Application of entire dental panorama image data in artificial intelligence model for age estimation.

BACKGROUND: Accurate age estimation is vital for clinical and forensic purposes. With the rapid advancement of artificial intelligence(AI) technologies, traditional methods relying on tooth development, while reliable, can be enhanced by leveraging deep learning, particularly neural networks. This study evaluated the efficiency of an AI model by applying the entire panoramic image for age estimation. The outcome performances were analyzed through supervised learning (SL) models. METHODS: Total of 27,877 dental panorama images from 5 to 90 years of age were classified by 2 types of grouping. In type 1 they were classified by each age and in type 2, applying heuristic grouping, the age over 20 years were classified by every 5 years. Wide ResNet (WRN) and DenseNet (DN) were used for supervised learning. In addition, the analysis with ± 3 years of deviation in both types were performed. RESULTS: For the DN model, while the type 1 grouping achieved an accuracy of 0.1016 and F1 score of 0.058, the type 2 achieved an accuracy of 0.3146 and F1 score of 0.2027. Incorporating ± 3years of deviation, the accuracy of type 1 and 2 were 0.281, 0.7323 respectively; and the F1 score were 0.1768, 0.6583 respectively. For the WRN model, while the type 1 grouping achieved an accuracy of 0.1041 and F1 score of 0.0599, the type 2 achieved an accuracy of 0.3182 and F1 score of 0.2071. Incorporating ± 3years of deviation, the accuracy of type 1 and 2 were 0.2716, 0.7323 respectively; and the F1 score were 0.1709, 0.6437 respectively. CONCLUSIONS: The application of entire panorama image data for supervised with classification by heuristics grouping with ± 3years of deviation for supervised learning models and demonstrated satisfactory outcome for the age estimation.

The efficacy of supervised learning and semi-supervised learning in diagnosis of impacted third molar on panoramic radiographs through artificial intelligence model.

OBJECTIVES: The aim of the study was to evaluate the efficacy of traditional supervised learning (SL) and semi-supervised learning (SSL) in the classification of mandibular third molars (Mn3s) on panoramic images. The simplicity of preprocessing step and the outcome of the performance of SL and SSL were analyzed. METHODS: Total 1625 Mn3s cropped images from 1000 panoramic images were labeled for classifications of the depth of impaction (D class), spatial relation with adjacent second molar (S class), and relationship with inferior alveolar nerve canal (N class). For the SL model, WideResNet (WRN) was applicated and for the SSL model, LaplaceNet (LN) was utilized. RESULTS: In the WRN model, 300 labeled images for D and S classes, and 360 labeled images for N class were used for training and validation. In the LN model, only 40 labeled images for D, S, and N classes were used for learning. The F1 score were 0.87, 0.87, and 0.83 in WRN model, 0.84, 0.94, and 0.80 for D class, S class, and N class in the LN model, respectively. CONCLUSIONS: These results confirmed that the LN model applied as SSL, even utilizing a small number of labeled images, demonstrated the satisfactory of the prediction accuracy similar to that of the WRN model as SL.