Related Experiment Video
Updated: Jun 13, 2026
10:50
A Finite Element Approach for Locating the Center of Resistance of Maxillary Teeth
Published on: April 8, 2020
SinusNet+: Deep Condition-Label-Free Segmentation of Maxillary Sinus Conditions in CBCT images.
Da-El Kim1, Su Yang2,3, Sang Heon Lim1
1Interdisciplinary Program in Bioengineering, Seoul National University, Gwanak-gu, Seoul, 08826, South Korea.
Dento Maxillo Facial Radiology
|June 11, 2026
Summary
A new deep learning framework, SinusNet+, enables condition-label-free segmentation of maxillary sinus conditions (MSC) in cone-beam computed tomography (CBCT) images. This method reduces the need for manual annotations, supporting dental implant planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oral and Maxillofacial Surgery
Background:
- Accurate segmentation of maxillary sinus conditions (MSC) in cone-beam computed tomography (CBCT) is crucial for preoperative assessment in posterior maxilla procedures like implant planning.
- Current supervised deep learning methods require extensive manual annotation of MSC, which is time-consuming and labor-intensive.
Purpose of the Study:
- To develop and evaluate SinusNet+, a novel deep learning framework for condition-label-free MSC segmentation in CBCT images.
- To eliminate the need for manual MSC annotations during network training by utilizing synthetic conditions.
Main Methods:
- A synthetic condition generator was developed to simulate various MSC within normal maxillary sinuses (MS) by manipulating texture, shape, and noise.
- These synthetic conditions approximate the radiographic appearance of MSC in CBCT images for network training.
Main Results:
- SinusNet+ achieved high segmentation performance with an average Dice similarity coefficient of 0.820, precision of 0.878, and recall of 0.777.
- The condition-label-free approach demonstrated comparable performance to supervised methods and outperformed unsupervised baselines.
Conclusions:
- The SinusNet+ framework proves the feasibility of condition-label-free MSC segmentation in CBCT images.
- While reducing annotation burden, anatomical annotation of the normal maxillary sinus is still necessary for dataset preparation.
