Detection of Caries and Hypodontia using Deep Learning and Explainable AI

In today's world of technology breakthroughs, incorporating cutting-edge procedures has become essential, especially in healthcare, where an early and precise diagnosis is critical for improving patient outcomes and saving lives. Dental caries is an infectious bacterial disease primarily caused by the interaction of oral bacteria with dietary sugars, leading to plaque formation on teeth. Millions of people worldwide are facing this disease. Hypodontia is another dental term that healthcare providers use to describe congenitally missing teeth. We employed an AI-driven healthcare solution by utilizing a transformer-based image classification model, Vision Transformer (VIT), for the detection of dental caries and hypodontia. A dataset containing both periapical and panoramic dental x-rays utilized to gain maximum insights into teeth. The dataset was divided into 60% for training, 20% for testing, and 20% for validation. Data augmentation techniques are applied to enhance the diversity of training data and mitigate overfitting. The proposed model demonstrated remarkable performance and achieved an accuracy of 96.63% with evaluation matrices Precision, Recall, and F1-score of 96.60, 96.63, and 96.57 % respectively. Unlike Convolutional Neural Networks (CNN) which are black boxes, we utilized the Explainable Artificial Intelligence (XAI) models such as LIME, SHAP, and GRAD-CAM to gain insights into the inner workings of the VIT model. The XAI models contributed to fostering trust among medical dentists and medical professionals.