,
PhD Student, Department of Computer Science, King Abdulaziz University , Jeddah , Saudi Arabia
,
Professor, Immersive virtual reality research group, Department of Computer Science, King Abdulaziz University , Jeddah , Saudi Arabia
,
Professor, consultant, and chairman, Department of Restorative Dentistry, Faculty of Dentistry, King Abdulaziz University , Jeddah , Saudi Arabia
,
Assistant Professor, Department of Computer Science, Faculty of Computing and Information Technology, King Abdulaziz University , Jeddah , Saudi Arabia
,
General Practitioner, Ministry of National Guard Health Affairs, King Abdulaziz Medical City , Jeddah , Saudi Arabia
General Practitioner, University Dental Hospital, King Abdulaziz University , Jeddah , Saudi Arabia
The study will be conducted to complement the diagnosis of proximal caries lesions that are difficult to improve due to their location between the teeth, as shown in Near-Infrared Light Transillumination (NILT) photographs. It was proposed to enhance caries detection with a semantic segmentation model based on YOLOv11, which is more specific in detecting mesial and distal caries lesions, which were not adequately explored in previous literature. The model was trained on a sample of 440 augmented grayscale NILT images collected from 17 patients at the Faculty of Nursing, King Abdulaziz University, Jeddah, Saudi Arabia. These pictures were categorized into five groups, i.e., enamel, dentin, background, mesial caries, and distal caries. The significance of the classes was addressed by optimizing hyperparameters and class weights for mesial and distal caries. The YOLOv11 model had an overall Dice coefficient of 87 and the highest scores of enamels (80) and dentin (89) due to their low prevalence in the dataset (mesial and distal caries, respectively). However, the model was deemed very specific, and its negative predictive value was 0 in all classes. The findings indicate that a well-represented class is the most effective with the model, whereas mesial and distal caries need further improvement. Future work will focus on improving the model's performance on underrepresented classes through data augmentation and class-balanced training, ultimately enhancing its clinical applicability for the intricate, non-invasive detection and diagnosis of caries.
This is an open access article distributed under the Creative Commons Attribution Non-Commercial License (CC BY-NC) License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
0
The statements, opinions and data contained in the journal are solely those of the individual authors and contributors and not of the publisher and the editor(s). We stay neutral with regard to jurisdictional claims in published maps and institutional affiliations.