Related Experiment Video
Updated: Jun 18, 2025
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
2.7K
Automatic segmentation of the maxillary sinus on cone beam computed tomographic images with U-Net deep learning model
Busra Ozturk1, Yavuz Selim Taspinar2, Murat Koklu3
1Department of Dentomaxillofacial Radiology, Faculty of Dentistry, Necmettin Erbakan University, Meram, Konya, 42050, Turkey.
Summary
This study developed a deep learning method for segmenting maxillary sinuses in cone beam computed tomography (CBCT) images. The U-Net model achieved high accuracy, improving diagnoses and surgical planning.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Radiology
Background:
- Medical imaging segmentation isolates specific structures in images.
- Maxillary sinus segmentation aids diagnosis and surgical planning.
- Deep learning offers advanced image processing capabilities.
Purpose of the Study:
- To develop a deep learning-based method for maxillary sinus segmentation.
- To improve image guidance for surgeons using cone beam computed tomography (CBCT) data.
- To enhance diagnostic accuracy and surgical intervention success rates.
Main Methods:
- Utilized axial CBCT images from 100 patients (200 sinuses).
- Employed the U-Net deep learning architecture for segmentation.
- Implemented early stopping to determine optimal training parameters (10 epochs, 100 iterations/epoch).
Main Results:
- Achieved high performance in maxillary sinus segmentation using the U-Net model.
- Intersection over Union (IoU) score of 0.9275.
- F1 Score of 0.9784.
Conclusions:
- The U-Net model demonstrates significant success in maxillary sinus segmentation.
- Enables fast, accurate evaluations, reducing clinician workload and subjective errors.
- Facilitates improved diagnostic and surgical outcomes in dentistry and related fields.
