Deep Learning for Automated Assessment of Mandibular Alveolar Bone Quality and Quantity on Cone Beam Computed Tomography for Dental Implant Planning: A Narrative Review
Kelsa Angelina Cen1, Eha Renwi Astuti2*, Ramadhan Hardani Putra2
Abstract
Background: Dental implants have become a preferred option for rehabilitating tooth loss, yet their success depends heavily on the quality and quantity of the alveolar bone at the planned implant site. Cone Beam Computed Tomography (CBCT) is the standard imaging modality for pre-surgical bone assessment, but manual measurement of bone dimensions across numerous cross-sections is time-consuming and operator-dependent, while the grey values of CBCT do not reliably represent bone density. Deep learning has therefore been explored to automate both aspects of bone evaluation. Purpose: This narrative review discusses the current application, performance, and limitations of deep learning for automated assessment of mandibular alveolar bone quality and quantity on CBCT to support dental implant planning. Review: A literature search was performed through PubMed, ScienceDirect, Web of Science, Scopus, and IEEE Xplore for articles published between 2012 and 2026. For bone quality, convolutional neural network (CNN) classifiers reached 86% accuracy in Lekholm and Zarb classification and outperformed implantologists; Nested U-Net enabled region-wise density grading, and a generative model reduced the error of CBCT-based bone mineral density measurement from 54.29% to 8.32%. For bone quantity, segmentation-based systems using U-Net++ and YOLOv8 delineated the alveolar ridge with Dice coefficients of 0.91–0.98, produced width measurements with a mean absolute error of 1.19 mm, and generated treatment suggestions agreeing with clinicians in 94% of cases. Small single-device datasets, heterogeneous reference standards, and the absence of links to clinical outcomes remain the main limitations. Conclusion: Deep learning shows promising performance for automated evaluation of mandibular bone quality and quantity on CBCT, but it should currently serve as a clinician-supervised decision-support tool.
Keywords
artificial intelligence; bone density; cone beam computed tomography; deep learning; dental implant
Cite This Article
Cen, K. A., Astuti, E. R., Putra, R. H. (2026). Deep Learning for Automated Assessment of Mandibular Alveolar Bone Quality and Quantity on Cone Beam Computed Tomography for Dental Implant Planning: A Narrative Review. International Journal of Scientific Advances (IJSCIA), Volume 7| Issue 5: Sep – Oct 2026, Pages 556-561 URL: https://www.ijscia.com/wp-content/uploads/2026/10/Volume7-Issue5-Sep-Oct-No.1076-556-561.pdf
Volume 7 | Issue 5: Sep – Oct 2026
