ANALYSIS OF DENTAL CARIES USING GRAY-LEVEL CO- OCCURRENCE MATRIX ON PERIAPICAL RADIOGRAPHIC IMAGES

• Main Author: Aries Putra, Andika
• Format: Theses
• Language: Indonesia
• Online Access: https://digilib.itb.ac.id/gdl/view/86648
• Institution: Institut Teknologi Bandung

Dental caries is one of the most prevalent chronic diseases worldwide, especially in Indonesia, where its prevalence reached 88.8% in 2018. Dental caries is caused by acid produced by bacteria in plaque, which damages tooth enamel and can lead to serious infections and tooth loss if left untreated. One method for early caries detection is through periapical radiographic imaging. Dentists typically identify carious teeth manually. This study develops an automated model for dental caries detection. The technique utilizes the k-Nearest Neighbours (KNN) algorithm with cubic interpolation, employing Gray-Level Co-occurrence Matrix (GLCM) features at a distance of 10 pixels and an image size of 20x20. The study's results show that the KNN model demonstrates excellent performance with high accuracy, both on validation and test data, and an Area Under the Curve (AUC) value indicating a strong ability to differentiate between healthy and carious teeth images. Additional evaluation metrics, including precision, recall, F1-score, and specificity, further indicate the model's effectiveness in identifying positive and negative samples with a low error rate. The most influential GLCM aspect in this analysis was dissimilarity, which successfully captured significant intensity variations between neighbouring pixels, making it a key feature in improving classification accuracy and model effectiveness. This research concludes that the combination of GLCM, with a focus on dissimilarity, and the KNN algorithm can be an effective approach in detecting dental caries, contributing significantly to enhancing early diagnosis and clinical treatment of caries.