Related Experiment Video
Updated: Mar 15, 2026
04:48
Application of Deep Learning-Based Medical Image Segmentation via Orbital Computed Tomography
Published on: November 30, 2022
3.6K
Evaluation of Maxillary Sinus Membrane Morphology Using a Novel Hybrid CNN-ViT-Based Deep Learning Model: An
Nurullah Duger1, Furkan Talo2, Gulucag Giray Tekin3
1Department of Periodontology, Faculty of Dentistry, Firat University, Elazig 23119, Turkey.
Diagnostics (Basel, Switzerland)
|March 14, 2026
Summary
A new hybrid deep learning model combining Convolutional Neural Networks (CNN) and Vision Transformers (ViT) accurately classifies maxillary sinus membrane morphologies from CBCT scans. This AI tool aids in assessing surgical risks before dental implant procedures.
Area of Science:
- Medical Imaging Analysis
- Artificial Intelligence in Healthcare
- Deep Learning for Medical Diagnostics
Background:
- Maxillary sinus membrane morphology classification is crucial for predicting surgical risks like perforation and sinusitis.
- Current methods may lack accuracy or efficiency in classifying diverse morphologies.
- Cone-Beam Computed Tomography (CBCT) provides detailed anatomical data for sinus analysis.
Purpose of the Study:
- To develop and validate a hybrid deep learning model for automated classification of maxillary sinus membrane morphologies.
- To distinguish between Normal, Flat, Polypoid, and Obstruction types using CBCT images.
- To enhance diagnostic accuracy and clinical decision support for implant surgery.
Main Methods:
- A dataset of 959 CBCT images was curated and classified into four morphological types.
- A hybrid model integrating CNN for local features and ViT for global context was developed.
- Performance was evaluated against established CNN and ViT models using accuracy, precision, recall, and F1-score.
Main Results:
- The hybrid CNN-ViT model achieved a superior overall accuracy of 98.44%, outperforming ResNet50 and ViT-B16.
- The model demonstrated exceptional performance for the 'Obstruction' class (100% accuracy) and high accuracy for 'Flat' (98.21%) and 'Polypoid' (98.04%) morphologies.
- The hybrid approach effectively addressed limitations of standard ViT models with limited datasets.
Conclusions:
- The proposed hybrid CNN-ViT model offers a highly accurate and reliable method for classifying maxillary sinus membrane morphologies.
- This AI tool can serve as a valuable clinical decision support system for pre-surgical risk assessment.
- Objective assessment of risk factors prior to implant surgery and sinus floor elevation is facilitated by this technology.