Related Experiment Video
Updated: May 30, 2025
07:32
Author Spotlight: 3D Movement Assessment of Maxillary Posterior Teeth in Clear Aligner Treatment
Published on: February 23, 2024
1.0K
A platform combining automatic segmentation and automatic measurement of the maxillary sinus and adjacent structures
Jiawei He1,2, Muxi Sun3, Youtong Huo1
1State Key Laboratory of Oral Diseases & National Center for Stomatology & National Clinical Research Center for Oral Diseases, West China Hospital of Stomatology, Sichuan University, Chengdu, Sichuan, 610041, China.
Clinical Oral Investigations
|January 25, 2025
Summary
A deep convolutional neural network (DCNN) platform accurately segments maxillary sinus (MS) and adjacent structures. This enables dentists to perform instant 3D reconstruction and measurements for improved clinical insights.
Area of Science:
- Utilizes advanced deep learning, specifically deep convolutional neural networks (DCNNs), for medical image analysis.
- Focuses on quantitative analysis of Cone Beam Computed Tomography (CBCT) data for dental and maxillofacial applications.
Background:
- Accurate segmentation and measurement of maxillary sinus (MS) and surrounding structures are crucial in dental diagnostics.
- Manual segmentation and measurement are time-consuming and prone to inter-observer variability.
Purpose of the Study:
- To develop an automated platform using a DCNN for segmenting the maxillary sinus (MS) and adjacent anatomical structures.
- To create algorithms for automatic measurement of 3D clinical parameters from CBCT scans.
- To assess the clinical reliability and accuracy of the developed DCNN and measurement algorithms.
Main Methods:
- Trained a DCNN with a 2.5D structure on 175 CBCT datasets (242 MS), including healthy and diseased sinuses with varying mucosal thickening.
- Developed automatic algorithms for measuring 3D clinical parameters post-segmentation.
- Validated segmentation accuracy using Dice Similarity Coefficient (DSC) and measurement reliability using Intra-class Correlation Coefficient (ICC).
Main Results:
- Achieved high median DSC values for segmentation: 0.990 (air cavity), 0.850 (mucosa), 0.961 (teeth), and 0.953 (maxillary bone).
- Demonstrated excellent reliability with ICC exceeding 0.975 for all automatic measurement algorithms.
- Reported low bias for volumetric (±0.5 cm³) and 2D (±1 mm) metrics, with satisfactory performance on incomplete MS and edentulous crests.
Conclusions:
- The DCNN-based platform provides clinically reliable automatic segmentation of the maxillary sinus and adjacent structures.
- Automatic measurement algorithms effectively extract 3D clinical information from CBCT 2D planes.
- This integrated platform facilitates instant 3D reconstruction and parameter measurement for dental professionals.
