Deep Learning Approaches for Automated Classification of Dental Conditions: A Comparative Study of InceptionV3 and ResNet18

This research paper proposes to develop enhanced detection and classification of six common dental conditions with deep learning models, which include Calculus, Data Caries, Gingivitis, Mouth Ulcer, Tooth Discoloration, and Hypodontia. It subsequently applies two state-of-the-art deep learning convolutional neural networks, namely, InceptionV3 and ResNet18, in an effort to classify clinical conditions from image data. InceptionV3 performed better than ResNet18 by achieving a high accuracy of 94% with higher precision and recall and F1 scores in most conditions. ResNet18 showed an accuracy of 91%, but it suffered more in the case of misclassifications, especially for Calculus and Gingivitis. The results indicate that deep learning models are very effective for automating dental disease detection and InceptionV3 yields a higher robustness. It also highlights the importance of this study in the utilization of state-of-the-art neural networks, not only to enhance the accuracy but also to speed up the diagnostics of oral diseases, perhaps opening new models for integration into clinical practice and supporting dental professionals at early detection and treatment.