Claim Missing Document
Check
Articles

Found 1 Documents
Search
Journal : Jurnal Teknik Informatika (JUTIF)

CLASSIFICATION OF DENTAL CARIES DISEASE IN TOOTH IMAGES USING A COMPARISON OF EFFICIENTNET-B0, MOBILENETV2, RESNET-50, INCEPTIONV3 ARCHITECTURES Wahyuningsih, Wahyuningsih; Nugraha, Gibran Satya; Dwiyansaputra, Ramaditia
Jurnal Teknik Informatika (Jutif) Vol. 5 No. 4 (2024): JUTIF Volume 5, Number 4, August 2024 - SENIKO
Publisher : Informatika, Universitas Jenderal Soedirman

Show Abstract | Download Original | Original Source | Check in Google Scholar | DOI: 10.52436/1.jutif.2024.5.4.2187

Abstract

Dental caries is a global metabolic disorder, influenced by complex interactions between the body and microbes, it's caused by prolonged exposure to a low pH environment, leading to demineralized carious lesions. If untreated, it can cause pain and eating difficulties, requiring emergency care and significantly impacting overall quality of life. Diagnosis process can be conducted through physical assessment and analyzing laboratory testing. Image-based artificial intelligence systems, particularly the EfficientNet-B0 model, is suggested as a resolution for classifying dental caries using tooth images. The study's goal is to assess EfficientNet-B0's performance in comparison to other CNN architectures and play a role in advancing medical image classification technology. The original dataset comprising 1554 images was initially collected. After augmentation, the dataset expanded to 6348 images. The data was then divided into three subsets of training, validation, and testing datasets with a distribution ratio of 70:15:15, respectively. From all the evaluated models, the EfficientNet-B0 demonstrated a quite commendable accuracy of 0.98% with overfitting tolerance of less than 2%. Having the same accuracy as the MobileNetV2 (0.98%). Despite its inability to exceed the accuracy achieved by ResNet-50 (0.99%), EfficientNet-B0 accomplished its accuracy level with roughly a quarter of the parameters than ResNet-50 and highger than InceptionV3 (0.97%), highlighting its efficiency in parameter utilization and computational resource management. These findings hold promise for enhancing models and guiding clinical decision-making.