Automated caries detection with smartphone color photography using machine learning

Untreated caries is significant problem that affected billion people over the world. Therefore, appropriate method and accuracy of detection in clinical decision-making dental practices as well oral epidemiology or research, are required urgently. The aim this study was to introduce a computational algorithm can automate recognize carious lesions on tooth occlusal surfaces smartphone images according International Caries Detection Assessment System (ICDAS). From group extracted teeth, 620 unrestored molars/premolars were photographed using smartphone. obtained evaluated for diagnosis with ICDAS II codes, labeled into three classes: “No Surface Change” ( NSC); “Visually Non-Cavitated” VNC); “Cavitated” C). Then, two steps scheme Support Vector Machine (SVM) has been proposed: “ C versus (VNC + NSC)” classification, VNC NSC” classification. accuracy, sensitivity, specificity best model 92.37%, 88.1%, 96.6% NSC),” whereas they 83.33%, 82.2%, 66.7% NSC.” Although proposed SVM system further improvement verification, data only imaged from smartphone, it performed an auspicious potential diagnostics reasonable minimal cost.