Alveolar Bone Segmentation Methods in Assessing the Effectiveness of Periodontal Defect Regeneration Through Machine Learning of CBCT Data: A Systematic Review

Description

Mahmud Mohammed, Tulio Fernandez-Medina, Manjunath Rajashekhar, Stephanie Baker, and Ernest Jennings review how well computers can outline jawbone in dental scans, in a 2025 paper. Dentists rebuilding lost gum bone need exact maps of the bone from cone-beam CT scans, and drawing those maps by hand is slow. The review gathered studies from the last five years that use machine learning for this outlining task, called segmentation.

From 834 papers found, 23 met the bar for inclusion. Deep learning methods, above all a design called U-Net, were the most common, and their accuracy scores ranged widely, from 76 to 98 percent overlap with hand-drawn maps. Only four studies used semi-automated methods. Bone with defects proved harder to outline than healthy bone. The authors call for better deep learning tools tuned to damaged bone, since exact maps guide regenerative treatment.

Catalog record

Source: https://europepmc.org/articles/PMC12752843