Classification of caries level using image processing and deep learning methods on dental images

The acids produced by the bacteria in the mouth cause dental caries by disrupting the mineral tissue of the teeth. Over time, bacteria accumulate in the mouth of people and these bacteria combine with food. When adequate care is not given to oral care, tooth enamel is damaged and tooth decay occurs. Until recently, infected teeth were extracted in order not to damage other areas. However, thanks to developing imaging techniques, the number and success rate of treatments applied without tooth extraction has increased rapidly. Tooth decay affects people's lives negatively. In this study, it is aimed to quickly detect dental caries on x-ray images and to minimize the tooth loss of the patients. Treatment methods applied using dental images and deep learning models are divided into filling, canal and bridge classes. In addition, it is aimed to increase the performance of deep learning models by applying the Luv-v channel and adaptive histogram equalization process as a preprocessing to tooth images. After segmentation processes on the data set consisting of 553 dental x-rays, experimental studies were carried out with Faster R-CNN and Yolov5 models, which are deep learning models. As a result of the experimental studies; while Faster R-CNN reached %86.7 accuracy, Yolov5 model achieved %92.7 accuracy. The decision support system obtained as a result of image processing and application of the Yolov5 hybrid model can be used in dental clinics.