Maulisa Oktiana, Maya Fitria, Hafidh Habibie, Khairun Saddami, Yasmina Elma, Handika Rahayu, Subhan Janura, Rizkika Putri, Rizki Novita
Dental caries is a disease that can damage the tooth structure, leading to cavities. However, the symptoms of decay are often unnoticed until a toothache develops, which can lead to cavities. Severe cases of dental caries, which cause cavities, require direct inspection. With early diagnosis of decay, preventive measures and treatment steps can be taken sooner. This helps prevent further development of tooth damage and reduces the need for more invasive restorative treatments. Therefore, this study develops a classification model for dental caries using Convolutional Neural Network architectures, namely Residual Network (ResNet)-50, EfficientNet B4, and MobileNet V2, where images are classified into two classes: decayed and normal teeth. The dataset comprises 700 dental images. Model evaluation includes accuracy, precision, recall, and F1-score. Experimental results indicate that the EfficientNet B4 model outperforms MobileNet V2 and Residual Network (ResNet)-50 architectures in classifying dental caries. Specifically, EfficientNet B4, trained on the lower occlusal dataset with a learning rate of 10⁻⁵, batch size of 16, and 200 epochs, achieves an accuracy of 77%, precision of 78%, recall of 77%, and F1-score of 77%. When the model is trained using a learning rate of 10⁻⁴, it achieves a better accuracy of 81%. The superiorperformance of EfficientNet B4 comes from its depthwise separable convolutions and squeeze-and-excitation (SE) blocks, which improve how features are extracted. These results show the promise of deep learning models in identifying dental cavities, leading to better diagnostics for oral health. However, an expanding dataset diversity is needed in further research for model generalization enhancement, aiming to ensure model’s robustness to apply in clinical settings. © 2026, Politeknik Negeri Padang. All rights reserved.
Department of Electrical and Computer Engineering, Universitas Syiah Kuala, Banda Aceh, Indonesia; Polyclinic of Dental and Oral, Regional General Hospital dr. Zainoel Abidin (RSUDZA), Banda Aceh, Indonesia; Faculty of Dentistry, Universitas Syiah Kuala, Banda Aceh, Indonesia
Research at a Glance
Register to unlock
Topics & SDG Alignment
Register to unlock
Collaboration
Register to unlock
Author Profile (Selected)
Register to unlock
References Overview
Register to unlock
Journal & Source
Register to unlock
Metadata & Integrity
Register to unlock