Related Experiment Video
Updated: Jul 2, 2025
05:49
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
Published on: February 23, 2024
843
Application of deep learning and feature selection technique on external root resorption identification on CBCT
Nor Hidayah Reduwan1,2, Azwatee Abdul Aziz3, Roziana Mohd Razi4
1Department of Oral and Maxillofacial Clinical Sciences, Faculty of Dentistry, Universiti Malaya, Kuala Lumpur, 50603, Malaysia.
BMC Oral Health
|February 19, 2024
Summary
Combining artificial intelligence models with feature selection techniques significantly improves external root resorption identification. The Random Forest and Visual Geometry Group 16 model, enhanced by feature selection, demonstrated superior diagnostic performance.
Area of Science:
- Dentistry
- Artificial Intelligence
- Medical Imaging
Background:
- Artificial intelligence (AI) shows promise in identifying maxillofacial lesions.
- External root resorption (ERR) is a dental condition requiring accurate identification.
Purpose of the Study:
- To evaluate four deep learning models (DLMs) for external root resorption (ERR) identification.
- To assess the impact of combining feature selection techniques (FST) with DLMs on ERR identification accuracy.
Main Methods:
- Simulated ERR on 88 premolars at varying depths (0.5-2 mm).
- Acquired Cone Beam CT scans for all samples.
- Trained and tested four DLMs (RF+VGG, RF+EFNET, SVM+VGG, SVM+EFNET) and their FST-hybridized counterparts.
Main Results:
- The Random Forest (RF) + Visual Geometry Group 16 (VGG) model showed the highest performance.
- FST combined with RF+VGG achieved 81.9% accuracy, 83% weighted accuracy, and 96% AUC.
- Significant differences (p=0.008) in prediction accuracy were observed among the eight models.
Conclusions:
- Deep learning models demonstrate comparable performance in ERR identification.
- Integrating feature selection techniques enhances the diagnostic capability of DLMs for ERR.
