Research Objective
White spot lesions are common early signs of dental caries in patients undergoing orthodontic treatment. This study aims to detect these lesions using a deep learning-based YOLOv5x algorithm.
Method and Dataset
435 post-orthodontic intraoral photographs were manually labeled for white spot lesions. This labeled dataset was used to train and test the deep learning algorithm.
Performance Metrics and Results
- Accuracy: 78.6%
- Recall: 61.8%
- F1 Score: 69.2%
- AUC (Area Under the Curve): 71.2%
- mAP (Mean Average Precision): 42.5%
Research Findings
The model performed below expectations in detecting white spot lesions but achieved an acceptable accuracy compared to previous studies. These results suggest that with larger datasets and algorithm improvements, enhanced models could be used in clinical settings.
Conclusion
This pilot study demonstrates the potential use of artificial intelligence in detecting white spot lesions in dentistry. The findings provide a significant foundation for further research and development. The integration of dentistry and artificial intelligence could pave the way for more accurate and rapid diagnostics in the future.
