Related Experiment Video
Updated: Jun 10, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Automatic maxillary sinus segmentation and pathology classification on cone-beam computed tomographic images using
Oğuzhan Altun1, Duygu Çelik Özen1, Şuayip Burak Duman2,3
1Department of Oral and Dentomaxillofacial Radiology, Faculty of Dentistry, Inonu University, Malatya, Turkey.
BMC Oral Health
|October 10, 2024
Summary
This study shows an AI model can accurately segment maxillary sinuses and detect diseases like sinusitis using CBCT scans. This deep learning approach aids physicians in virtual planning for maxillofacial surgeries.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Oral and Maxillofacial Radiology
Background:
- Automated segmentation of the maxillofacial complex can enhance virtual surgical planning.
- Deep learning (DL) systems can aid physicians in detecting maxillary sinus pathologies.
Purpose of the Study:
- To segment maxillary sinuses and diseases using a modified YOLOv5x architecture with transfer learning.
- To evaluate the AI model's performance on Cone-Beam Computed Tomographic (CBCT) images.
Main Methods:
- Utilized a dataset of 307 anonymized CBCT scans from the Department of Oral and Maxillofacial Radiology.
- Employed a modified YOLOv5x architecture for segmentation of healthy sinuses, mucous retention cysts (MRC), mucosal thickenings (MT), and sinusitis.
Main Results:
- Achieved high performance metrics: F1 scores of 0.992 for total sinus segmentation, 0.964 for healthy sinuses, 0.889 for MT, 0.924 for MRC, and 0.970 for sinusitis.
- Demonstrated excellent recall and precision across all segmentation tasks.
Conclusions:
- The developed AI model effectively segments maxillary sinuses and accurately detects associated diseases.
- This automated approach supports clinical decision-making and surgical preparation in maxillofacial radiology.
