EVALUATION OF AN ARTIFICIAL INTELLIGENCE MODEL FOR THE DETECTION OF DENTAL ABSCESS ON ORTHOPANTOMOGRAMS

Background: Dental X-rays are fundamental for diagnosing intra-bony lesions, such as dental abscesses. Recently, deep learning techniques have shown promise in enhancing radiographic interpretation accuracy and efficiency. Objective:This study aimed to develop and evaluate a deep learning model for the automated detection of dental abscesses on orthopantomograms. Materials and Methods: A classification model based on the EfficientNetB3 architecture was developed. The model was trained and validated on a unique dataset of 238 confirmed abscess cases, which was expanded to 714 images using a multilevel data augmentation strategy to improve generalization. The performance of the model was then evaluated on a separate, independent test set of 40 cases. Results: The EfficientNetB3 model achieved 96.40% validation accuracy on the augmented dataset. The model demonstrated 83.3% recall (95% CI: 68.6%–93.0%) and 53.0% precision (95% CI: 38.5%–67.1%) for abscess detection on the independent test set. This corresponds to a true positive rate of 87.5% and a false negative rate of 12.5%. Conclusion: This study demonstrates that employing a specialized model for the classification task constitutes an effective methodological approach for clearly the Detection of abscesses. Nevertheless, the model’s performance is substantially constrained by the size and class distribution of the training dataset, thereby reaffirming that access to large-scale, well-balanced data is fundamental for the development of robust and reliable clinical predictive models.