Texture features-based automated classification for dental caries level images

Abstrak

Dental caries is a globally prevalent oral health issue posing substantial challenges regarding health outcomes and economic burden. Early detection is critical to prevent the progression of the disease and ensure effective treatment. This study aims to develop a machine learning-based system for classifying dental caries severity using X-ray radiographic images. The proposed system integrates two prominent feature extraction techniques: Histogram of Oriented Gradients (HOG) and Haar Wavelet Transform, applied at varying levels (HOG 50×50, HOG 70×70, Haar Level 1, and Haar Level 2) to capture both texture and frequency-based features. These extracted features are subsequently classified using two machine learning algorithms, Support Vector Machines (SVM) and k-Nearest Neighbors (KNN), across four models: Cubic SVM, Quadratic SVM, Weighted KNN, and Fine KNN. A dataset of 347 dental X-ray images was expanded to 1,388 through augmentation techniques and pre-processed into grayscale for consistency. The results unveiled that combining Haar Wavelet features with the KNN classifier yielded the highest classification accuracy, reaching 97.99% during training and an AUC of 0.99. These findings underscore the potential of combining advanced feature extraction methods with robust machine learning algorithms to enhance the precision of dental caries detection in clinical practice. This system presents a significant step forward in automating diagnostic procedures, providing a reliable and efficient tool for early caries detection, ultimately contributing to improved patient outcomes.