Related Experiment Video
Updated: Aug 2, 2025
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
1.8K
Identification of Dental Implant Systems Using a Large-Scale Multicenter Data Set
W Park1,2, F Schwendicke3,4, J Krois3,4
1Department of Advanced General Dentistry, Yonsei University College of Dentistry, Seoul, Korea.
Journal of Dental Research
|April 22, 2023
Summary
Deep learning (DL) accurately identifies and classifies dental implant systems (DISs), outperforming dental professionals in accuracy and speed. This AI tool shows promise as a clinical decision support aid for DIS identification.
Area of Science:
- Dental radiology
- Artificial intelligence in dentistry
- Medical imaging analysis
Background:
- Accurate identification of dental implant systems (DISs) is crucial for effective treatment planning and patient care.
- Current methods rely on manual interpretation by dental professionals, which can be time-consuming and prone to errors.
- The increasing complexity and variety of DISs necessitate advanced diagnostic tools.
Purpose of the Study:
- To evaluate the efficacy of deep learning (DL) in identifying and classifying various types of dental implant systems (DISs).
- To compare the classification accuracy and time efficiency of DL models against dental professionals.
- To assess the potential of DL as a decision support tool in clinical dental practice.
Main Methods:
- A large-scale multicenter dataset of 37,442 periapical and 113,291 panoramic radiographic images was utilized.
- Images represented 10 manufacturers and 25 different types of DISs.
- A pretrained and modified ResNet-50 architecture was employed for DL classification, with performance compared to specialized and non-specialized dental professionals via questionnaire.
Main Results:
- Deep learning achieved a classification accuracy of 82.0% (95% CI, 75.9%-87.0%), significantly outperforming dental professionals (mean accuracy: 23.5% ± 18.5%).
- Specialized implantologists achieved a mean accuracy of 43.3% ± 20.4%, while non-specialized dentists achieved 16.8% ± 9.0%.
- DL required significantly less time (4.5 min) for classification compared to specialized (75.6 ± 31.0 min) and non-specialized (91.3 ± 38.3 min) dentists.
Conclusions:
- Deep learning demonstrates reliable outcomes in the identification and classification of diverse dental implant systems.
- The classification accuracy of DL significantly surpasses that of both specialized and non-specialized dental professionals.
- DL can be effectively implemented as a decision support aid for identifying and classifying DISs in clinical settings.
