Related Experiment Video
Updated: Aug 20, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.8K
Automatic and quantitative measurement of alveolar bone level in OCT images using deep learning
Sul-Hee Kim1,2, Jin Kim3,2, Su Yang4
1Department of Periodontology, School of Dentistry, Seoul National University, Seoul, 03080, Republic of Korea.
Biomedical Optics Express
|November 25, 2022
Summary
This study introduces a convolutional neural network (CNN) method for automatically segmenting tooth enamel and alveolar bone in optical coherence tomography (OCT) images. The approach accurately measures alveolar bone level (ABL), crucial for diagnosing periodontal disease.
Area of Science:
- Biomedical Engineering
- Medical Imaging
- Periodontology
Background:
- Periodontal disease diagnosis relies on accurate measurement of alveolar bone level (ABL).
- Manual measurement of ABL from optical coherence tomography (OCT) images is time-consuming and subjective.
- Automated segmentation and measurement methods are needed for efficient and reliable ABL assessment.
Purpose of the Study:
- To develop and validate a convolutional neural network (CNN) based method for automatic segmentation of tooth enamel and alveolar bone in OCT images.
- To quantitatively measure the alveolar bone level (ABL) by automatically detecting the cemento-enamel junction (CEJ) and alveolar bone crest (ABC).
- To assess the accuracy and reliability of the proposed CNN models for ABL measurement.
Main Methods:
- Utilized U-Net, Dense-UNet, and U2-Net convolutional neural network (CNN) architectures for automatic segmentation of tooth enamel and alveolar bone.
- Employed image processing techniques to measure alveolar bone level (ABL) as the distance between the detected cemento-enamel junction (CEJ) and alveolar bone crest (ABC).
- Evaluated segmentation accuracy and ABL measurement precision using metrics such as mean distance difference (MDD) and mean absolute error (MAE).
Main Results:
- CNN models achieved high segmentation accuracies for tooth enamel and alveolar bone.
- The proposed method demonstrated a mean distance difference (MDD) ranging from 0.18 to 0.32 mm for CEJ and 0.19 to 0.22 mm for ABC.
- All tested CNN models exhibited a mean absolute error (MAE) below 0.25 mm and a successful detection rate (SDR) greater than 90% at 0.5 mm for both CEJ and ABC.
- ABL measurements in incisors showed high correlation and reliability with ground truth data in OCT images.
Conclusions:
- The developed CNN-based method enables accurate and automatic segmentation of periodontal structures and quantitative measurement of alveolar bone level (ABL) from OCT images.
- The high accuracy and reliability of the proposed method support its potential clinical application in the diagnosis and monitoring of periodontal diseases.
- This automated approach offers a promising advancement for objective and efficient assessment of periodontal health using OCT imaging.
