UNet Architecture Based Dental Panoramic Image Segmentation

Abstract

This paper proposes an UNet architecture that uses convolutional neural networks to achieve accurate segmentation of Dental panoramic x-ray images. In dentistry, Radiographic images help medical experts to identify and diagnose the disease in an accurate manner X-rays, Computed Tomography (CT), Magnetic resonance imaging (MRI) are some of the radiographic images. Generally, X-ray images are complex in nature. The presence of noise results in lack of reliable separation between the various parts of teeth. This makes the segmentation process very difficult. Dental image segmentation helps the dentist to detect the impacted teeth, find the accurate position for the placement of dental implants and determine the orientation of teeth structure. UNet architecture model is a recent approach used for medical image segmentation. In this paper we took an advantage of the UNet architecture for dental x-ray image segmentation and achieved an accuracy of 97 % and Dice score of 94 %. Also, the performance of UNet architecture for dental x-ray image segmentation is compared with other image segmentation algorithms.