Related Experiment Video
Updated: May 8, 2026
Accuracy in Dental Medicine, A New Way to Measure Trueness and Precision
Published on: April 29, 2014
Accuracy in diagnosing caries in young permanent molars using interproximal radiographic imaging and validation by
Débora Heloísa Silva de Brito1, Thaysa Gomes Ferreira Tenório Dos Santos1, Samylla Glória de Araújo Costa2
1DDS, MSc, PhD student. University of Pernambuco, School of Dentistry - Department of Pediatric Dentistry, Recife, Pernambuco, Brazil.
Background:
Caries lesions, in their early stages, can be challenging to identify clinically, as they often do not cause symptoms or are in areas that are difficult to access. Caries diagnosis involves high subjectivity and can lack consistency among professionals with different backgrounds and levels of experience. Discrepancies may occur between examiners or even with the same examiner at other times. With technological advancements, increasingly efficient methods for diagnosing dental caries are available, and new techniques and tools are under study. This study aims to evaluate the accuracy of diagnosing caries lesions in young permanent molars using interproximal radiographs by training object detection algorithms with an artificial intelligence (AI) system and comparing them to inter-examiner diagnoses.
Material And Methods:
A descriptive study was conducted in interproximal images of the first permanent molars of children aged between 6 and 9 years. The radiographs were obtained from three private radiological clinics in The Federal District The training was conducted by graduate dentists and calibrated using Professor of Radiology (MCF) as the gold standard. The YOLOv8 model architecture and a pre-trained classifier (EfficientNet-B0) were used.
Results:
The kappa agreement index was obtained to evaluate the degree of agreement between examiners. The inter-examiner agreement in the caries diagnosis was considered excellent, being 97.4%, with a kappa value of 0.88. The YOLOv8 model was applied to detect carious teeth using AI. The results show that the model achieved excellent performance, with accuracy metrics of 91% and precision of 98%. The EfficientNet-B0 classifier categorized teeth with and without caries lesions. The classifier achieved an accuracy of 89%.
Conclusions:
There was excellent inter-examiner agreement in evaluating caries diagnosis for the teeth assessed. The AI-based method proposed in this study showed good performance and proved effective in recognizing caries lesions in radiographic images. Key words:Dental caries, Artificial intelligence, Radiography, Bitewing, Child.
