Transfer Learning-based Fine Tuned MobileNetV2 Model with Explainable Artificial Intelligence for Identifying Dental Diseases

This study presents a pioneering approach to identifying and categorizing six prevalent dental diseases—calculus, caries, gingivitis, hypodontia, mouth ulcers, and tooth discoloration using transfer learning-based deep learning models enhanced with explainable artificial intelligence (XAI). To support medical professionals and address significant diagnostic challenges, we evaluated four pre-trained models: DenseNet169, InceptionV3, MobileNetV2, and ResNet50V2. Model efficacy was assessed through various metrics, including accuracy, precision, recall, f1-score, and confusion matrix. Our findings indicate that MobileNetV2 excelled, achieving the highest accuracy rate of 89.02%, with a precision of 89.45%, recall of 89.02%, and an F1-score of 89.13%. To enhance model interpretability and trust, we employed LIME and Grad-Cam techniques, which provided valuable insights into model predictions by highlighting critical areas of focus. This comprehensive and rigorous analysis not only advances multiclass image classification for oral diseases but also strengthens the reliability of the results. The integration of XAI techniques ensures that the models are transparent and trustworthy, thereby enhancing the confidence of both dentists and patients in the diagnostic process. The proposed work significantly outperforms existing works, providing a reliable and interpretable solution for dental diseases classification.