Deep learning-based fully automatic segmentation of the maxillary sinus on cone-beam computed tomographic images

Area of Science:

• Dentistry and Medical Imaging
• Artificial Intelligence in Healthcare

Background:

• Accurate maxillary sinus segmentation is crucial for dental implant surgery, tooth extraction, and diagnosing odontogenic diseases.
• Current segmentation methods may lack the precision required for complex cases, including hazy sinus states.

Purpose of the Study:

• To develop and evaluate a deep learning-based method for fully automatic segmentation of maxillary sinuses in cone-beam computed tomographic (CBCT) images.
• To assess the model's performance on both clear and hazy maxillary sinus conditions.

Main Methods:

• A U-Net convolutional neural network was employed for maxillary sinus segmentation.
• The model was trained and validated on a dataset of 19,350 CBCT images from 90 maxillary sinuses.
• Post-processing techniques were applied to refine the segmentation results and reduce prediction errors.

Main Results:

• The U-Net model achieved an average Dice Similarity Coefficient (DSC) of 0.9090 and Hausdorff Distance (HD) of 2.7013 before post-processing.
• After post-processing, the average DSC improved to 0.9099 and HD to 2.1470.
• The model demonstrated robust performance in segmenting both clear and hazy maxillary sinuses.

Conclusions:

• The proposed deep learning model, enhanced with post-processing, provides accurate and reliable maxillary sinus segmentation.
• This automated approach has the potential to significantly assist dental clinicians in diagnosis and treatment planning.
• The model's consistent accuracy across various sinus conditions makes it a valuable tool for dental practice.