AI-driven Classification of Dental Restorations: CNN and KNN Techniques

The Summary encapsulates the results and methodologies of a research project utilizing CNN and KNN techniques to forecast various types of dental restoration. Precision, recall, and F1-Score metrics demonstrate outstanding accuracy across all categories: Amalgam (Precision: 98.21%, Recall: 98.21%, F1-Score: 98.21%), Composite (Precision: 98.26%, Recall: 98.26%, F1-Score: 98.26%), and Metal-Ceramic (Precision: 98.24%, Recall: 98.24%, F1-Score: 98.24%). These metrics confirm the models’ ability to accurately identify each type of dental restoration. The support values indicate a well-balanced dataset: Amalgam (3360 samples), Composite (3450 samples), and Metal-Ceramic (3410 samples), ensuring comprehensive coverage during training and evaluation. The model was trained on a dataset of 10,220 photos, and it obtained an accuracy of 98.23%. The overall accuracy of 99% underscores the models’ effectiveness in classification tasks, suggesting their potential for practical implementation in dental healthcare settings. Additionally, the study utilized a hybrid CNN-KNN approach with specific architectural details: 4 convolutional layers, 4 max-pooling layers, and 2 fully connected layers. This robust model architecture facilitated strong feature extraction and classification, contributing to the high precision and recall observed. In summary, the Abstract offers a concise overview of the research findings, highlighting the models’ high accuracy in distinguishing between Amalgam, Composite, and Metal-Ceramic dental restorations. The results indicate promising applications in improving diagnostic accuracy and treatment planning in dental practice, demonstrating the usefulness of machine learning in advancing dental healthcare.