sharing of best practices, ultimately accelerating the development and implementation of AI technologies in dental practice. A key strength of this review is its comprehensive scope, ensuring that all published studies using AI for root canal segmentation are included, thus providing a detailed map of the current literature. However, a significant limitation is the heterogeneity in the parameters and metrics reported across the studies, making it impossible to perform a full systematic review or meta-analysis. This variability underscores the need for standardized reporting and methodological approaches in future research. In addition, in line with established methods for scoping reviews, we did not conduct a risk of bias assessment of the included studies; this could be done in a future, full systematic review. While the AI models reviewed show great potential for improving endodontic treatment planning and outcomes, the results should be interpreted with caution due to certain limitations. Some studies lacked detailed methodology reporting, including their ground truth sources, which can affect the reliability of the findings. Additionally, as this is a scoping review aimed at mapping the literature, we did not perform a Risk of Bias analysis on the studies, further emphasizing the need for cautious interpretation of the results for clinical practice. The use of AI in root canal segmentation and endodontics is still in its initial stages, indicating significant room for refinement and improvement of the models. Future studies should address these gaps and standardize methodologies to enhance the reliability and clinical utility of AI in dental practice. Declarations Data Availability Statement All data supporting the findings of this study are included in this article. As this study is a scoping review, no new datasets were generated. Funding statement This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors. Ethics approval statement Ethics approval is not required for this scoping review as the review uses solely published literature. Relevant reporting guideline This scoping review was carried out using a predefined protocol in accordance with the PRISMA Extension for Scoping Reviews guidelines(17,18). The document is included in Supplementary material 2. Ethics and Integrity Statement This research adheres to rigorous ethical standards and integrity policies, ensuring transparency, accountability, and respect for all participants. Upholding these standards is essential for the credibility and reliability of our findings. Acknowledgements The authors would like to thank Risa Shorr (Learning Services, The Ottawa Hospital) for assisting in the creation of the literature search. References Tabassum S, Khan FR. Failure of endodontic treatment: The usual suspects. Eur J Dent. 2016;10(1):144–7. Martins JNR, Alkhawas MBAM, Altaki Z, Bellardini G, Berti L, Boveda C, et al. Worldwide Analyses of Maxillary First Molar Second Mesiobuccal Prevalence: A Multicenter Cone-beam Computed Tomographic Study. J Endod. 2018 Nov;44(11):1641-1649.e1. Pattanshetti N, Gaidhane M, Al Kandari AM. Root and canal morphology of the mesiobuccal and distal roots of permanent first molars in a Kuwait population--a clinical study. Int Endod J. 2008 Sep;41(9):755–62. Lin X, Fu Y, Ren G, Yang X, Duan W, Chen Y, et al. Micro-Computed Tomography-Guided Artificial Intelligence for Pulp Cavity and Tooth Segmentation on Cone-beam Computed Tomography. J Endod. 2021 Dec;47(12):1933–41. Song Y, Yang H, Ge Z, Du H, Li G. Age estimation based on 3D pulp segmentation of first molars from CBCT images using U-Net. Dento Maxillo Facial Radiol. 2023 Oct;52(7):20230177. Davenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019 Jun;6(2):94–8. Ching T, Himmelstein DS, Beaulieu-Jones BK, Kalinin AA, Do BT, Way GP, et al. Opportunities and obstacles for deep learning in biology and medicine. J R Soc Interface. 2018 Apr;15(141):20170387. Schwendicke F, Samek W, Krois J. Artificial Intelligence in Dentistry: Chances and Challenges. J Dent Res. 2020 Jul;99(7):769–74. Park WJ, Park JB. History and application of artificial neural networks in dentistry. Eur J Dent. 2018;12(4):594–601. Cheng PM, Montagnon E, Yamashita R, Pan I, Cadrin-Chênevert A, Perdigón Romero F, et al. Deep Learning: An Update for Radiologists. Radiogr Rev Publ Radiol Soc N Am Inc. 2021;41(5):1427–45. Musri N, Christie B, Ichwan SJA, Cahyanto A. Deep learning convolutional neural network algorithms for the early detection and diagnosis of dental caries on periapical radiographs: A systematic review. Imaging Sci Dent. 2021 Sep;51(3):237–42. Umer F, Habib S. Critical Analysis of Artificial Intelligence in Endodontics: A Scoping Review. J Endod. 2022;48(2):152–60. Asiri A.F., Altuwalah A.S. The role of neural artificial intelligence for diagnosis and treatment planning in endodontics: A qualitative review. Saudi Dent J. 2022;34(4):270–81. Ahmed ZH, Almuharib AM, Abdulkarim AA, Alhassoon AH, Alanazi AF, Alhaqbani MA, et al. Artificial Intelligence and Its Application in Endodontics: A Review. J Contemp Dent Pract. 2023;24(11):912–7. Khanagar SB, Alfadley A, Alfouzan K, Awawdeh M, Alaqla A, Jamleh A. Developments and Performance of Artificial Intelligence Models Designed for Application in Endodontics: A Systematic Review. Diagn Basel Switz. 2023;13(3). Alfadley A, Shujaat S, Jamleh A, Riaz M, Aboalela AA, Ma H, et al. Progress of Artificial Intelligence-Driven Solutions for Automated Segmentation of Dental Pulp Space on Cone-Beam Computed Tomography Images. A Systematic Review. J Endod. 2024 May;S0099239924003364. Khalil H, Peters MD, Tricco AC, Pollock D, Alexander L, McInerney P, et al. Conducting high quality scoping reviews-challenges and solutions. J Clin Epidemiol. 2021 Feb;130:156–60. Peters MDJ, Marnie C, Tricco AC, Pollock D, Munn Z, Alexander L, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evid Synth. 2020 Oct;18(10):2119–26. Arksey H, O’Malley L. Scoping studies: towards a methodological framework. Int J Soc Res Methodol. 2005 Feb;8(1):19–32. Levac D, Colquhoun H, O’Brien KK. Scoping studies: advancing the methodology. Implement Sci. 2010 Dec;5(1):69. Covidence systematic review software [Internet]. Melbourne, Australia: Veritas HealthInnovation; [cited 2024 Mar 1]. Available from: www.covidence.org Szilágyi L, Dobó-Nagy C, Benyó B. Identification of the root canal from dental micro-CT records. In 2011. p. 339–46. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-81855226116&doi=10.1007%2f978-3-642-25085-9_40&partnerID=40&md5=4318c1e0537e79cea3271b4510b9f05c Benyo B. Identification of dental root canals and their medial line from micro-CT and cone-beam CT records. Biomed Eng Online. 2012;11((Benyo) Budapest University of Technology and Economics, Department of Control Engineering and Information Technology, Hungary.):81. Hatvani J, Horvath A, Michetti J, Basarab A, Kouame D, Gyongy M. Deep learning-based super-resolution applied to dental computed tomography. IEEE Trans Radiat Plasma Med Sci. 2019;3(2):120–8. Lin X, Fu YJ, Ren GQ, Wen JH, Chen YF, Zhang Q. [Segmentation and accuracy validation of mandibular molar and pulp cavity on cone-beam CT images by U-net neural network]. Shanghai Kou Qiang Yi Xue Shanghai J Stomatol. 2022;31(5):454–9. Dumont M, Prieto JC, Brosset S, Cevidanes L, Bianchi J, Ruellas A, et al. Patient Specific Classification of Dental Root Canal and Crown Shape. Shape Med Imaging Int Workshop ShapeMI 2020 Held Conjunction MICCAI 2020 Lima Peru Oct 4 2020 Proc. 2020;12474(101773457):145–53. Deleat-Besson R, Le C, Zhang W, Turkestani NA, Cevidanes L, Bianchi J, et al. Merging and Annotating Teeth and Roots from Automated Segmentation of Multimodal Images. In Springer Science and Business Media Deutschland GmbH; 2021. p. 81–92. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85118156750&doi=10.1007%2f978-3-030-89847-2_8&partnerID=40&md5=9a4c5b428267988b9b1aac6d2a797dcd Deleat-Besson R, Le C, Al Turkestani N, Zhang W, Dumont M, Brosset S, et al. Automatic Segmentation of Dental Root Canal and Merging with Crown Shape. Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Annu Int Conf. 2021;2021(101763872):2948–51. Sherwood AA, Sherwood AI, Setzer FC, K SD, Shamili JV, John C, et al. A Deep Learning Approach to Segment and Classify C-Shaped Canal Morphologies in Mandibular Second Molars Using Cone-beam Computed Tomography. J Endod. 2021;47(12):1907–16. Zhang J, Xia W, Dong J, Tang Z, Zhao Q. Root Canal Segmentation in CBCT Images by 3D U-Net with Global and Local Combination Loss. Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Annu Int Conf. 2021;2021(101763872):3097–100. Duman SB, Celik Ozen D, Bayrakdar IS, Baydar O, Alhaija ESA, Helvacioglu Yigit D, et al. Second mesiobuccal canal segmentation with YOLOv5 architecture using cone beam computed tomography images. Odontology. 2023;(101134822). Kim SH, Song IS, Baek SJ. Automatic Segmentation of Internal Tooth Structure from CBCT Images Using Hierarchical Deep Learning. In Springer Science and Business Media Deutschland GmbH; 2023. p. 703–13. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85174730962&doi=10.1007%2f978-3-031-43898-1_67&partnerID=40&md5=e03879beefbd8314dc165d7f73d05b34 Wang Y, Xia W, Yan Z, Zhao L, Bian X, Liu C, et al. Root canal treatment planning by automatic tooth and root canal segmentation in dental CBCT with deep multi-task feature learning. Med Image Anal. 2023;85(c8s, 9713490):102750. Tan M, Cui Z, Zhong T, Fang Y, Zhang Y, Shen D. A progressive framework for tooth and substructure segmentation from cone-beam CT images. Comput Biol Med. 2024;169(doc, 1250250):107839. Baydar O., Rozylo-Kalinowska I., Futyma-Gabka K., Saglam H. The U-Net Approaches to Evaluation of Dental Bite-Wing Radiographs: An Artificial Intelligence Study. Diagnostics. 2023;13(3):453. Vinayahalingam S, Goey RS, Kempers S, Schoep J, Cherici T, Moin DA, et al. Automated chart filing on panoramic radiographs using deep learning. J Dent. 2021;115(hx1, 0354422):103864. Rohrer C, Krois J, Patel J, Meyer-Lueckel H, Rodrigues JA, Schwendicke F. Segmentation of Dental Restorations on Panoramic Radiographs Using Deep Learning. Diagn Basel Switz. 2022;12(6). Gardiyanoglu E, Unsal G, Akkaya N, Aksoy S, Orhan K. Automatic Segmentation of Teeth, Crown-Bridge Restorations, Dental Implants, Restorative Fillings, Dental Caries, Residual Roots, and Root Canal Fillings on Orthopantomographs: Convenience and Pitfalls. Diagn Basel Switz. 2023;13(8). Ari T, Saglam H, Oksuzoglu H, Kazan O, Bayrakdar IS, Duman SB, et al. Automatic Feature Segmentation in Dental Periapical Radiographs. Diagn Basel Switz. 2022;12(12). Lee J, Seo H, Choi YJ, Lee C, Kim S, Lee YS, et al. An Endodontic Forecasting Model Based on the Analysis of Preoperative Dental Radiographs: A Pilot Study on an Endodontic Predictive Deep Neural Network. J Endod. 2023;49(6):710–9. Gamal M, Baraka M, Torki M. Automatic Mandibular Semantic Segmentation of Teeth Pulp Cavity and Root Canals, and Inferior Alveolar Nerve on Pulpy3D Dataset. In: Linguraru MG, Dou Q, Feragen A, Giannarou S, Glocker B, Lekadir K, et al., editors. Medical Image Computing and Computer Assisted Intervention – MICCAI 2024 [Internet]. Cham: Springer Nature Switzerland; 2024 [cited 2025 Jan 11]. p. 14–23. (Lecture Notes in Computer Science; vol. 15008). Available from: https://link.springer.com/10.1007/978-3-031-72111-3_2 Slim ML, Jacobs R, De Souza Leal RM, Fontenele RC. AI-driven segmentation of the pulp cavity system in mandibular molars on CBCT images using convolutional neural networks. Clin Oral Investig. 2024 Nov 21;28(12):650. Çelik B, Genç MZ, Çelik ME. Evaluation of root canal filling length on periapical radiograph using artificial intelligence. Oral Radiol. 2025 Jan;41(1):102–10. Albitar L., Zhao T., Huang C., Mahdian M. Artificial Intelligence (AI) for Detection and Localization of Unobturated Second Mesial Buccal (MB2) Canals in Cone-Beam Computed Tomography (CBCT). Diagnostics. 2022;12(12):3214. Orhan K, Aktuna Belgin C, Manulis D, Golitsyna M, Bayrak S, Aksoy S, et al. Determining the reliability of diagnosis and treatment using artificial intelligence software with panoramic radiographs. Imaging Sci Dent. 2023;53(3):199–208. Yang S, Kim KD, Kise Y, Nozawa M, Mori M, Takata N, et al. External Validation of the Effect of the Combined Use of Object Detection for the Classification of the C-Shaped Canal Configuration of the Mandibular Second Molar in Panoramic Radiographs: A Multicenter Study. J Endod. 2024;(i1k, 7511484, 17910050r). Yang S, Kim KD, Ariji E, Takata N, Kise Y. Evaluating the performance of generative adversarial network-synthesized periapical images in classifying C-shaped root canals. Sci Rep. 2023;13(1):18038. Orhan K, Bilgir E, Bayrakdar IS, Ezhov M, Gusarev M, Shumilov E. Evaluation of artificial intelligence for detecting impacted third molars on cone-beam computed tomography scans. J Stomatol Oral Maxillofac Surg. 2021;122(4):333–7. Madhuri J, Mohan S, Praveen S, Chandan T, Reddy Gurram T, Anitha Kumari R, et al. Detection of Mid Mesial Canal in Dental CBCT Images Using Combined YOLO and 2D U-Net Model. In Institute of Electrical and Electronics Engineers Inc.; 2023. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85175401195&doi=10.1109%2fNMITCON58196.2023.10275873&partnerID=40&md5=332dd5428d5350c6ef3f5eab2d082c5c Zhang L, Xu F, Li Y, Zhang H, Xi Z, Xiang J, et al. A lightweight convolutional neural network model with receptive field block for C-shaped root canal detection in mandibular second molars. Sci Rep. 2022;12(1):17373. Zhou Y, Zhang H. Root Canal Therapy Evaluation Based on Rule Embedded Neural Networks. In Institute of Electrical and Electronics Engineers Inc.; 2021. p. 749–54. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85124403156&doi=10.1109%2fICAA53760.2021.00136&partnerID=40&md5=fc353ba1d5b1d5258753caa5fcab9551 Hasan HA, Saad FH, Ahmed S, Mohammed N, Farook TH, Dudley J. Experimental validation of computer-vision methods for the successful detection of endodontic treatment obturation and progression from noisy radiographs. Oral Radiol. 2023;39(4):683–98. Lin X, Fu Y, Ren G, Yang X, Duan W, Chen Y, et al. Micro-Computed Tomography-Guided Artificial Intelligence for Pulp Cavity and Tooth Segmentation on Cone-beam Computed Tomography. J Endod. 2021;47(12):1933–41. Schneider L, Arsiwala-Scheppach L, Krois J, Meyer-Lueckel H, Bressem KK, Niehues SM, et al. Benchmarking Deep Learning Models for Tooth Structure Segmentation. J Dent Res. 2022;101(11):1343–9. Mnekhir HJ. THE US-CHINESE RACE IN ARTIFICIAL INTELLIGENCE CHALLENGES AND OPPORTUNITIES. Russ Law J [Internet]. 2023 Apr 7 [cited 2024 Jul 21];11(3). Available from: https://russianlawjournal.org/index.php/journal/article/view/2182 Valente NA, Tichy A, Schwendicke F, Chaurasia A, Hamdan M, Amanabi M, et al. Completeness of METADATA Reporting in AI Dental Research: Scoping Review Protocol. 2024 [cited 2025 Jan 22]; Available from: https://osf.io/jxmsf/ Schwendicke F, Uribe SE, Issa J. Quality and Completeness of Reporting in Dental AI Research. 2024 [cited 2025 Jan 22]; Available from: https://osf.io/f2uzy/ Jakubovics NS, Schwendicke F. Toward Better Reporting in Oral Health Research. J Dent Res. 2024 Oct;103(11):1045–6. Liu J, Peng G, Yan S. An Intelligent Evaluation Method of Root Canal Therapy Quality Based on Deep Learning. In Institute of Electrical and Electronics Engineers Inc.; 2022. p. 6254–9. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85151151347&doi=10.1109%2fCAC57257.2022.10056075&partnerID=40&md5=afa692730e693b9ee522c7d6a8a5a8a2 Table Table 1 is available in the Supplementary Files section Additional Declarations No competing interests reported. Supplementary Files Table.docx Supplementarymaterial1SearchStrategy.docx Supplementarymaterial2PRISMAScRChecklist.docx Cite Share Download PDF Status: Posted Version 1 posted 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-5952801","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Systematic Review","associatedPublications":[],"authors":[{"id":413071279,"identity":"9d008b4f-6842-44d4-9911-129c83aaf620","order_by":0,"name":"Maryam Ghiasi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAAtklEQVRIiWNgGAWjYFACHoYDDAwScgwSDCAGCVqMSdMCAokNEsQ6S7d/7cEDP/5YpPdL9z48wFBjR1iL2Y13CQd72yRyZ845bnCA4VgyMVrOGBzgbZDI3XAjjeEAYwMzcVoO/vkjkW4A0VJPhJbzPQaHedgkEqBaDhNjC4/BYdk2CcOZM4BaEo4dJ8aWM8Yf3/ypk+eXSGP+8KGmmrAWBokEJE4CDkWogP8AUcpGwSgYBaNgJAMAikY+a4Hr7IgAAAAASUVORK5CYII=","orcid":"","institution":"Ottawa Hospital Research Institute","correspondingAuthor":true,"prefix":"","firstName":"Maryam","middleName":"","lastName":"Ghiasi","suffix":""},{"id":413071280,"identity":"1c05bc9a-3392-47a3-958c-b8068945c78b","order_by":1,"name":"Ava Zaboli","email":"","orcid":"","institution":"UCLA School of Dentistry","correspondingAuthor":false,"prefix":"","firstName":"Ava","middleName":"","lastName":"Zaboli","suffix":""},{"id":413071281,"identity":"8726a564-566a-43ca-a728-83583da34529","order_by":2,"name":"Mina Mahdian","email":"","orcid":"","institution":"Stony Brook University","correspondingAuthor":false,"prefix":"","firstName":"Mina","middleName":"","lastName":"Mahdian","suffix":""},{"id":413071282,"identity":"3c958c53-5292-4b80-b390-b76cbcd7f244","order_by":3,"name":"Sanjay M. Mallya","email":"","orcid":"","institution":"UCLA School of Dentistry","correspondingAuthor":false,"prefix":"","firstName":"Sanjay","middleName":"M.","lastName":"Mallya","suffix":""}],"badges":[],"createdAt":"2025-02-03 17:08:23","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5952801/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5952801/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":75995477,"identity":"3df61158-296e-4f30-998f-fe16470c84a0","added_by":"auto","created_at":"2025-02-11 09:51:32","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":228986,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/6af461dce72512547b25e24c.png"},{"id":75995486,"identity":"876ce37f-8884-4168-b627-27939249bd40","added_by":"auto","created_at":"2025-02-11 09:51:32","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":35235,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"2.png","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/00f5b9c300631018cf3cf001.png"},{"id":75997712,"identity":"f5274bef-96e2-4521-a7cb-f415c767781d","added_by":"auto","created_at":"2025-02-11 10:07:32","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":45888,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"3.png","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/691da36e74b95e77919c7a51.png"},{"id":75997111,"identity":"83335cb7-3774-4cc2-8753-e169c49c7c30","added_by":"auto","created_at":"2025-02-11 09:59:32","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":41920,"visible":true,"origin":"","legend":"\u003cp\u003eLegend not included with this version.\u003c/p\u003e","description":"","filename":"4.png","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/51df8aa0f4d7e319bc54c24f.png"},{"id":75997713,"identity":"bac012b4-a8e3-451b-88d3-f4e06a736afa","added_by":"auto","created_at":"2025-02-11 10:07:37","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":771875,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/8ebc7cdf-694d-4cbb-8d97-26accbd9ef68.pdf"},{"id":75995468,"identity":"f46c9f0b-a5db-4c58-8ba6-b1235c90c2dc","added_by":"auto","created_at":"2025-02-11 09:51:32","extension":"docx","order_by":1,"title":"","display":"","copyAsset":false,"role":"supplement","size":78631,"visible":true,"origin":"","legend":"","description":"","filename":"Table.docx","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/fd676ac20ec08a60857761e0.docx"},{"id":75995470,"identity":"4346b8e3-54d8-423b-922d-43bbebbd7921","added_by":"auto","created_at":"2025-02-11 09:51:32","extension":"docx","order_by":2,"title":"","display":"","copyAsset":false,"role":"supplement","size":17816,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial1SearchStrategy.docx","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/db663396eada77e46eea79ae.docx"},{"id":75997114,"identity":"691a3e91-4a91-407a-ad6a-12dc665e6b1f","added_by":"auto","created_at":"2025-02-11 09:59:32","extension":"docx","order_by":3,"title":"","display":"","copyAsset":false,"role":"supplement","size":86643,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementarymaterial2PRISMAScRChecklist.docx","url":"https://assets-eu.researchsquare.com/files/rs-5952801/v1/d5b8fcc06a99b8b127c8980d.docx"}],"financialInterests":"No competing interests reported.","formattedTitle":"Artificial Intelligence for Root Canal Segmentation on Radiographic Images: A Scoping Review","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eThe number, size, patency, and location of pulp canals are critical in endodontic treatment planning. Accurate knowledge of these parameters is essential to ensure successful endodontic outcomes, as the inability to locate, debride and disinfect all canals is a common cause of endodontic failure(1). Variations in canal anatomy, which can be influenced by factors such as age, ethnicity, and root anatomy, further complicate this task(2,3). A comprehensive three-dimensional understanding of pulp morphology is essential for successful endodontic treatment outcomes(4). Traditional methods for pulp canal identification and segmentation involve manual or semi-automated approaches, both of which are time-consuming and labor-intensive. Manual segmentation, in particular, requires extensive user interaction and can take considerable time, depending on the complexity of the cases(5).\u003c/p\u003e \u003cp\u003eArtificial intelligence (AI) offers a promising solution to these challenges. AI consists of a variety of technologies that emulate intelligent human behavior through experiential learning(6). Deep learning (DL), a subset of AI, uses neural networks with multiple hidden layers that function similarly to biological neurons, learning by identifying features from input data(7). Convolutional neural networks (CNNs), a subset of neural networks, have been particularly successful in processing complex data structures like images(8). In healthcare, AI and DL are increasingly being used to enhance disease prediction, diagnosis, and management(7,9,10).\u003c/p\u003e \u003cp\u003eIn dentistry, AI has shown potential in various applications, including the detection of caries, periapical lesions, cracks, root fractures, and the determination of working length and tooth labeling(11\u0026ndash;15). However, the application of AI in endodontics, particularly for root canal segmentation, remains relatively underexplored. Existing reviews have had limited scope(16). AI-driven approaches can potentially automate the segmentation of root canals from radiographic images, providing efficiency, consistency, and accuracy.\u003c/p\u003e \u003cp\u003eGiven the rapidly evolving nature of AI technology and its growing application in dental practice, it is timely to conduct a comprehensive scoping review to map the literature on AI-driven solutions for automated root canal segmentation.\u003c/p\u003e \u003cp\u003eThis scoping review aims to map the existing literature on the use of AI to automate root canal segmentation on radiographic images. It provides an updated overview of the current state of research in this area, highlights the gaps in existing research, and propose future research directions, emphasizing the potential of AI to transform endodontic practice and enhance treatment results.\u003c/p\u003e"},{"header":"2. Materials and Methods","content":"\u003cp\u003eThis scoping review was carried out using a predefined protocol in accordance with the PRISMA Extension for Scoping Reviews guidelines(17,18). Protocol development and review conduct were guided by best practice principles outlined by the Cochrane Handbook for Systematic Reviews of Interventions. There were no deviations from the protocol. Our approach followed the five-step framework originally developed by Arksey and O\u0026rsquo;Malley(19), and later refined by Levac et al(20). The framework involves: (1) formulating the research question, (2) identifying relevant studies, (3) selecting the studies, (4) charting the data, and (5) collating, summarizing, and reporting the findings. Due to the novelty and limited research available on this topic, resulting in a heterogeneous body of literature, the scoping review method was deemed the most suitable. This study did not require ethical approval.\u003c/p\u003e\n\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\n \u003ch2\u003e2.1 Information sources and search strategy\u003c/h2\u003e\n \u003cp\u003eAn experienced medical information specialist developed and refined the search strategies through an iterative process in collaboration with the main reviewer. The MEDLINE (Ovid), Embase, Scopus, and Web of Science were searched for all studies published since the database\u0026apos;s inception up to January 8, 2025. These four health sciences major electronic databases were comprehensively searched using Medical Subject Headings (MeSH), keywords, truncations, adjacency functions, and Boolean operators. Additionally, Google Scholar was searched using the same keywords applied in the above-mentioned electronic databases to find further relevant sources of gray literature. The complete search strategy used to locate articles in electronic databases is detailed in Supplementary material 1. Following the electronic database search, the reference lists of the included articles were manually searched for relevant articles.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\n \u003ch2\u003e2.2 Eligibility criteria\u003c/h2\u003e\n \u003cp\u003eStudies were included in the review if they investigated or reported on application of artificial intelligence for dental canal segmentation. Studies that did not meet the inclusion criteria were excluded from this review.\u003c/p\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eInclusion criteria:\u003c/p\u003e\n \u003c/div\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eAll types of radiographic images including cone-beam computed tomography (CBCT), panoramic, periapical, and bitewing are included.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eStudy selection is not limited to any specific study design; observational and interventional studies are included.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eThere is no restriction on the date of publication or language of the studies.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eThere is no limitation on the type of teeth investigated.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eThere is no limitation on the status of the obturation of the root canals (i.e., both obturated and unobturated canals are included).\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n \u003cdiv class=\"BlockQuote\"\u003e\n \u003cp\u003eExclusion criteria:\u003c/p\u003e\n \u003c/div\u003e\n \u003cul\u003e\n \u003cli\u003e\n \u003cp\u003eStudies not utilizing artificial Intelligence are excluded.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eStudies not using radiographic images to assess the root canals are excluded.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eStudies not focusing on pulp canal segmentation are excluded (e.g., studies reporting on tooth segmentation without focusing on root canals are excluded).\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eStudies not having canal segmentation as their objective with no details on the segmentation methods and results are excluded.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eCommentaries, retracted articles, and inaccessible full-text articles are excluded.\u003c/p\u003e\n \u003c/li\u003e\n \u003cli\u003e\n \u003cp\u003eReview articles are excluded if they do not contain relevant studies that are not already included in our review.\u003c/p\u003e\n \u003c/li\u003e\n \u003c/ul\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\n \u003ch2\u003e2.3 Study selection\u003c/h2\u003e\n \u003cp\u003eStudy selection was performed independently by two reviewers (MG and AZ). Initially, titles and abstracts were screened to remove irrelevant studies (stage 1) using the Covidence web application(21). For potentially relevant studies, full texts were retrieved and comprehensively assessed to make sure they were compliant with the eligibility criteria (stage 2). References from all included publications that were identified as potentially relevant or which the reviewers were unclear of its inclusion were also considered for full-text screening. When there was uncertainty, a third-reviewer with expertise in subject matter was consulted. A pilot-test screening was done for 20 studies before the main screening in each stage to train and calibrate the reviewers to the eligibility criteria and ensure consistent application of the criteria. Reasons for the exclusion of each study were documented.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec6\" class=\"Section2\"\u003e\n \u003ch2\u003e2.4 Charting the data\u003c/h2\u003e\n \u003cp\u003eData extraction was independently performed by two reviewers (MG and AZ) using the Covidence web application a pre-developed extraction form. Any disagreements were resolved through discussion and consensus, or by involving a third reviewer if necessary. The data mapping process involved two main steps. First, the data was organized by examining the included studies, classifying them, and sorting them according to the research objectives. Second, an explanatory summary of the data was generated to derive insights. Reviewers thoroughly reviewed and charted all included full-text articles, with the data being summarized narratively and compiled into tables. Extracted data consisted of the title, authorship, year of publication, country of origin, study design, imaging modalities, obturation status, type of teeth analyzed, AI models used, and results.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec8\" class=\"Section2\"\u003e\n \u003ch2\u003e3.1 Study selection\u003c/h2\u003e\n \u003cp\u003eOur electronic search initially identified 836 articles. After examining the titles and abstracts, 79 articles were selected for a full-text review. Of these, 13 were excluded for not investigating root canal/pulp specifically, 14 were excluded for not performing segmentation, and 6 were excluded as canal segmentation was not the objective of the studies without providing any details on the segmentation methods and results. Additionally, 8 review articles were excluded because their references did not include any new eligible study not already covered in our review. Four studies were excluded for the following reasons: not using AI in their methods, being retracted, being a commentary, and the full-text not being found. This left 35 studies that met the eligibility criteria and were included in the final synthesis. Data extraction was then performed on these eligible articles, and the results are presented in this paper. The screening and selection process of the studies is illustrated in Fig. \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e, adhering to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) flowchart.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e\n \u003ch2\u003e3.2 Characteristics of included studies\u003c/h2\u003e\n \u003cp\u003eThe included studies were published between 2011 and 2024, with one study from 2011(22), another from 2012(23), and the remaining studies published from 2019 to 2024. Most of the studies originated from China (26%), followed by Turkey (17%) and the United States (11%). Figure \u003cspan class=\"InternalRef\"\u003e2\u003c/span\u003e provides a detailed breakdown of the number of studies by country of origin.\u003c/p\u003e\n \u003cp\u003eThe characteristics of the 35 included articles are summarized in Table \u003cspan class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec10\" class=\"Section2\"\u003e\n \u003ch2\u003e3.3 Content analysis\u003c/h2\u003e\n \u003cp\u003eTwo studies were conducted prospectively and ex vivo(4,24), while the remaining 33 were conducted retrospectively in vivo. Most of the studies used CBCT images (51%) to train their AI models, followed by panoramic (17%) and periapical images (14%). Figure \u003cspan class=\"InternalRef\"\u003e3\u003c/span\u003e illustrates the distribution of studies based on the diagnostic techniques used for AI model training.\u003c/p\u003e\n \u003cp\u003eRegarding the obturation status of the teeth used for training, 34% of the studies did not specify whether they used obturated or unobturated canals, 23% used obturated canals, 26% used unobturated canals, and 17% used both types. Most of the studies did not specify the type of teeth used for AI model training. Figure 4 presents the number of studies categorized by each tooth type.\u003c/p\u003e\n \u003cp\u003eTwo older studies, published in 2011 and 2012, employed histogram-based AI models: Histogram-based Quick Fuzzy C-means Clustering(22) and a Histogram-based Enhanced Version(23). The remaining 33 studies utilized CNNs. Regarding the ground truth, 7 studies did not specify their ground truth sources. Three studies used micro-CT(4,24,25), while the remaining studies relied on dental experts\u0026mdash;including dentists, endodontists, and dentomaxillofacial radiologists\u0026mdash;for clinical confirmation or manual segmentation of the root canals.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eOutcome metrics\u003c/strong\u003e\u003c/p\u003e\n \u003cp\u003eThe performance of the AI models for canal segmentation was primarily evaluated using the Dice Similarity Coefficient (DSC), also known as the F1 score, which ranged from 0.6 to 0.98 ( 22 studies(4,5,24\u0026ndash;43)). The majority of the studies reported several metrics. Sensitivity values ranged from 0.72 to 1 (11 studies(26,28,29,31,35,39,40,44\u0026ndash;47)), while specificity values varied from 0.23 to 1 (7 studies(26,28,40,44\u0026ndash;47)). Accuracy was reported between 0.73 and 0.99 (15 studies(23,26,28,38,40,42\u0026ndash;44,46\u0026ndash;52)), and precision ranged from 0.66 to 1 (12 studies(27,31,32,35,36,39,40,42,46,47,49,53)). The area under the curve (AUC) spanned from 0.58 to 0.99 (7 studies(26\u0026ndash;28,31,40,46,50)). Less frequent metrics were also reported including Positive Predicte Value(29,44), Negative Predictive Value(44), Hausdorff distance (HD)(25,33,53), and inter-over union (IoU)(31,41\u0026ndash;43).\u003c/p\u003e\n \u003cp\u003eLin et al.(25) reported that only thick root canals could be segmented, unlike the thin root canals, such as those in the apical third and lateral root canals. Lee et al.(40) showed that certain features, including canal visibility, were less distinguishable. Szil\u0026aacute;gyi et al.(22) indicated that while their algorithm performed automatically in most cases, some decisions required interactive input. Wang et al.(33) demonstrated that their AI workflow reduced the time needed to obtain a 3D model of the tooth and root canal from 6 hours to 2 minutes, resulting in satisfactory outcomes for complex root canal treatments. Similarly, Song et al(5), showed that automatic segmentation took 13\u0026ndash;33 s while manual segmentation needs 15\u0026ndash;25 min. Schneider et al.(54) found that deeper, more complex models did not necessarily outperform less complex alternatives. Rohrer et al.(37) used a tiling approach on panoramic images, resulting in a performance improvement of +\u0026thinsp;294%; the mean F1-score increased from 0.33 to 0.9. Gardiyanoglu et al.(38) reported a fair DSC value of 0.78 for successful root canal treatments but noted that cases with inadequate root canal fillings and gutta-perchas superimposed with neighboring teeth\u0026apos;s roots caused relatively lower DSC values. Orhan et al.(45) found a sensitivity of 82.1% for root canal fillings, but only 23.0% for the assessment of voids in the root canal. Four studies investigated c-shaped canal segmentation in mandibular second molars(29,46,47,50). One of them(47) showed that the neural network\u0026apos;s classification performance, when augmented with GAN-synthesized periapical images, improved compared to using real data alone. Hasan et al.(52) demonstrated that imbalanced datasets with noise removal led to YOLOv5x\u0026apos;s prediction accuracy dropping to 72%, while balancing and noise removal led to all models (YOLOv5s, YOLOv5x, and YOLOv7) performing at over 95% accuracy. Mean average precision (mAP) improved from 52\u0026ndash;92% following balancing and denoising.\u003c/p\u003e\n\u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eThis study is the most comprehensive review of AI models for root canal segmentation conducted to date, encompassing a wide range of studies and providing a detailed understanding of the current landscape in this field. Our analysis of studies published between 2011 and 2024 highlights significant progress and ongoing challenges in the development and application of AI in endodontics.\u003c/p\u003e\n\u003cp\u003eIt is also noteworthy that the majority of studies in this field have originated from China (26%), followed by Turkey (17%) and the United States (11%). Several factors might explain why China leads in publications, including strong government support for research and development, a large domestic market that drives demand for advanced dental technologies, widespread availability of data, effective collaboration between industry and academia, and their investment in education and research infrastructure(55). Understanding the regional strengths and focus areas can help in fostering international collaborations and addressing any gaps in research coverage. However, the development of deep learning algorithms trained on regional morphology may impact the global applicability of such tools.\u003c/p\u003e\n\u003cp\u003eBaydar et al.(35) and Ari et al.(39) both reported on the effectiveness of these models in both unobturated and obturated canals separately. Performance metrics such as F1 scores, sensitivity, and precision were higher for obturated canals, which may suggest that the AI models are particularly adept at identifying high contrast objects such root canal fillings. This should be further investigated to confirm that the observed differences are not due to chance.\u003c/p\u003e\n\u003cp\u003eLee et al.(40) reported low AUC (0.58) and specificity (0.23), which were attributed to the difficulty in distinguishing less visible canal features. This suggests that their DCNN model has room for improvement, and refining machine learning algorithms and incorporating advanced image processing in future studies could enhance both feature detection and endodontic prognostication accuracy.\u003c/p\u003e\n\u003cp\u003eA notable gap in the literature is the lack of consideration for patient age. None of the studies accounted for the impact of age on AI performance, despite the known tendency for dental canals to become more calcified with age, potentially affecting model accuracy. Future research should investigate this variable to ensure the robustness of AI models across different age groups.\u003c/p\u003e\n\u003cp\u003eThe majority of studies did not specify the type of teeth used for training, implying that their datasets included various tooth types and conditions. This generalizability is beneficial for developing AI models applicable to a wide range of clinical scenarios. However, focusing on specific tooth types could also allow for more targeted model development and potentially higher accuracy for application in those particular tooth types.\u003c/p\u003e\n\u003cp\u003eAdditionally, while some studies demonstrated the potential of AI to significantly reduce the time required for clinical tasks, such as Wang et al.\u0026apos;s reduction from 6 hours to 2 minutes for 3D modeling(33) and Song et al.\u0026rsquo;s reduction from 15\u0026ndash;20 minutes to 13\u0026ndash;33 seconds for segmentation (manual versus automatic), it is essential to balance speed with accuracy. The findings by Schneider et al.(54) that more complex models do not always yield better results emphasize the importance of optimizing model complexity.\u003c/p\u003e\n\u003cp\u003eFuture studies should strive for more standardized and detailed methodologies, along with greater transparency in their reporting. Standardized reporting would allow for a more homogeneous set of studies, facilitating a more comprehensive analysis of various features, such as different CNN models, obturated versus unobturated canals, and the types of teeth used in the datasets. By establishing uniform criteria for data collection, annotation, and evaluation metrics, researchers can ensure that studies are comparable and replicable, which is critical for validating AI models and their clinical utility(56\u0026ndash;58).\u003c/p\u003e\n\u003cp\u003eMoreover, detailed reporting should include comprehensive descriptions of the datasets, including patient demographics, imaging modalities, and the ground truth used for training and validation. This level of detail will be essential in identifying the factors that contribute to the success or failure of AI models, enabling more accurate interpretation of results.\u003c/p\u003e\n\u003cp\u003eIn addition, future research should consider stratifying results based on clinically relevant subgroups, such as age, sex, and specific dental conditions, to assess the robustness and generalizability of AI models across diverse patient populations. This stratification could reveal insights into how different variables impact model performance and help in developing more tailored AI solutions for various clinical scenarios.\u003c/p\u003e\n\u003cp\u003eCollaborations among researchers, clinicians, computer science experts, and institutions can further enhance the quality of AI research in endodontics. Collaborative efforts can lead to larger, more diverse datasets and the sharing of best practices, ultimately accelerating the development and implementation of AI technologies in dental practice.\u003c/p\u003e\n\u003cp\u003eA key strength of this review is its comprehensive scope, ensuring that all published studies using AI for root canal segmentation are included, thus providing a detailed map of the current literature. However, a significant limitation is the heterogeneity in the parameters and metrics reported across the studies, making it impossible to perform a full systematic review or meta-analysis. This variability underscores the need for standardized reporting and methodological approaches in future research. In addition, in line with established methods for scoping reviews, we did not conduct a risk of bias assessment of the included studies; this could be done in a future, full systematic review.\u003c/p\u003e\n\u003cp\u003eWhile the AI models reviewed show great potential for improving endodontic treatment planning and outcomes, the results should be interpreted with caution due to certain limitations. Some studies lacked detailed methodology reporting, including their ground truth sources, which can affect the reliability of the findings. Additionally, as this is a scoping review aimed at mapping the literature, we did not perform a Risk of Bias analysis on the studies, further emphasizing the need for cautious interpretation of the results for clinical practice. The use of AI in root canal segmentation and endodontics is still in its initial stages, indicating significant room for refinement and improvement of the models. Future studies should address these gaps and standardize methodologies to enhance the reliability and clinical utility of AI in dental practice.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData Availability Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll data supporting the findings of this study are included in this article. As this study is a scoping review, no new datasets were generated.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthics approval is not required for this scoping review as the review uses solely published literature.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eRelevant reporting guideline\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis scoping review was carried out using a predefined protocol in accordance with the PRISMA Extension for Scoping Reviews guidelines(17,18). The document is included in \u003cstrong\u003eSupplementary material 2.\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics and Integrity Statement\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis research adheres to rigorous ethical standards and integrity policies, ensuring transparency, accountability, and respect for all participants. Upholding these standards is essential for the credibility and reliability of our findings.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors would like to thank Risa Shorr (Learning Services, The Ottawa Hospital) for assisting in the creation of the literature search.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n \u003cli\u003eTabassum S, Khan FR. Failure of endodontic treatment: The usual suspects. Eur J Dent. 2016;10(1):144\u0026ndash;7.\u003c/li\u003e\n \u003cli\u003eMartins JNR, Alkhawas MBAM, Altaki Z, Bellardini G, Berti L, Boveda C, et al. Worldwide Analyses of Maxillary First Molar Second Mesiobuccal Prevalence: A Multicenter Cone-beam Computed Tomographic Study. J Endod. 2018 Nov;44(11):1641-1649.e1.\u003c/li\u003e\n \u003cli\u003ePattanshetti N, Gaidhane M, Al Kandari AM. Root and canal morphology of the mesiobuccal and distal roots of permanent first molars in a Kuwait population--a clinical study. Int Endod J. 2008 Sep;41(9):755\u0026ndash;62.\u003c/li\u003e\n \u003cli\u003eLin X, Fu Y, Ren G, Yang X, Duan W, Chen Y, et al. Micro-Computed Tomography-Guided Artificial Intelligence for Pulp Cavity and Tooth Segmentation on Cone-beam Computed Tomography. J Endod. 2021 Dec;47(12):1933\u0026ndash;41.\u003c/li\u003e\n \u003cli\u003eSong Y, Yang H, Ge Z, Du H, Li G. Age estimation based on 3D pulp segmentation of first molars from CBCT images using U-Net. Dento Maxillo Facial Radiol. 2023 Oct;52(7):20230177.\u003c/li\u003e\n \u003cli\u003eDavenport T, Kalakota R. The potential for artificial intelligence in healthcare. Future Healthc J. 2019 Jun;6(2):94\u0026ndash;8.\u003c/li\u003e\n \u003cli\u003eChing T, Himmelstein DS, Beaulieu-Jones BK, Kalinin AA, Do BT, Way GP, et al. Opportunities and obstacles for deep learning in biology and medicine. J R Soc Interface. 2018 Apr;15(141):20170387.\u003c/li\u003e\n \u003cli\u003eSchwendicke F, Samek W, Krois J. Artificial Intelligence in Dentistry: Chances and Challenges. J Dent Res. 2020 Jul;99(7):769\u0026ndash;74.\u003c/li\u003e\n \u003cli\u003ePark WJ, Park JB. History and application of artificial neural networks in dentistry. Eur J Dent. 2018;12(4):594\u0026ndash;601.\u003c/li\u003e\n \u003cli\u003eCheng PM, Montagnon E, Yamashita R, Pan I, Cadrin-Ch\u0026ecirc;nevert A, Perdig\u0026oacute;n Romero F, et al. Deep Learning: An Update for Radiologists. Radiogr Rev Publ Radiol Soc N Am Inc. 2021;41(5):1427\u0026ndash;45.\u003c/li\u003e\n \u003cli\u003eMusri N, Christie B, Ichwan SJA, Cahyanto A. Deep learning convolutional neural network algorithms for the early detection and diagnosis of dental caries on periapical radiographs: A systematic review. Imaging Sci Dent. 2021 Sep;51(3):237\u0026ndash;42.\u003c/li\u003e\n \u003cli\u003eUmer F, Habib S. Critical Analysis of Artificial Intelligence in Endodontics: A Scoping Review. J Endod. 2022;48(2):152\u0026ndash;60.\u003c/li\u003e\n \u003cli\u003eAsiri A.F., Altuwalah A.S. The role of neural artificial intelligence for diagnosis and treatment planning in endodontics: A qualitative review. Saudi Dent J. 2022;34(4):270\u0026ndash;81.\u003c/li\u003e\n \u003cli\u003eAhmed ZH, Almuharib AM, Abdulkarim AA, Alhassoon AH, Alanazi AF, Alhaqbani MA, et al. Artificial Intelligence and Its Application in Endodontics: A Review. J Contemp Dent Pract. 2023;24(11):912\u0026ndash;7.\u003c/li\u003e\n \u003cli\u003eKhanagar SB, Alfadley A, Alfouzan K, Awawdeh M, Alaqla A, Jamleh A. Developments and Performance of Artificial Intelligence Models Designed for Application in Endodontics: A Systematic Review. Diagn Basel Switz. 2023;13(3).\u003c/li\u003e\n \u003cli\u003eAlfadley A, Shujaat S, Jamleh A, Riaz M, Aboalela AA, Ma H, et al. Progress of Artificial Intelligence-Driven Solutions for Automated Segmentation of Dental Pulp Space on Cone-Beam Computed Tomography Images. A Systematic Review. J Endod. 2024 May;S0099239924003364.\u003c/li\u003e\n \u003cli\u003eKhalil H, Peters MD, Tricco AC, Pollock D, Alexander L, McInerney P, et al. Conducting high quality scoping reviews-challenges and solutions. J Clin Epidemiol. 2021 Feb;130:156\u0026ndash;60.\u003c/li\u003e\n \u003cli\u003ePeters MDJ, Marnie C, Tricco AC, Pollock D, Munn Z, Alexander L, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evid Synth. 2020 Oct;18(10):2119\u0026ndash;26.\u003c/li\u003e\n \u003cli\u003eArksey H, O\u0026rsquo;Malley L. Scoping studies: towards a methodological framework. Int J Soc Res Methodol. 2005 Feb;8(1):19\u0026ndash;32.\u003c/li\u003e\n \u003cli\u003eLevac D, Colquhoun H, O\u0026rsquo;Brien KK. Scoping studies: advancing the methodology. Implement Sci. 2010 Dec;5(1):69.\u003c/li\u003e\n \u003cli\u003eCovidence systematic review software [Internet]. Melbourne, Australia: Veritas HealthInnovation; [cited 2024 Mar 1]. Available from: www.covidence.org\u003c/li\u003e\n \u003cli\u003eSzil\u0026aacute;gyi L, Dob\u0026oacute;-Nagy C, Beny\u0026oacute; B. Identification of the root canal from dental micro-CT records. In 2011. p. 339\u0026ndash;46. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-81855226116\u0026amp;doi=10.1007%2f978-3-642-25085-9_40\u0026amp;partnerID=40\u0026amp;md5=4318c1e0537e79cea3271b4510b9f05c\u003c/li\u003e\n \u003cli\u003eBenyo B. Identification of dental root canals and their medial line from micro-CT and cone-beam CT records. Biomed Eng Online. 2012;11((Benyo) Budapest University of Technology and Economics, Department of Control Engineering and Information Technology, Hungary.):81.\u003c/li\u003e\n \u003cli\u003eHatvani J, Horvath A, Michetti J, Basarab A, Kouame D, Gyongy M. Deep learning-based super-resolution applied to dental computed tomography. IEEE Trans Radiat Plasma Med Sci. 2019;3(2):120\u0026ndash;8.\u003c/li\u003e\n \u003cli\u003eLin X, Fu YJ, Ren GQ, Wen JH, Chen YF, Zhang Q. [Segmentation and accuracy validation of mandibular molar and pulp cavity on cone-beam CT images by U-net neural network]. Shanghai Kou Qiang Yi Xue Shanghai J Stomatol. 2022;31(5):454\u0026ndash;9.\u003c/li\u003e\n \u003cli\u003eDumont M, Prieto JC, Brosset S, Cevidanes L, Bianchi J, Ruellas A, et al. Patient Specific Classification of Dental Root Canal and Crown Shape. Shape Med Imaging Int Workshop ShapeMI 2020 Held Conjunction MICCAI 2020 Lima Peru Oct 4 2020 Proc. 2020;12474(101773457):145\u0026ndash;53.\u003c/li\u003e\n \u003cli\u003eDeleat-Besson R, Le C, Zhang W, Turkestani NA, Cevidanes L, Bianchi J, et al. Merging and Annotating Teeth and Roots from Automated Segmentation of Multimodal Images. In Springer Science and Business Media Deutschland GmbH; 2021. p. 81\u0026ndash;92. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85118156750\u0026amp;doi=10.1007%2f978-3-030-89847-2_8\u0026amp;partnerID=40\u0026amp;md5=9a4c5b428267988b9b1aac6d2a797dcd\u003c/li\u003e\n \u003cli\u003eDeleat-Besson R, Le C, Al Turkestani N, Zhang W, Dumont M, Brosset S, et al. Automatic Segmentation of Dental Root Canal and Merging with Crown Shape. Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Annu Int Conf. 2021;2021(101763872):2948\u0026ndash;51.\u003c/li\u003e\n \u003cli\u003eSherwood AA, Sherwood AI, Setzer FC, K SD, Shamili JV, John C, et al. A Deep Learning Approach to Segment and Classify C-Shaped Canal Morphologies in Mandibular Second Molars Using Cone-beam Computed Tomography. J Endod. 2021;47(12):1907\u0026ndash;16.\u003c/li\u003e\n \u003cli\u003eZhang J, Xia W, Dong J, Tang Z, Zhao Q. Root Canal Segmentation in CBCT Images by 3D U-Net with Global and Local Combination Loss. Annu Int Conf IEEE Eng Med Biol Soc IEEE Eng Med Biol Soc Annu Int Conf. 2021;2021(101763872):3097\u0026ndash;100.\u003c/li\u003e\n \u003cli\u003eDuman SB, Celik Ozen D, Bayrakdar IS, Baydar O, Alhaija ESA, Helvacioglu Yigit D, et al. Second mesiobuccal canal segmentation with YOLOv5 architecture using cone beam computed tomography images. Odontology. 2023;(101134822).\u003c/li\u003e\n \u003cli\u003eKim SH, Song IS, Baek SJ. Automatic Segmentation of Internal Tooth Structure from CBCT Images Using Hierarchical Deep Learning. In Springer Science and Business Media Deutschland GmbH; 2023. p. 703\u0026ndash;13. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85174730962\u0026amp;doi=10.1007%2f978-3-031-43898-1_67\u0026amp;partnerID=40\u0026amp;md5=e03879beefbd8314dc165d7f73d05b34\u003c/li\u003e\n \u003cli\u003eWang Y, Xia W, Yan Z, Zhao L, Bian X, Liu C, et al. Root canal treatment planning by automatic tooth and root canal segmentation in dental CBCT with deep multi-task feature learning. Med Image Anal. 2023;85(c8s, 9713490):102750.\u003c/li\u003e\n \u003cli\u003eTan M, Cui Z, Zhong T, Fang Y, Zhang Y, Shen D. A progressive framework for tooth and substructure segmentation from cone-beam CT images. Comput Biol Med. 2024;169(doc, 1250250):107839.\u003c/li\u003e\n \u003cli\u003eBaydar O., Rozylo-Kalinowska I., Futyma-Gabka K., Saglam H. The U-Net Approaches to Evaluation of Dental Bite-Wing Radiographs: An Artificial Intelligence Study. Diagnostics. 2023;13(3):453.\u003c/li\u003e\n \u003cli\u003eVinayahalingam S, Goey RS, Kempers S, Schoep J, Cherici T, Moin DA, et al. Automated chart filing on panoramic radiographs using deep learning. J Dent. 2021;115(hx1, 0354422):103864.\u003c/li\u003e\n \u003cli\u003eRohrer C, Krois J, Patel J, Meyer-Lueckel H, Rodrigues JA, Schwendicke F. Segmentation of Dental Restorations on Panoramic Radiographs Using Deep Learning. Diagn Basel Switz. 2022;12(6).\u003c/li\u003e\n \u003cli\u003eGardiyanoglu E, Unsal G, Akkaya N, Aksoy S, Orhan K. Automatic Segmentation of Teeth, Crown-Bridge Restorations, Dental Implants, Restorative Fillings, Dental Caries, Residual Roots, and Root Canal Fillings on Orthopantomographs: Convenience and Pitfalls. Diagn Basel Switz. 2023;13(8).\u003c/li\u003e\n \u003cli\u003eAri T, Saglam H, Oksuzoglu H, Kazan O, Bayrakdar IS, Duman SB, et al. Automatic Feature Segmentation in Dental Periapical Radiographs. Diagn Basel Switz. 2022;12(12).\u003c/li\u003e\n \u003cli\u003eLee J, Seo H, Choi YJ, Lee C, Kim S, Lee YS, et al. An Endodontic Forecasting Model Based on the Analysis of Preoperative Dental Radiographs: A Pilot Study on an Endodontic Predictive Deep Neural Network. J Endod. 2023;49(6):710\u0026ndash;9.\u003c/li\u003e\n \u003cli\u003eGamal M, Baraka M, Torki M. Automatic Mandibular Semantic Segmentation of Teeth Pulp Cavity and Root Canals, and Inferior Alveolar Nerve on Pulpy3D Dataset. In: Linguraru MG, Dou Q, Feragen A, Giannarou S, Glocker B, Lekadir K, et al., editors. Medical Image Computing and Computer Assisted Intervention \u0026ndash; MICCAI 2024 [Internet]. Cham: Springer Nature Switzerland; 2024 [cited 2025 Jan 11]. p. 14\u0026ndash;23. (Lecture Notes in Computer Science; vol. 15008). Available from: https://link.springer.com/10.1007/978-3-031-72111-3_2\u003c/li\u003e\n \u003cli\u003eSlim ML, Jacobs R, De Souza Leal RM, Fontenele RC. AI-driven segmentation of the pulp cavity system in mandibular molars on CBCT images using convolutional neural networks. Clin Oral Investig. 2024 Nov 21;28(12):650.\u003c/li\u003e\n \u003cli\u003e\u0026Ccedil;elik B, Gen\u0026ccedil; MZ, \u0026Ccedil;elik ME. Evaluation of root canal filling length on periapical radiograph using artificial intelligence. Oral Radiol. 2025 Jan;41(1):102\u0026ndash;10.\u003c/li\u003e\n \u003cli\u003eAlbitar L., Zhao T., Huang C., Mahdian M. Artificial Intelligence (AI) for Detection and Localization of Unobturated Second Mesial Buccal (MB2) Canals in Cone-Beam Computed Tomography (CBCT). Diagnostics. 2022;12(12):3214.\u003c/li\u003e\n \u003cli\u003eOrhan K, Aktuna Belgin C, Manulis D, Golitsyna M, Bayrak S, Aksoy S, et al. Determining the reliability of diagnosis and treatment using artificial intelligence software with panoramic radiographs. Imaging Sci Dent. 2023;53(3):199\u0026ndash;208.\u003c/li\u003e\n \u003cli\u003eYang S, Kim KD, Kise Y, Nozawa M, Mori M, Takata N, et al. External Validation of the Effect of the Combined Use of Object Detection for the Classification of the C-Shaped Canal Configuration of the Mandibular Second Molar in Panoramic Radiographs: A Multicenter Study. J Endod. 2024;(i1k, 7511484, 17910050r).\u003c/li\u003e\n \u003cli\u003eYang S, Kim KD, Ariji E, Takata N, Kise Y. Evaluating the performance of generative adversarial network-synthesized periapical images in classifying C-shaped root canals. Sci Rep. 2023;13(1):18038.\u003c/li\u003e\n \u003cli\u003eOrhan K, Bilgir E, Bayrakdar IS, Ezhov M, Gusarev M, Shumilov E. Evaluation of artificial intelligence for detecting impacted third molars on cone-beam computed tomography scans. J Stomatol Oral Maxillofac Surg. 2021;122(4):333\u0026ndash;7.\u003c/li\u003e\n \u003cli\u003eMadhuri J, Mohan S, Praveen S, Chandan T, Reddy Gurram T, Anitha Kumari R, et al. Detection of Mid Mesial Canal in Dental CBCT Images Using Combined YOLO and 2D U-Net Model. In Institute of Electrical and Electronics Engineers Inc.; 2023. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85175401195\u0026amp;doi=10.1109%2fNMITCON58196.2023.10275873\u0026amp;partnerID=40\u0026amp;md5=332dd5428d5350c6ef3f5eab2d082c5c\u003c/li\u003e\n \u003cli\u003eZhang L, Xu F, Li Y, Zhang H, Xi Z, Xiang J, et al. A lightweight convolutional neural network model with receptive field block for C-shaped root canal detection in mandibular second molars. Sci Rep. 2022;12(1):17373.\u003c/li\u003e\n \u003cli\u003eZhou Y, Zhang H. Root Canal Therapy Evaluation Based on Rule Embedded Neural Networks. In Institute of Electrical and Electronics Engineers Inc.; 2021. p. 749\u0026ndash;54. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85124403156\u0026amp;doi=10.1109%2fICAA53760.2021.00136\u0026amp;partnerID=40\u0026amp;md5=fc353ba1d5b1d5258753caa5fcab9551\u003c/li\u003e\n \u003cli\u003eHasan HA, Saad FH, Ahmed S, Mohammed N, Farook TH, Dudley J. Experimental validation of computer-vision methods for the successful detection of endodontic treatment obturation and progression from noisy radiographs. Oral Radiol. 2023;39(4):683\u0026ndash;98.\u003c/li\u003e\n \u003cli\u003eLin X, Fu Y, Ren G, Yang X, Duan W, Chen Y, et al. Micro-Computed Tomography-Guided Artificial Intelligence for Pulp Cavity and Tooth Segmentation on Cone-beam Computed Tomography. J Endod. 2021;47(12):1933\u0026ndash;41.\u003c/li\u003e\n \u003cli\u003eSchneider L, Arsiwala-Scheppach L, Krois J, Meyer-Lueckel H, Bressem KK, Niehues SM, et al. Benchmarking Deep Learning Models for Tooth Structure Segmentation. J Dent Res. 2022;101(11):1343\u0026ndash;9.\u003c/li\u003e\n \u003cli\u003eMnekhir HJ. THE US-CHINESE RACE IN ARTIFICIAL INTELLIGENCE CHALLENGES AND OPPORTUNITIES. Russ Law J [Internet]. 2023 Apr 7 [cited 2024 Jul 21];11(3). Available from: https://russianlawjournal.org/index.php/journal/article/view/2182\u003c/li\u003e\n \u003cli\u003eValente NA, Tichy A, Schwendicke F, Chaurasia A, Hamdan M, Amanabi M, et al. Completeness of METADATA Reporting in AI Dental Research: Scoping Review Protocol. 2024 [cited 2025 Jan 22]; Available from: https://osf.io/jxmsf/\u003c/li\u003e\n \u003cli\u003eSchwendicke F, Uribe SE, Issa J. Quality and Completeness of Reporting in Dental AI Research. 2024 [cited 2025 Jan 22]; Available from: https://osf.io/f2uzy/\u003c/li\u003e\n \u003cli\u003eJakubovics NS, Schwendicke F. Toward Better Reporting in Oral Health Research. J Dent Res. 2024 Oct;103(11):1045\u0026ndash;6.\u003c/li\u003e\n \u003cli\u003eLiu J, Peng G, Yan S. An Intelligent Evaluation Method of Root Canal Therapy Quality Based on Deep Learning. In Institute of Electrical and Electronics Engineers Inc.; 2022. p. 6254\u0026ndash;9. Available from: https://www.scopus.com/inward/record.uri?eid=2-s2.0-85151151347\u0026amp;doi=10.1109%2fCAC57257.2022.10056075\u0026amp;partnerID=40\u0026amp;md5=afa692730e693b9ee522c7d6a8a5a8a2\u003c/li\u003e\n\u003c/ol\u003e"},{"header":"Table","content":"\u003cp\u003eTable 1 is available in the Supplementary Files section\u003c/p\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"hideJournal":true,"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":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"","lastPublishedDoi":"10.21203/rs.3.rs-5952801/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5952801/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003e\u003cem\u003eObjectives:\u003c/em\u003e The number, size, patency, and location of pulp canals are critical in endodontic treatment planning. This information is currently obtained through visual radiographic assessment, which is time-consuming and labor-intensive. Artificial intelligence (AI) could automate this task via accurate segmentation of root canals providing efficiency and consistency. This scoping review maps existing literature on the use of AI to automate root canal segmentation on radiographic images.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eMaterials and Methods:\u003c/em\u003e We searched MEDLINE (Ovid), Embase, Scopus, and Web of Science for relevant studies up to January 8, 2025. Studies that used AI for root canal segmentation were included. Study selection was not limited by design, language, or date. Commentaries, retracted articles, and inaccessible full-text articles were excluded. Titles and abstracts were screened based on eligibility criteria, and the full-text of potentially relevant studies was assessed. Screening and data extraction were conducted in duplicate by independent reviewers, with disagreements resolved via consensus or a third-reviewer if necessary.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eResults:\u003c/em\u003e Out of 836 articles identified, 35 studies met the eligibility criteria and were retained for synthesis. Data extraction focused on the country of origin, study design, imaging modalities, obturation status, type of teeth analyzed, AI models used, and results. Modalities included were mostly cone beam computed tomography (CBCT, 51%), followed by panoramic (17%) and periapical (14%). AI-based models, particularly those employing CNNs, reported accuracies ranging from 0.73 to 0.99 and sensitivities from 0.72 to 1. These models were effective across all imaging modalities with most studies reporting improved diagnostic precision and reduced time compared with manual methods.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eConclusions:\u003c/em\u003e AI-based root canal segmentation has clinical value by increasing accuracy in identifying root canal anatomy prior to treatment. This will preserve clinicians' time and reduce the risk of treatment failure. This review highlights current status of this technological application and identifies areas to refine these technologies for broad clinical application to enhance patient outcomes in endodontic care.\u003c/p\u003e\n\u003cp\u003e\u003cem\u003eClinical Relevance:\u003c/em\u003e The application of AI in root canal segmentation offers significant clinical benefits by improving the accuracy and efficiency of identifying root canal anatomy. This can lead to better treatment planning, reduced procedure times, and lower risk of endodontic failure. As AI technology continues to evolve, its integration into endodontic practice has the potential to enhance patient outcomes and streamline clinical workflows.\u003c/p\u003e","manuscriptTitle":"Artificial Intelligence for Root Canal Segmentation on Radiographic Images: A Scoping Review","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-02-11 09:51:27","doi":"10.21203/rs.3.rs-5952801/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"6834dfe3-3ffc-4eea-9466-4eadcdbe83ac","owner":[],"postedDate":"February 11th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2025-02-11T09:51:27+00:00","versionOfRecord":[],"versionCreatedAt":"2025-02-11 09:51:27","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5952801","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5952801","identity":"rs-5952801","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}