Automatic tooth instance segmentation and identification from panoramic X-Ray images using deep CNN

Due to deep learning's success in many medical applications, it is regarded as a significant technique for use in the area of dentistry, where several panoramic X-ray images are frequently used to assess oral health and identify disorders that damage the teeth and bones (e.g., cavities). In order to give diagnostic information for the management of dental problems and diseases, automatic segmentation is a crucial role in medical image processing and analysis. There are just a few datasets for panoramic radiograph images now accessible, according to the state of the art. For this endeavor we present a real dataset composed of 107 panoramic x-ray image were collected from two dental clinics and annotated. A Convolutional Neural Network (CNN) was trained using the annotated data for identification and instance segmentation. which enables the development of deep learning-based automatic tooth detection systems to segment teeth and assess oral status. Mask-RCNN is a deep CNN model for object detection, object localization, and object instance segmentation of medical images. In this paper, we demonstrate that Mask-RCNN can be used to perform highly effective and efficient automatic segmentation and identification of the tooth from panoramic radiographs images. The performance of the implemented networks is 90% of mean average precision (mAP), 63% of F1-sores and 96% of precision.