Related Experiment Video
Updated: Aug 26, 2025
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
Published on: February 23, 2024
Detection of Proximal Caries Lesions on Bitewing Radiographs Using Deep Learning Method
Xiaotong Chen1, Jiachang Guo2, Jiaxue Ye1
1Department of Cariology and Endodontology, Peking University School and Hospital of Stomatology and National Clinical Research Center for Oral Diseases and National Engineering Research of Oral Biomaterials and Digital Medical Devices and Beijing Key Laboratory of Digital Stomatology, Beijing, China.
A deep learning convolutional neural network (CNN) shows promise as an assistant tool for detecting proximal caries on dental bitewing radiographs, outperforming postgraduate students in identifying early-stage lesions.
Area of Science:
- Dentistry
- Artificial Intelligence
- Medical Imaging
Background:
- Proximal caries detection on bitewing radiographs is crucial for timely dental treatment.
- Traditional methods rely on human interpretation, which can be subjective and prone to errors.
- Deep learning offers a potential solution for improving diagnostic accuracy.
Purpose of the Study:
- To evaluate the diagnostic performance of a deep learning-based convolutional neural network (CNN) for detecting proximal caries.
- To compare the CNN's performance against postgraduate dental students.
- To assess the CNN's ability to detect caries at different stages of severity.
Main Methods:
- A Faster R-CNN deep learning model was trained on 978 bitewing radiographs (10,899 proximal surfaces).
- The dataset included 2,719 annotated proximal caries lesions and 8,180 sound surfaces.
- Model performance was evaluated on a separate test set and compared to postgraduate students' interpretations.
Main Results:
- The CNN achieved an overall accuracy of 0.87, with a sensitivity of 0.72, specificity of 0.93, PPV of 0.77, and NPV of 0.91.
- The CNN demonstrated significantly higher sensitivity (≥0.65) for early-stage lesions compared to students (<0.40).
- The CNN's F1-score (0.74) surpassed that of students (0.57).
Conclusions:
- The deep learning CNN shows potential as an effective assistant tool for detecting proximal caries on bitewing radiographs.
- The CNN exhibits superior performance in identifying early-stage enamel and outer dentin lesions compared to human interpretation.
- Further integration of AI can enhance diagnostic capabilities in dentistry.
