Related Experiment Video
Updated: Sep 12, 2026
Measuring Maxillary Posterior Tooth Movement: A Model Assessment using Palatal and Dental Superimposition
Published on: February 23, 2024
Application of machine learning in predicting of MB2 canal in permanent maxillary first and second molars: A CBCT
1Department of Operative Dentistry and Endodontics, College of Dentistry, University of Sulaimani, Sulaimani, Iraq.
Objectives:
The aim of this study is to apply Machine learning (ML), a key subset of Artificial Intelligence (AI) in predicting second mesiobuccal (MB2) canals in both maxillary first molars (MFMs) and maxillary second molars (MSMs) using the cone-beam computed tomography (CBCT) imaging in an Iraqi sub-population.
Methods:
In this retrospective study, 144 CBCT scans from 63 female and 81 male patients were retrieved from the archives of a radiology department at B&R Dental Center. The presence and absence of MB2 canal in both MFMs and MSMs in relation to age, sex, and side, area of triangle (A) between MB, DB and P canals, semiperimeter (SP), and the distances between the orifices of mesio-buccal (MB), disto-buccal (DB), and palatal (P) canals in both MFMs and MSMs in the axial plane were measured on CBCT scans. Descriptive and inferential statistics were employed for data analysis. The receiver operating characteristic (ROC) curve analysis was employed to assess the diagnostic accuracy of all data in predicting the existence of an MB2 canal in both molars. The optimal cut-off point was established based on sensitivity and specificity. The models' classification measures, including area under the curve (AUC), accuracy, F1-score, and precision, were evaluated.
Results:
Overall, the prevalence of the MB2 canal was 65.24% in maxillary first molars (MFMs) and 36.89% in maxillary second molars (MSMs). A significantly higher prevalence of the MB2 canal among females was observed only in MSMs (p < 0.001), whereas no significant sex difference was found in MFMs (p = 0.915). Logistic regression (LR) demonstrated the best performance among all machine learning (ML) models, achieving an AUC of 0.88 for MFMs and 0.90 for MSMs. Feature importance analysis using Random Forest (RF) and Decision Tree (DT) models identified the mesiobuccal-to-palatal (MB-P) distance as the most influential predictor of MB2 canal presence.
Conclusions:
The present study showed promising results in the ML based predicting of MB2 canal in both MFMs and MSMs using axial CBCT slices based on MB-P distance parameter as the most influential predictor for the presence of the MB2 canal in both MFMs and MSMs using all ML models. These findings might allow clinicians to identify MB2 canals in maxillary molars for effective endodontic treatment.
