Related Experiment Video
Updated: Sep 23, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.9K
Estimation of Alveolar Bone Loss in Periodontitis Using Machine Learning
Nektarios Tsoromokos1, Sarah Parinussa2, Frank Claessen2
1Department of Periodontology, Academic Centre for Dentistry Amsterdam (ACTA), University of Amsterdam and Vrije Universiteit Amsterdam, Amsterdam, The Netherlands.
International Dental Journal
|May 15, 2022
Summary
A new convolutional neural network (CNN) algorithm shows moderate to good reliability in automatically detecting and quantifying percentage alveolar bone loss (ABL) from dental radiographs. This AI tool aids in analyzing periodontitis progression using periapical imaging.
Area of Science:
- Dentistry
- Artificial Intelligence
- Medical Imaging
Background:
- Periodontitis is a common inflammatory disease affecting tooth-supporting structures.
- Accurate quantification of alveolar bone loss (ABL) is crucial for diagnosing and monitoring periodontitis.
- Manual analysis of periapical radiographs for ABL is time-consuming and subjective.
Purpose of the Study:
- To develop and evaluate a supervised machine learning model, specifically a convolutional neural network (CNN), for automated analysis of periapical radiographs.
- To quantify the percentage of alveolar bone loss (ABL) on approximal tooth surfaces in patients with and without periodontitis.
Main Methods:
- A dataset of 1546 approximal sites from 54 participants' periapical radiographs was manually annotated (MA).
- The data was divided into training (n=1308), validation (n=98), and test (n=140) sets.
- A CNN algorithm was developed and trained to identify key anatomical landmarks and quantify %ABL.
Main Results:
- The CNN algorithm achieved a mean %ABL of 23.1 ± 11.8%, compared to 27.8 ± 13.8% for manual annotation in the test set.
- The intraclass correlation (ICC) between the CNN and manual analysis was 0.601, indicating moderate reliability.
- Excellent reliability (ICC = 0.763) was observed for %ABL quantification on nonmolar teeth.
Conclusions:
- A CNN-based algorithm demonstrates moderate to good diagnostic performance for detecting and quantifying %ABL in periapical radiographs.
- This automated approach shows potential for improving the efficiency and objectivity of periodontitis assessment.
- Further development may enhance accuracy across all tooth types and bone loss patterns.
