Related Experiment Video
Updated: Jul 12, 2025
05:49
Author Spotlight: Advancing CBCT and Digital Dental Image Integration with AI-Assisted Digitization
Published on: February 23, 2024
875
Improving accuracy of early dental carious lesions detection using deep learning-based automated method
Paula Dresch Portella1, Lucas Ferrari de Oliveira2, Mateus Felipe de Cássio Ferreira2
1Stomatology Department, Universidade Federal do Paraná, Avenida Prefeito Lothário Meissner, 632, Jardim Botânico, Curitiba, PR, 80210-170, Brazil. pauladresch@hotmail.com.
Clinical Oral Investigations
|October 31, 2023
Summary
A convolutional neural network (CNN) effectively detects early tooth decay. This deep learning tool significantly improves diagnostic accuracy when used by dental professionals as an auxiliary aid.
Area of Science:
- Dentistry
- Artificial Intelligence
- Medical Imaging
Background:
- Early detection of dental caries is crucial for effective treatment.
- Traditional diagnostic methods can be subjective and may miss subtle signs of early decay.
- Deep learning offers potential for objective and accurate image analysis in dentistry.
Purpose of the Study:
- To evaluate the effectiveness of a specific convolutional neural network (CNN), VGG-19, in identifying healthy teeth and early-stage carious lesions on occlusal surfaces.
- To assess the utility of this deep learning algorithm as a supplementary tool for dental practitioners.
Main Methods:
- A dataset of 2,481 posterior teeth images, classified using the International Caries Detection and Assessment System (ICDAS), was utilized.
- The VGG-19 CNN was trained and tested on these images, with results compared against a reference examiner.
- Examiners' diagnostic accuracy was compared before and after using the CNN as an assistive tool.
Main Results:
- The VGG-19 model achieved high performance metrics, including an accuracy of 0.879 and an F1-score of 0.887.
- Dental examiners showed improved accuracy when using the CNN as an auxiliary aid, with statistically significant increases observed.
- The CNN demonstrated robust performance in detecting early carious lesions.
Conclusions:
- The VGG-19 CNN shows promise for the reliable detection of early dental caries.
- Deep learning-based automated detection serves as a valuable aid, enhancing diagnostic accuracy and supporting clinical decision-making in caries management.
